{ "schemaVersion": 3, "recordingCount": 21, "episodeCount": 40, "rubricSha256": "ea4b6dd9b099318d454d62e8650f27595f97426d34c4515bf34f1efc0696e48a", "method": "Three AI reviewers inspected synchronized top-down and wrist camera frames. Whole-episode overview sampling was followed by denser event-window inspection; ambiguous windows were decoded at native resolution and selected cases received a second review. Source summaries and telemetry locate events but do not determine scores. The default extraction is approximately 2.5 Hz, so brief events between frames and occluded contact remain limitations.", "timeBasis": "Seconds from each episode rollout start; acquisition ends after a retained lift of at least one second. Subsequent losses belong to transportation, including before lateral carry.", "scope": "The 40 evaluation episodes listed in the episode ledger. Development, continuation, and unlisted runs are excluded.", "progressSource": "Operator-assigned task milestones on a 0–100 scale. Progress is separate from execution quality and completion; 100 indicates that the brick is inside the box, without requiring release and rest.", "usageSource": "analysis/per_episode.json: per-episode model cost, calls, and tokens; excludes development. Total tokens include cached input, uncached input, and output. Reasoning tokens are a subset of output, not an additional chargeable total.", "lengthDefinition": "Recorded episode duration in seconds. Release times are reported separately for completed placements; incomplete episodes have no observed qualifying release.", "inventory": [ { "id": "enpire-test-0__20260919-233406", "source": "enpire-test-0/yam_red_lego_in_dark_gray_box/20260919-233406/recording.rrd", "bytes": 13232472837, "episodes": [ "ep_0000", "ep_0001", "ep_0002", "ep_0003", "ep_0004", "ep_0005", "ep_0006", "ep_0007", "ep_0008", "ep_0009", "ep_0010", "ep_0011", "ep_0012", "ep_0013", "ep_0014", "ep_0015", "ep_0016", "ep_0017", "ep_0018", "ep_0019" ], "task": "yam_red_lego_in_dark_gray_box", "summarySha256": "cbd56c044080533e86bec4f344afa71dd65de36bb89c770ccdee417a19ec9d63" }, { "id": "se3labs-eval-rac__20260920-211500", "source": "se3labs-eval-rac/yam_red_lego_in_dark_gray_box/20260920-211500/recording.rrd", "bytes": 455233205, "episodes": [ "ep_0000" ], "task": "yam_red_lego_in_dark_gray_box", "summarySha256": "6f76df208db2868560fd10552d0014c389ce0249a6ecc2a960b477b053a618f3" }, { "id": "se3labs-eval-rac__20260920-211841", "source": "se3labs-eval-rac/yam_red_lego_in_dark_gray_box/20260920-211841/recording.rrd", "bytes": 807042166, "episodes": [ "ep_0000" ], "task": "yam_red_lego_in_dark_gray_box", "summarySha256": "7afdcac8d93f1faf4e90186486ce0d13dc59f67630af3a65cd94e9a1c207cb94" }, { "id": "se3labs-eval-rac__20260920-212446", "source": "se3labs-eval-rac/yam_red_lego_in_dark_gray_box/20260920-212446/recording.rrd", "bytes": 632660965, "episodes": [ "ep_0000" ], "task": "yam_red_lego_in_dark_gray_box", "summarySha256": "16393e4f3f3b4f1c3a035b8401aca1df565a0bead0eda45e224662db7302d811" }, { "id": "se3labs-eval-rac__20260920-213042", "source": "se3labs-eval-rac/yam_red_lego_in_dark_gray_box/20260920-213042/recording.rrd", "bytes": 1043001179, "episodes": [ "ep_0000" ], "task": "yam_red_lego_in_dark_gray_box", "summarySha256": "95e2e1ea3b83df4817788a552627bc185a1ec7f5039ab0bfd4b2afc149ffcb59" }, { "id": "se3labs-eval-rac__20260920-214242", "source": "se3labs-eval-rac/yam_red_lego_in_dark_gray_box/20260920-214242/recording.rrd", "bytes": 672808574, "episodes": [ "ep_0000" ], "task": "yam_red_lego_in_dark_gray_box", "summarySha256": "6e640a1716dce40f0ef50ec1c828a4cec8272d2039aef8f43b8301e7c35194f4" }, { "id": "se3labs-eval-rac__20260920-214816", "source": "se3labs-eval-rac/yam_red_lego_in_dark_gray_box/20260920-214816/recording.rrd", "bytes": 1043895106, "episodes": [ "ep_0000" ], "task": "yam_red_lego_in_dark_gray_box", "summarySha256": "94eb9bfb527d85fa1edb28a2867fc151a565ef2785801958c96dd66684bdd3ac" }, { "id": "se3labs-eval-rac__20260920-215546", "source": "se3labs-eval-rac/yam_red_lego_in_dark_gray_box/20260920-215546/recording.rrd", "bytes": 1155618715, "episodes": [ "ep_0000" ], "task": "yam_red_lego_in_dark_gray_box", "summarySha256": "a18c0b9b79fc776cf7adb724f6d3c764004ed6ae7e1058f66075a523ee2d431f" }, { "id": "se3labs-eval-rac__20260920-222019", "source": "se3labs-eval-rac/yam_red_lego_in_dark_gray_box/20260920-222019/recording.rrd", "bytes": 1133168297, "episodes": [ "ep_0000" ], "task": "yam_red_lego_in_dark_gray_box", "summarySha256": "efb4664c5b99600cea0859a61bb15c22384c421727e096b06d1d4622f60d4114" }, { "id": "se3labs-eval-rac__20260920-224459", "source": "se3labs-eval-rac/yam_red_lego_in_dark_gray_box/20260920-224459/recording.rrd", "bytes": 505546155, "episodes": [ "ep_0000" ], "task": "yam_red_lego_in_dark_gray_box", "summarySha256": "0f29dc1a0781f8422465d2253c1fa1c8ceb4e841a0923dafcd944d25f4f6340c" }, { "id": "se3labs-eval-rac__20260920-224900", "source": "se3labs-eval-rac/yam_red_lego_in_dark_gray_box/20260920-224900/recording.rrd", "bytes": 541843016, "episodes": [ "ep_0000" ], "task": "yam_red_lego_in_dark_gray_box", "summarySha256": "615d170f341dee82524f444512dae3a976ae14daac6b0f1839f4cb669af60a59" }, { "id": "se3labs-eval-rac__20260920-230058", "source": "se3labs-eval-rac/yam_red_lego_in_dark_gray_box/20260920-230058/recording.rrd", "bytes": 550602059, "episodes": [ "ep_0000" ], "task": "yam_red_lego_in_dark_gray_box", "summarySha256": "ab95a3ba4cb431d830f7060e1902d25f01a92a022cbf86e4f5cc5b46163cf194" }, { "id": "se3labs-eval-rac__20260920-230618", "source": "se3labs-eval-rac/yam_red_lego_in_dark_gray_box/20260920-230618/recording.rrd", "bytes": 467074569, "episodes": [ "ep_0000" ], "task": "yam_red_lego_in_dark_gray_box", "summarySha256": "fb4e47987cb7e6244b934e5983d376b14dbe8f77c02f179718d0156162d6f377" }, { "id": "se3labs-eval-rac__20260920-230929", "source": "se3labs-eval-rac/yam_red_lego_in_dark_gray_box/20260920-230929/recording.rrd", "bytes": 831278452, "episodes": [ "ep_0000" ], "task": "yam_red_lego_in_dark_gray_box", "summarySha256": "db16fd1d305803aaed96038736e55201374f1676775315682ecfca1f97a74483" }, { "id": "se3labs-eval-rac__20260920-231850", "source": "se3labs-eval-rac/yam_red_lego_in_dark_gray_box/20260920-231850/recording.rrd", "bytes": 561846355, "episodes": [ "ep_0000" ], "task": "yam_red_lego_in_dark_gray_box", "summarySha256": "3735f7c4dc8956be38853d09f9d638a41ba1002b662a858a443a90686c00b29a" }, { "id": "se3labs-eval-rac__20260920-232318", "source": "se3labs-eval-rac/yam_red_lego_in_dark_gray_box/20260920-232318/recording.rrd", "bytes": 550654259, "episodes": [ "ep_0000" ], "task": "yam_red_lego_in_dark_gray_box", "summarySha256": "002a7c39f9492dccccc27fec18721359a0a2f9cb1fdeda165b8b326cf2651207" }, { "id": "se3labs-eval-rac__20260920-232815", "source": "se3labs-eval-rac/yam_red_lego_in_dark_gray_box/20260920-232815/recording.rrd", "bytes": 673885455, "episodes": [ "ep_0000" ], "task": "yam_red_lego_in_dark_gray_box", "summarySha256": "263b93136cb65af08bec32e8ab7eb0d85fe899710f11e798340e06e4ed9b4dc3" }, { "id": "se3labs-eval-rac__20260920-233401", "source": "se3labs-eval-rac/yam_red_lego_in_dark_gray_box/20260920-233401/recording.rrd", "bytes": 590929871, "episodes": [ "ep_0000" ], "task": "yam_red_lego_in_dark_gray_box", "summarySha256": "29d13b9b7c4c5fde83ef4926d5c3fe4f79f3ac332b89fcd2047170559e6d0286" }, { "id": "se3labs-eval-rac__20260920-234503", "source": "se3labs-eval-rac/yam_red_lego_in_dark_gray_box/20260920-234503/recording.rrd", "bytes": 867486177, "episodes": [ "ep_0000" ], "task": "yam_red_lego_in_dark_gray_box", "summarySha256": "1a4e8398b671ff102832bd77986d2deaf3717d6a424510e2e360510f1a3179e6" }, { "id": "se3labs-eval-rac__20260920-235124", "source": "se3labs-eval-rac/yam_red_lego_in_dark_gray_box/20260920-235124/recording.rrd", "bytes": 840475727, "episodes": [ "ep_0000" ], "task": "yam_red_lego_in_dark_gray_box", "summarySha256": "6512872dbff05274f489cd2fbb8ae7141f7e0d5b385895f2c02d1c80d8bc4601" }, { "id": "se3labs-eval-rac__20260921-000118", "source": "se3labs-eval-rac/yam_red_lego_in_dark_gray_box/20260921-000118/recording.rrd", "bytes": 711650270, "episodes": [ "ep_0000" ], "task": "yam_red_lego_in_dark_gray_box", "summarySha256": "0488d3f9a6d6df4cb807bfc0bee75f3ef9de183a1af08ad2a504bf11e9204812" } ], "records": [ { "id": "enpire-test-0__20260919-233406__ep_0000", "run": "20260919-233406", "episode": "ep_0000", "approach": "code-as-policy", "cohort": "frozen-eval", "grasp": { "score": 30, "item": "First attempt: stable grasp and lift, without corrective grip adjustment.", "evidence": "First closure squares the target before lifting; its pose stays fixed in the fingers.", "timestamps": [ 80.8, 84, 91.2 ] }, "transport": { "score": 30, "item": "Release pose reached with no slipping, drops, or unintended contact.", "evidence": "Smooth carry crosses the right rim into the box opening by 128 s.", "timestamps": [ 104, 122, 128 ] }, "placement": { "score": 5, "item": "Held above the opening or inside the box, but never released.", "evidence": "The red target remains pinched fully above the opening through the last frame; no release.", "timestamps": [ 130, 177, 179.5 ] }, "complete": false, "attempts": 1, "drops": 0, "placementAttempts": 0, "confidence": "medium", "reviewNote": "Topdown and active right-wrist frames reviewed over the full rollout at 10 s intervals, grasp at 0.8 s, carry at 3 s and releases at 0.4 s. Final 177–179.5 s frames establish end state. Scores reflect visible behavior at this sampling resolution; brief unsampled contact cannot be excluded.", "reviewedEvidence": [ "enpire-test-0/yam_red_lego_in_dark_gray_box/20260919-233406/recording.rrd", "analysis/sheets/enpire-test-0__20260919-233406__ep_0000.jpg", "analysis/scoring-review/code-sheets/enpire-test-0__20260919-233406__ep_0000__overview.jpg", "analysis/scoring-review/code-sheets/enpire-test-0__20260919-233406__ep_0000__grasp.jpg", "analysis/scoring-review/code-sheets/enpire-test-0__20260919-233406__ep_0000__transfer.jpg" ], "total": 65, "reviewStatus": "scored", "source": "enpire-test-0/yam_red_lego_in_dark_gray_box/20260919-233406/recording.rrd", "durationSeconds": 180.033, "rolloutOriginSeconds": 1.446, "terminalCode": "COMPLETED", "reviewer": "code-policy", "ledgerLabel": "233406 / ep 00", "progress": 75, "ledgerIndex": 1, "modelCostUsd": 0.0, "modelCalls": 0, "activeArm": "right", "tokenUsage": { "total": 0, "uncachedInput": 0, "cachedInput": 0, "output": 0, "reasoning": 0 }, "video": { "mp4": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/episode-enpire-test-0__20260919-233406__ep_0000-10x.mp4", "webm": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/episode-enpire-test-0__20260919-233406__ep_0000-10x.webm", "poster": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/episode-enpire-test-0__20260919-233406__ep_0000-10x-poster.jpg", "speedup": 10, "width": 1280, "height": 360 }, "evidenceImage": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/evidence/enpire-test-0__20260919-233406__ep_0000.jpg", "episodeNumber": 1, "displayName": "Code-as-policy - episode 1" }, { "id": "enpire-test-0__20260919-233406__ep_0001", "run": "20260919-233406", "episode": "ep_0001", "approach": "code-as-policy", "cohort": "frozen-eval", "grasp": { "score": 30, "item": "First attempt: stable grasp and lift, without corrective grip adjustment.", "evidence": "First closure retains the target; lift shows no visible relative slip.", "timestamps": [ 82, 85.2, 92.4 ] }, "transport": { "score": 30, "item": "Release pose reached with no slipping, drops, or unintended contact.", "evidence": "Carries target from the left side of the box to a position inside its opening.", "timestamps": [ 108, 123.2, 129.2 ] }, "placement": { "score": 5, "item": "Held above the opening or inside the box, but never released.", "evidence": "Fingers remain closed on the target above the opening until timeout.", "timestamps": [ 130, 177, 179.5 ] }, "complete": false, "attempts": 1, "drops": 0, "placementAttempts": 0, "confidence": "medium", "reviewNote": "Topdown and active right-wrist frames reviewed over the full rollout at 10 s intervals, grasp at 0.8 s, carry at 3 s and releases at 0.4 s. Final 177–179.5 s frames establish end state. Scores reflect visible behavior at this sampling resolution; brief unsampled contact cannot be excluded.", "reviewedEvidence": [ "enpire-test-0/yam_red_lego_in_dark_gray_box/20260919-233406/recording.rrd", "analysis/sheets/enpire-test-0__20260919-233406__ep_0001.jpg", "analysis/scoring-review/code-sheets/enpire-test-0__20260919-233406__ep_0001__overview.jpg", "analysis/scoring-review/code-sheets/enpire-test-0__20260919-233406__ep_0001__grasp.jpg", "analysis/scoring-review/code-sheets/enpire-test-0__20260919-233406__ep_0001__transfer.jpg" ], "total": 65, "reviewStatus": "scored", "source": "enpire-test-0/yam_red_lego_in_dark_gray_box/20260919-233406/recording.rrd", "durationSeconds": 180.048, "rolloutOriginSeconds": 508.32, "terminalCode": "COMPLETED", "reviewer": "code-policy", "ledgerLabel": "233406 / ep 01", "progress": 100, "ledgerIndex": 2, "modelCostUsd": 0.0, "modelCalls": 0, "activeArm": "right", "tokenUsage": { "total": 0, "uncachedInput": 0, "cachedInput": 0, "output": 0, "reasoning": 0 }, "video": { "mp4": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/episode-enpire-test-0__20260919-233406__ep_0001-10x.mp4", "webm": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/episode-enpire-test-0__20260919-233406__ep_0001-10x.webm", "poster": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/episode-enpire-test-0__20260919-233406__ep_0001-10x-poster.jpg", "speedup": 10, "width": 1280, "height": 360 }, "evidenceImage": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/evidence/enpire-test-0__20260919-233406__ep_0001.jpg", "episodeNumber": 2, "displayName": "Code-as-policy - episode 2" }, { "id": "enpire-test-0__20260919-233406__ep_0002", "run": "20260919-233406", "episode": "ep_0002", "approach": "code-as-policy", "cohort": "frozen-eval", "grasp": { "score": 0, "item": "No purposeful target contact, or only a distractor is grasped.", "evidence": "Arm moves to an observation pose, but the gripper stays open and never contacts the target.", "timestamps": [ 0, 10, 20, 179.5 ] }, "transport": { "score": 0, "item": "No transport beyond the initial lift, or no retained lift.", "evidence": "Target remains on the table; no acquired target is transported.", "timestamps": [ 20, 90, 179.5 ] }, "placement": { "score": 0, "item": "Finishes outside the box after a miss or bounce, or placement is not reached.", "evidence": "No placement; final target is outside the box.", "timestamps": [ 20, 177, 179.5 ] }, "complete": false, "attempts": 0, "drops": 0, "placementAttempts": 0, "confidence": "high", "reviewNote": "At rollout start the red target is still in the box from the preceding setup. It is moved onto the table within the first 10 s; scene setup overlaps the rollout. Do not count its initial in-box state as robot completion. Topdown and active right-wrist frames reviewed over the full rollout at 10 s intervals, grasp at 0.8 s, carry at 3 s and releases at 0.4 s. Final 177–179.5 s frames establish end state. Scores reflect visible behavior at this sampling resolution; brief unsampled contact cannot be excluded.", "reviewedEvidence": [ "enpire-test-0/yam_red_lego_in_dark_gray_box/20260919-233406/recording.rrd", "analysis/sheets/enpire-test-0__20260919-233406__ep_0002.jpg", "analysis/scoring-review/code-sheets/enpire-test-0__20260919-233406__ep_0002__overview.jpg" ], "total": 0, "reviewStatus": "scored", "source": "enpire-test-0/yam_red_lego_in_dark_gray_box/20260919-233406/recording.rrd", "durationSeconds": 180.047, "rolloutOriginSeconds": 720.675, "terminalCode": "COMPLETED", "reviewer": "code-policy", "ledgerLabel": "233406 / ep 02", "progress": 0, "ledgerIndex": 3, "modelCostUsd": 0.0, "modelCalls": 0, "activeArm": "right", "tokenUsage": { "total": 0, "uncachedInput": 0, "cachedInput": 0, "output": 0, "reasoning": 0 }, "video": { "mp4": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/episode-enpire-test-0__20260919-233406__ep_0002-10x.mp4", "webm": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/episode-enpire-test-0__20260919-233406__ep_0002-10x.webm", "poster": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/episode-enpire-test-0__20260919-233406__ep_0002-10x-poster.jpg", "speedup": 10, "width": 1280, "height": 360 }, "evidenceImage": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/evidence/enpire-test-0__20260919-233406__ep_0002.jpg", "episodeNumber": 3, "displayName": "Code-as-policy - episode 3" }, { "id": "enpire-test-0__20260919-233406__ep_0003", "run": "20260919-233406", "episode": "ep_0003", "approach": "code-as-policy", "cohort": "frozen-eval", "grasp": { "score": 30, "item": "First attempt: stable grasp and lift, without corrective grip adjustment.", "evidence": "The first closure retains the brick with a stable lift lasting several seconds.", "timestamps": [ 78.9, 82.1, 89.3 ] }, "transport": { "score": 30, "item": "Release pose reached with no slipping, drops, or unintended contact.", "evidence": "Brick remains fixed in fingers during translation and reaches inside the box opening.", "timestamps": [ 102.1, 117.1, 123.1 ] }, "placement": { "score": 5, "item": "Held above the opening or inside the box, but never released.", "evidence": "Robot holds the target over the box without opening the gripper.", "timestamps": [ 130, 177, 179.5 ] }, "complete": false, "attempts": 1, "drops": 0, "placementAttempts": 0, "confidence": "medium", "reviewNote": "Topdown and active right-wrist frames reviewed over the full rollout at 10 s intervals, grasp at 0.8 s, carry at 3 s and releases at 0.4 s. Final 177–179.5 s frames establish end state. Scores reflect visible behavior at this sampling resolution; brief unsampled contact cannot be excluded.", "reviewedEvidence": [ "enpire-test-0/yam_red_lego_in_dark_gray_box/20260919-233406/recording.rrd", "analysis/sheets/enpire-test-0__20260919-233406__ep_0003.jpg", "analysis/scoring-review/code-sheets/enpire-test-0__20260919-233406__ep_0003__overview.jpg", "analysis/scoring-review/code-sheets/enpire-test-0__20260919-233406__ep_0003__grasp.jpg", "analysis/scoring-review/code-sheets/enpire-test-0__20260919-233406__ep_0003__transfer.jpg" ], "total": 65, "reviewStatus": "scored", "source": "enpire-test-0/yam_red_lego_in_dark_gray_box/20260919-233406/recording.rrd", "durationSeconds": 180.047, "rolloutOriginSeconds": 945.152, "terminalCode": "COMPLETED", "reviewer": "code-policy", "ledgerLabel": "233406 / ep 03", "progress": 100, "ledgerIndex": 4, "modelCostUsd": 0.0, "modelCalls": 0, "activeArm": "right", "tokenUsage": { "total": 0, "uncachedInput": 0, "cachedInput": 0, "output": 0, "reasoning": 0 }, "video": { "mp4": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/episode-enpire-test-0__20260919-233406__ep_0003-10x.mp4", "webm": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/episode-enpire-test-0__20260919-233406__ep_0003-10x.webm", "poster": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/episode-enpire-test-0__20260919-233406__ep_0003-10x-poster.jpg", "speedup": 10, "width": 1280, "height": 360 }, "evidenceImage": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/evidence/enpire-test-0__20260919-233406__ep_0003.jpg", "episodeNumber": 4, "displayName": "Code-as-policy - episode 4" }, { "id": "enpire-test-0__20260919-233406__ep_0004", "run": "20260919-233406", "episode": "ep_0004", "approach": "code-as-policy", "cohort": "frozen-eval", "grasp": { "score": 30, "item": "First attempt: stable grasp and lift, without corrective grip adjustment.", "evidence": "Corner-oriented first grasp holds securely; orientation stays constant after lift.", "timestamps": [ 80.9, 84.1, 91.3 ] }, "transport": { "score": 5, "item": "Transfer started but release pose never reached; includes an unrecovered drop.", "evidence": "Carry starts toward the box but stalls with target visibly outside its right edge.", "timestamps": [ 101.1, 110.1, 113.1 ] }, "placement": { "score": 0, "item": "Finishes outside the box after a miss or bounce, or placement is not reached.", "evidence": "Target remains held outside the opening; placement is not reached.", "timestamps": [ 120, 177, 179.5 ] }, "complete": false, "attempts": 1, "drops": 0, "placementAttempts": 0, "confidence": "medium", "reviewNote": "The diagonal target orientation is present at initial contact and stays stable; it is not evidence of a later slip. Topdown and active right-wrist frames reviewed over the full rollout at 10 s intervals, grasp at 0.8 s, carry at 3 s and releases at 0.4 s. Final 177–179.5 s frames establish end state. Scores reflect visible behavior at this sampling resolution; brief unsampled contact cannot be excluded.", "reviewedEvidence": [ "enpire-test-0/yam_red_lego_in_dark_gray_box/20260919-233406/recording.rrd", "analysis/sheets/enpire-test-0__20260919-233406__ep_0004.jpg", "analysis/scoring-review/code-sheets/enpire-test-0__20260919-233406__ep_0004__overview.jpg", "analysis/scoring-review/code-sheets/enpire-test-0__20260919-233406__ep_0004__grasp.jpg", "analysis/scoring-review/code-sheets/enpire-test-0__20260919-233406__ep_0004__transfer.jpg" ], "total": 35, "reviewStatus": "scored", "source": "enpire-test-0/yam_red_lego_in_dark_gray_box/20260919-233406/recording.rrd", "durationSeconds": 180.035, "rolloutOriginSeconds": 1182.608, "terminalCode": "COMPLETED", "reviewer": "code-policy", "ledgerLabel": "233406 / ep 04", "progress": 50, "ledgerIndex": 5, "modelCostUsd": 0.0, "modelCalls": 0, "activeArm": "right", "tokenUsage": { "total": 0, "uncachedInput": 0, "cachedInput": 0, "output": 0, "reasoning": 0 }, "video": { "mp4": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/episode-enpire-test-0__20260919-233406__ep_0004-10x.mp4", "webm": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/episode-enpire-test-0__20260919-233406__ep_0004-10x.webm", "poster": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/episode-enpire-test-0__20260919-233406__ep_0004-10x-poster.jpg", "speedup": 10, "width": 1280, "height": 360 }, "evidenceImage": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/evidence/enpire-test-0__20260919-233406__ep_0004.jpg", "episodeNumber": 5, "displayName": "Code-as-policy - episode 5" }, { "id": "enpire-test-0__20260919-233406__ep_0005", "run": "20260919-233406", "episode": "ep_0005", "approach": "code-as-policy", "cohort": "frozen-eval", "grasp": { "score": 30, "item": "First attempt: stable grasp and lift, without corrective grip adjustment.", "evidence": "First grasp aligns the target during closure and lifts it securely.", "timestamps": [ 76.1, 79.3, 86.5 ] }, "transport": { "score": 30, "item": "Release pose reached with no slipping, drops, or unintended contact.", "evidence": "Target stays fixed in fingers along the carry and is fully over the opening before release.", "timestamps": [ 114.3, 126.3, 133.8 ] }, "placement": { "score": 40, "item": "Complete on the first placement attempt, without rim contact or correction.", "evidence": "Single release around 135 s lands inside with no visible rim contact; brick remains at rest through episode end.", "timestamps": [ 135.4, 138.2, 177, 179.5 ] }, "complete": true, "attempts": 1, "drops": 0, "placementAttempts": 1, "confidence": "medium", "reviewNote": "Topdown and active right-wrist frames reviewed over the full rollout at 10 s intervals, grasp at 0.8 s, carry at 3 s and releases at 0.4 s. Final 177–179.5 s frames establish end state. Scores reflect visible behavior at this sampling resolution; brief unsampled contact cannot be excluded.", "reviewedEvidence": [ "enpire-test-0/yam_red_lego_in_dark_gray_box/20260919-233406/recording.rrd", "analysis/sheets/enpire-test-0__20260919-233406__ep_0005.jpg", "analysis/scoring-review/code-sheets/enpire-test-0__20260919-233406__ep_0005__overview.jpg", "analysis/scoring-review/code-sheets/enpire-test-0__20260919-233406__ep_0005__grasp.jpg", "analysis/scoring-review/code-sheets/enpire-test-0__20260919-233406__ep_0005__transfer.jpg", "analysis/scoring-review/code-sheets/enpire-test-0__20260919-233406__ep_0005__release.jpg" ], "total": 100, "reviewStatus": "scored", "source": "enpire-test-0/yam_red_lego_in_dark_gray_box/20260919-233406/recording.rrd", "durationSeconds": 180.011, "rolloutOriginSeconds": 1427.727, "terminalCode": "COMPLETED", "reviewer": "code-policy", "ledgerLabel": "233406 / ep 05", "progress": 100, "ledgerIndex": 6, "modelCostUsd": 0.0, "modelCalls": 0, "activeArm": "right", "tokenUsage": { "total": 0, "uncachedInput": 0, "cachedInput": 0, "output": 0, "reasoning": 0 }, "video": { "mp4": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/episode-enpire-test-0__20260919-233406__ep_0005-10x.mp4", "webm": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/episode-enpire-test-0__20260919-233406__ep_0005-10x.webm", "poster": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/episode-enpire-test-0__20260919-233406__ep_0005-10x-poster.jpg", "speedup": 10, "width": 1280, "height": 360 }, "evidenceImage": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/evidence/enpire-test-0__20260919-233406__ep_0005.jpg", "episodeNumber": 6, "displayName": "Code-as-policy - episode 6" }, { "id": "enpire-test-0__20260919-233406__ep_0006", "run": "20260919-233406", "episode": "ep_0006", "approach": "code-as-policy", "cohort": "frozen-eval", "grasp": { "score": 30, "item": "First attempt: stable grasp and lift, without corrective grip adjustment.", "evidence": "Target is retained on the first closure and remains stable during the lift.", "timestamps": [ 78.3, 81.5, 88.7 ] }, "transport": { "score": 30, "item": "Release pose reached with no slipping, drops, or unintended contact.", "evidence": "The retained target travels into the opening without visible slip or drop.", "timestamps": [ 104.5, 116.5, 122.8 ] }, "placement": { "score": 40, "item": "Complete on the first placement attempt, without rim contact or correction.", "evidence": "Single release around 124 s; target settles fully inside, well clear of the rim, and stays at rest through end.", "timestamps": [ 124.4, 127.2, 177, 179.5 ] }, "complete": true, "attempts": 1, "drops": 0, "placementAttempts": 1, "confidence": "medium", "reviewNote": "Topdown and active right-wrist frames reviewed over the full rollout at 10 s intervals, grasp at 0.8 s, carry at 3 s and releases at 0.4 s. Final 177–179.5 s frames establish end state. Scores reflect visible behavior at this sampling resolution; brief unsampled contact cannot be excluded.", "reviewedEvidence": [ "enpire-test-0/yam_red_lego_in_dark_gray_box/20260919-233406/recording.rrd", "analysis/sheets/enpire-test-0__20260919-233406__ep_0006.jpg", "analysis/scoring-review/code-sheets/enpire-test-0__20260919-233406__ep_0006__overview.jpg", "analysis/scoring-review/code-sheets/enpire-test-0__20260919-233406__ep_0006__grasp.jpg", "analysis/scoring-review/code-sheets/enpire-test-0__20260919-233406__ep_0006__transfer.jpg", "analysis/scoring-review/code-sheets/enpire-test-0__20260919-233406__ep_0006__release.jpg" ], "total": 100, "reviewStatus": "scored", "source": "enpire-test-0/yam_red_lego_in_dark_gray_box/20260919-233406/recording.rrd", "durationSeconds": 180.027, "rolloutOriginSeconds": 1645.027, "terminalCode": "COMPLETED", "reviewer": "code-policy", "ledgerLabel": "233406 / ep 06", "progress": 100, "ledgerIndex": 7, "modelCostUsd": 0.0, "modelCalls": 0, "activeArm": "right", "tokenUsage": { "total": 0, "uncachedInput": 0, "cachedInput": 0, "output": 0, "reasoning": 0 }, "video": { "mp4": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/episode-enpire-test-0__20260919-233406__ep_0006-10x.mp4", "webm": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/episode-enpire-test-0__20260919-233406__ep_0006-10x.webm", "poster": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/episode-enpire-test-0__20260919-233406__ep_0006-10x-poster.jpg", "speedup": 10, "width": 1280, "height": 360 }, "evidenceImage": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/evidence/enpire-test-0__20260919-233406__ep_0006.jpg", "episodeNumber": 7, "displayName": "Code-as-policy - episode 7" }, { "id": "enpire-test-0__20260919-233406__ep_0007", "run": "20260919-233406", "episode": "ep_0007", "approach": "code-as-policy", "cohort": "frozen-eval", "grasp": { "score": 0, "item": "No purposeful target contact, or only a distractor is grasped.", "evidence": "Neither arm makes purposeful target contact; right gripper stays in its starting pose.", "timestamps": [ 0, 60, 120, 179.5 ] }, "transport": { "score": 0, "item": "No transport beyond the initial lift, or no retained lift.", "evidence": "Target remains on the table throughout; no lift or transfer.", "timestamps": [ 0, 90, 179.5 ] }, "placement": { "score": 0, "item": "Finishes outside the box after a miss or bounce, or placement is not reached.", "evidence": "No placement; target is outside the box in final topdown and wrist views.", "timestamps": [ 0, 177, 179.5 ] }, "complete": false, "attempts": 0, "drops": 0, "placementAttempts": 0, "confidence": "high", "reviewNote": "Topdown and active right-wrist frames reviewed over the full rollout at 10 s intervals, grasp at 0.8 s, carry at 3 s and releases at 0.4 s. Final 177–179.5 s frames establish end state. Scores reflect visible behavior at this sampling resolution; brief unsampled contact cannot be excluded.", "reviewedEvidence": [ "enpire-test-0/yam_red_lego_in_dark_gray_box/20260919-233406/recording.rrd", "analysis/sheets/enpire-test-0__20260919-233406__ep_0007.jpg", "analysis/scoring-review/code-sheets/enpire-test-0__20260919-233406__ep_0007__overview.jpg" ], "total": 0, "reviewStatus": "scored", "source": "enpire-test-0/yam_red_lego_in_dark_gray_box/20260919-233406/recording.rrd", "durationSeconds": 180.046, "rolloutOriginSeconds": 1863.552, "terminalCode": "COMPLETED", "reviewer": "code-policy", "ledgerLabel": "233406 / ep 07", "progress": 0, "ledgerIndex": 8, "modelCostUsd": 0.0, "modelCalls": 0, "activeArm": "right", "tokenUsage": { "total": 0, "uncachedInput": 0, "cachedInput": 0, "output": 0, "reasoning": 0 }, "video": { "mp4": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/episode-enpire-test-0__20260919-233406__ep_0007-10x.mp4", "webm": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/episode-enpire-test-0__20260919-233406__ep_0007-10x.webm", "poster": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/episode-enpire-test-0__20260919-233406__ep_0007-10x-poster.jpg", "speedup": 10, "width": 1280, "height": 360 }, "evidenceImage": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/evidence/enpire-test-0__20260919-233406__ep_0007.jpg", "episodeNumber": 8, "displayName": "Code-as-policy - episode 8" }, { "id": "enpire-test-0__20260919-233406__ep_0008", "run": "20260919-233406", "episode": "ep_0008", "approach": "code-as-policy", "cohort": "frozen-eval", "grasp": { "score": 30, "item": "First attempt: stable grasp and lift, without corrective grip adjustment.", "evidence": "First closure produces a stable grasp and lift.", "timestamps": [ 82.4, 85.6, 92.8 ] }, "transport": { "score": 30, "item": "Release pose reached with no slipping, drops, or unintended contact.", "evidence": "Target travels across the right side of the box into the opening with no visible loss of grip.", "timestamps": [ 108.6, 120.6, 125.2 ] }, "placement": { "score": 40, "item": "Complete on the first placement attempt, without rim contact or correction.", "evidence": "Released once around 126.4 s; brick flips during the free fall, settles fully inside, and remains at rest. No rim strike or correction observed.", "timestamps": [ 126.8, 129.6, 177, 179.5 ] }, "complete": true, "attempts": 1, "drops": 0, "placementAttempts": 1, "confidence": "medium", "reviewNote": "Rotation after intentional release is not a grasp instability or transport drop. Topdown and active right-wrist frames reviewed over the full rollout at 10 s intervals, grasp at 0.8 s, carry at 3 s and releases at 0.4 s. Final 177–179.5 s frames establish end state. Scores reflect visible behavior at this sampling resolution; brief unsampled contact cannot be excluded.", "reviewedEvidence": [ "enpire-test-0/yam_red_lego_in_dark_gray_box/20260919-233406/recording.rrd", "analysis/sheets/enpire-test-0__20260919-233406__ep_0008.jpg", "analysis/scoring-review/code-sheets/enpire-test-0__20260919-233406__ep_0008__overview.jpg", "analysis/scoring-review/code-sheets/enpire-test-0__20260919-233406__ep_0008__grasp.jpg", "analysis/scoring-review/code-sheets/enpire-test-0__20260919-233406__ep_0008__transfer.jpg", "analysis/scoring-review/code-sheets/enpire-test-0__20260919-233406__ep_0008__release.jpg" ], "total": 100, "reviewStatus": "scored", "source": "enpire-test-0/yam_red_lego_in_dark_gray_box/20260919-233406/recording.rrd", "durationSeconds": 180.061, "rolloutOriginSeconds": 2091.728, "terminalCode": "COMPLETED", "reviewer": "code-policy", "ledgerLabel": "233406 / ep 08", "progress": 100, "ledgerIndex": 9, "modelCostUsd": 0.0, "modelCalls": 0, "activeArm": "right", "tokenUsage": { "total": 0, "uncachedInput": 0, "cachedInput": 0, "output": 0, "reasoning": 0 }, "video": { "mp4": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/episode-enpire-test-0__20260919-233406__ep_0008-10x.mp4", "webm": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/episode-enpire-test-0__20260919-233406__ep_0008-10x.webm", "poster": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/episode-enpire-test-0__20260919-233406__ep_0008-10x-poster.jpg", "speedup": 10, "width": 1280, "height": 360 }, "evidenceImage": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/evidence/enpire-test-0__20260919-233406__ep_0008.jpg", "episodeNumber": 9, "displayName": "Code-as-policy - episode 9" }, { "id": "enpire-test-0__20260919-233406__ep_0009", "run": "20260919-233406", "episode": "ep_0009", "approach": "code-as-policy", "cohort": "frozen-eval", "grasp": { "score": 30, "item": "First attempt: stable grasp and lift, without corrective grip adjustment.", "evidence": "Stable first grasp and clear lift are visible for more than 1 s.", "timestamps": [ 89.2, 92.4, 97.2 ] }, "transport": { "score": 5, "item": "Transfer started but release pose never reached; includes an unrecovered drop.", "evidence": "After lift, target moves a short distance laterally at 97–101 s but never reaches the box opening.", "timestamps": [ 97, 99, 101 ] }, "placement": { "score": 0, "item": "Finishes outside the box after a miss or bounce, or placement is not reached.", "evidence": "The target remains held over the table at end; no release or placement.", "timestamps": [ 110, 177, 179.5 ] }, "complete": false, "attempts": 1, "drops": 0, "placementAttempts": 0, "confidence": "medium", "reviewNote": "Stage boundary reviewed by a second assessor: approximately 27 mm lateral motion accompanies final lift settling. Controller evidence in analysis/taxonomy.json identifies transport-pose settling failure, supporting transfer started (5) rather than lift-only (0). Topdown and active right-wrist frames reviewed over the full rollout at 10 s intervals, grasp at 0.8 s, carry at 3 s and releases at 0.4 s. Final 177–179.5 s frames establish end state. Scores reflect visible behavior at this sampling resolution; brief unsampled contact cannot be excluded.", "reviewedEvidence": [ "enpire-test-0/yam_red_lego_in_dark_gray_box/20260919-233406/recording.rrd", "analysis/sheets/enpire-test-0__20260919-233406__ep_0009.jpg", "analysis/scoring-review/code-sheets/enpire-test-0__20260919-233406__ep_0009__overview.jpg", "analysis/scoring-review/code-sheets/enpire-test-0__20260919-233406__ep_0009__grasp.jpg", "analysis/scoring-review/code-sheets/enpire-test-0__20260919-233406__ep_0009__transfer.jpg", "analysis/taxonomy.json" ], "total": 35, "reviewStatus": "scored", "source": "enpire-test-0/yam_red_lego_in_dark_gray_box/20260919-233406/recording.rrd", "durationSeconds": 180.047, "rolloutOriginSeconds": 2323.51, "terminalCode": "COMPLETED", "reviewer": "code-policy", "ledgerLabel": "233406 / ep 09", "progress": 50, "ledgerIndex": 10, "modelCostUsd": 0.0, "modelCalls": 0, "activeArm": "right", "tokenUsage": { "total": 0, "uncachedInput": 0, "cachedInput": 0, "output": 0, "reasoning": 0 }, "video": { "mp4": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/episode-enpire-test-0__20260919-233406__ep_0009-10x.mp4", "webm": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/episode-enpire-test-0__20260919-233406__ep_0009-10x.webm", "poster": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/episode-enpire-test-0__20260919-233406__ep_0009-10x-poster.jpg", "speedup": 10, "width": 1280, "height": 360 }, "evidenceImage": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/evidence/enpire-test-0__20260919-233406__ep_0009.jpg", "episodeNumber": 10, "displayName": "Code-as-policy - episode 10" }, { "id": "enpire-test-0__20260919-233406__ep_0010", "run": "20260919-233406", "episode": "ep_0010", "approach": "code-as-policy", "cohort": "frozen-eval", "grasp": { "score": 30, "item": "First attempt: stable grasp and lift, without corrective grip adjustment.", "evidence": "Target aligns during the initial closure, then remains fixed while lifting.", "timestamps": [ 84.5, 87.7, 94.9 ] }, "transport": { "score": 30, "item": "Release pose reached with no slipping, drops, or unintended contact.", "evidence": "Carry includes pauses but no visible slip, contact or drop; target reaches opening.", "timestamps": [ 113.7, 140.7, 152.7 ] }, "placement": { "score": 40, "item": "Complete on the first placement attempt, without rim contact or correction.", "evidence": "One release around 154.7 s lands cleanly inside; brick stays at rest through 179.5 s.", "timestamps": [ 155.1, 157.9, 177, 179.5 ] }, "complete": true, "attempts": 1, "drops": 0, "placementAttempts": 1, "confidence": "medium", "reviewNote": "Pauses and repeated visual checks are not penalties. Topdown and active right-wrist frames reviewed over the full rollout at 10 s intervals, grasp at 0.8 s, carry at 3 s and releases at 0.4 s. Final 177–179.5 s frames establish end state. Scores reflect visible behavior at this sampling resolution; brief unsampled contact cannot be excluded.", "reviewedEvidence": [ "enpire-test-0/yam_red_lego_in_dark_gray_box/20260919-233406/recording.rrd", "analysis/sheets/enpire-test-0__20260919-233406__ep_0010.jpg", "analysis/scoring-review/code-sheets/enpire-test-0__20260919-233406__ep_0010__overview.jpg", "analysis/scoring-review/code-sheets/enpire-test-0__20260919-233406__ep_0010__grasp.jpg", "analysis/scoring-review/code-sheets/enpire-test-0__20260919-233406__ep_0010__transfer.jpg", "analysis/scoring-review/code-sheets/enpire-test-0__20260919-233406__ep_0010__release.jpg" ], "total": 100, "reviewStatus": "scored", "source": "enpire-test-0/yam_red_lego_in_dark_gray_box/20260919-233406/recording.rrd", "durationSeconds": 180.082, "rolloutOriginSeconds": 2555.43, "terminalCode": "COMPLETED", "reviewer": "code-policy", "ledgerLabel": "233406 / ep 10", "progress": 100, "ledgerIndex": 11, "modelCostUsd": 0.0, "modelCalls": 0, "activeArm": "right", "tokenUsage": { "total": 0, "uncachedInput": 0, "cachedInput": 0, "output": 0, "reasoning": 0 }, "video": { "mp4": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/episode-enpire-test-0__20260919-233406__ep_0010-10x.mp4", "webm": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/episode-enpire-test-0__20260919-233406__ep_0010-10x.webm", "poster": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/episode-enpire-test-0__20260919-233406__ep_0010-10x-poster.jpg", "speedup": 10, "width": 1280, "height": 360 }, "evidenceImage": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/evidence/enpire-test-0__20260919-233406__ep_0010.jpg", "episodeNumber": 11, "displayName": "Code-as-policy - episode 11" }, { "id": "enpire-test-0__20260919-233406__ep_0011", "run": "20260919-233406", "episode": "ep_0011", "approach": "code-as-policy", "cohort": "frozen-eval", "grasp": { "score": 30, "item": "First attempt: stable grasp and lift, without corrective grip adjustment.", "evidence": "First grasp holds the target stably through lift.", "timestamps": [ 89, 92.2, 99.4 ] }, "transport": { "score": 30, "item": "Release pose reached with no slipping, drops, or unintended contact.", "evidence": "Target is carried into the box opening with its pose fixed between the fingers.", "timestamps": [ 112.2, 127.2, 131.6 ] }, "placement": { "score": 40, "item": "Complete on the first placement attempt, without rim contact or correction.", "evidence": "Single release around 132.8 s lands fully inside with no visible rim collision; final rest exceeds 2 s.", "timestamps": [ 133.2, 136, 177, 179.5 ] }, "complete": true, "attempts": 1, "drops": 0, "placementAttempts": 1, "confidence": "medium", "reviewNote": "Topdown and active right-wrist frames reviewed over the full rollout at 10 s intervals, grasp at 0.8 s, carry at 3 s and releases at 0.4 s. Final 177–179.5 s frames establish end state. Scores reflect visible behavior at this sampling resolution; brief unsampled contact cannot be excluded.", "reviewedEvidence": [ "enpire-test-0/yam_red_lego_in_dark_gray_box/20260919-233406/recording.rrd", "analysis/sheets/enpire-test-0__20260919-233406__ep_0011.jpg", "analysis/scoring-review/code-sheets/enpire-test-0__20260919-233406__ep_0011__overview.jpg", "analysis/scoring-review/code-sheets/enpire-test-0__20260919-233406__ep_0011__grasp.jpg", "analysis/scoring-review/code-sheets/enpire-test-0__20260919-233406__ep_0011__transfer.jpg", "analysis/scoring-review/code-sheets/enpire-test-0__20260919-233406__ep_0011__release.jpg" ], "total": 100, "reviewStatus": "scored", "source": "enpire-test-0/yam_red_lego_in_dark_gray_box/20260919-233406/recording.rrd", "durationSeconds": 180.013, "rolloutOriginSeconds": 2793.191, "terminalCode": "COMPLETED", "reviewer": "code-policy", "ledgerLabel": "233406 / ep 11", "progress": 100, "ledgerIndex": 12, "modelCostUsd": 0.0, "modelCalls": 0, "activeArm": "right", "tokenUsage": { "total": 0, "uncachedInput": 0, "cachedInput": 0, "output": 0, "reasoning": 0 }, "video": { "mp4": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/episode-enpire-test-0__20260919-233406__ep_0011-10x.mp4", "webm": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/episode-enpire-test-0__20260919-233406__ep_0011-10x.webm", "poster": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/episode-enpire-test-0__20260919-233406__ep_0011-10x-poster.jpg", "speedup": 10, "width": 1280, "height": 360 }, "evidenceImage": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/evidence/enpire-test-0__20260919-233406__ep_0011.jpg", "episodeNumber": 12, "displayName": "Code-as-policy - episode 12" }, { "id": "enpire-test-0__20260919-233406__ep_0012", "run": "20260919-233406", "episode": "ep_0012", "approach": "code-as-policy", "cohort": "frozen-eval", "grasp": { "score": 0, "item": "No purposeful target contact, or only a distractor is grasped.", "evidence": "The robot moves to an observation pose then stops; target remains untouched and gripper open.", "timestamps": [ 0, 20, 90, 179.5 ] }, "transport": { "score": 0, "item": "No transport beyond the initial lift, or no retained lift.", "evidence": "No retained lift or transport.", "timestamps": [ 20, 90, 179.5 ] }, "placement": { "score": 0, "item": "Finishes outside the box after a miss or bounce, or placement is not reached.", "evidence": "Target stays outside the box; placement not reached.", "timestamps": [ 20, 177, 179.5 ] }, "complete": false, "attempts": 0, "drops": 0, "placementAttempts": 0, "confidence": "high", "reviewNote": "Target begins on its side; its shape in the wrist image is not a failed grasp. Topdown and active right-wrist frames reviewed over the full rollout at 10 s intervals, grasp at 0.8 s, carry at 3 s and releases at 0.4 s. Final 177–179.5 s frames establish end state. Scores reflect visible behavior at this sampling resolution; brief unsampled contact cannot be excluded.", "reviewedEvidence": [ "enpire-test-0/yam_red_lego_in_dark_gray_box/20260919-233406/recording.rrd", "analysis/sheets/enpire-test-0__20260919-233406__ep_0012.jpg", "analysis/scoring-review/code-sheets/enpire-test-0__20260919-233406__ep_0012__overview.jpg" ], "total": 0, "reviewStatus": "scored", "source": "enpire-test-0/yam_red_lego_in_dark_gray_box/20260919-233406/recording.rrd", "durationSeconds": 180.064, "rolloutOriginSeconds": 3050.414, "terminalCode": "COMPLETED", "reviewer": "code-policy", "ledgerLabel": "233406 / ep 12", "progress": 0, "ledgerIndex": 13, "modelCostUsd": 0.0, "modelCalls": 0, "activeArm": "right", "tokenUsage": { "total": 0, "uncachedInput": 0, "cachedInput": 0, "output": 0, "reasoning": 0 }, "video": { "mp4": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/episode-enpire-test-0__20260919-233406__ep_0012-10x.mp4", "webm": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/episode-enpire-test-0__20260919-233406__ep_0012-10x.webm", "poster": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/episode-enpire-test-0__20260919-233406__ep_0012-10x-poster.jpg", "speedup": 10, "width": 1280, "height": 360 }, "evidenceImage": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/evidence/enpire-test-0__20260919-233406__ep_0012.jpg", "episodeNumber": 13, "displayName": "Code-as-policy - episode 13" }, { "id": "enpire-test-0__20260919-233406__ep_0013", "run": "20260919-233406", "episode": "ep_0013", "approach": "code-as-policy", "cohort": "frozen-eval", "grasp": { "score": 0, "item": "No purposeful target contact, or only a distractor is grasped.", "evidence": "Open gripper approaches but remains separated from the target; no purposeful contact observed.", "timestamps": [ 0, 50, 90, 179.5 ] }, "transport": { "score": 0, "item": "No transport beyond the initial lift, or no retained lift.", "evidence": "No target acquisition or transfer.", "timestamps": [ 50, 90, 179.5 ] }, "placement": { "score": 0, "item": "Finishes outside the box after a miss or bounce, or placement is not reached.", "evidence": "Target stays on the table outside the box.", "timestamps": [ 50, 177, 179.5 ] }, "complete": false, "attempts": 0, "drops": 0, "placementAttempts": 0, "confidence": "high", "reviewNote": "Topdown and active right-wrist frames reviewed over the full rollout at 10 s intervals, grasp at 0.8 s, carry at 3 s and releases at 0.4 s. Final 177–179.5 s frames establish end state. Scores reflect visible behavior at this sampling resolution; brief unsampled contact cannot be excluded.", "reviewedEvidence": [ "enpire-test-0/yam_red_lego_in_dark_gray_box/20260919-233406/recording.rrd", "analysis/sheets/enpire-test-0__20260919-233406__ep_0013.jpg", "analysis/scoring-review/code-sheets/enpire-test-0__20260919-233406__ep_0013__overview.jpg" ], "total": 0, "reviewStatus": "scored", "source": "enpire-test-0/yam_red_lego_in_dark_gray_box/20260919-233406/recording.rrd", "durationSeconds": 180.014, "rolloutOriginSeconds": 3262.309, "terminalCode": "COMPLETED", "reviewer": "code-policy", "ledgerLabel": "233406 / ep 13", "progress": 0, "ledgerIndex": 14, "modelCostUsd": 0.0, "modelCalls": 0, "activeArm": "right", "tokenUsage": { "total": 0, "uncachedInput": 0, "cachedInput": 0, "output": 0, "reasoning": 0 }, "video": { "mp4": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/episode-enpire-test-0__20260919-233406__ep_0013-10x.mp4", "webm": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/episode-enpire-test-0__20260919-233406__ep_0013-10x.webm", "poster": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/episode-enpire-test-0__20260919-233406__ep_0013-10x-poster.jpg", "speedup": 10, "width": 1280, "height": 360 }, "evidenceImage": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/evidence/enpire-test-0__20260919-233406__ep_0013.jpg", "episodeNumber": 14, "displayName": "Code-as-policy - episode 14" }, { "id": "enpire-test-0__20260919-233406__ep_0014", "run": "20260919-233406", "episode": "ep_0014", "approach": "code-as-policy", "cohort": "frozen-eval", "grasp": { "score": 30, "item": "First attempt: stable grasp and lift, without corrective grip adjustment.", "evidence": "Target starts on its side and is held in that stable orientation on the first grasp.", "timestamps": [ 92.4, 95.6, 102 ] }, "transport": { "score": 30, "item": "Release pose reached with no slipping, drops, or unintended contact.", "evidence": "Carries retained target fully over the opening without visible slippage or drop.", "timestamps": [ 112.6, 124.6, 133.6 ] }, "placement": { "score": 5, "item": "Held above the opening or inside the box, but never released.", "evidence": "Gripper stays closed with target above the box through episode end.", "timestamps": [ 140, 177, 179.5 ] }, "complete": false, "attempts": 1, "drops": 0, "placementAttempts": 0, "confidence": "medium", "reviewNote": "The sideways brick orientation precedes the grasp and is retained; it is not a slip. Topdown and active right-wrist frames reviewed over the full rollout at 10 s intervals, grasp at 0.8 s, carry at 3 s and releases at 0.4 s. Final 177–179.5 s frames establish end state. Scores reflect visible behavior at this sampling resolution; brief unsampled contact cannot be excluded.", "reviewedEvidence": [ "enpire-test-0/yam_red_lego_in_dark_gray_box/20260919-233406/recording.rrd", "analysis/sheets/enpire-test-0__20260919-233406__ep_0014.jpg", "analysis/scoring-review/code-sheets/enpire-test-0__20260919-233406__ep_0014__overview.jpg", "analysis/scoring-review/code-sheets/enpire-test-0__20260919-233406__ep_0014__grasp.jpg", "analysis/scoring-review/code-sheets/enpire-test-0__20260919-233406__ep_0014__transfer.jpg" ], "total": 65, "reviewStatus": "scored", "source": "enpire-test-0/yam_red_lego_in_dark_gray_box/20260919-233406/recording.rrd", "durationSeconds": 180.068, "rolloutOriginSeconds": 3625.127, "terminalCode": "COMPLETED", "reviewer": "code-policy", "ledgerLabel": "233406 / ep 14", "progress": 75, "ledgerIndex": 15, "modelCostUsd": 0.0, "modelCalls": 0, "activeArm": "right", "tokenUsage": { "total": 0, "uncachedInput": 0, "cachedInput": 0, "output": 0, "reasoning": 0 }, "video": { "mp4": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/recovery-A-14-10x.mp4", "webm": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/recovery-A-14-10x.webm", "poster": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/recovery-A-14-10x-poster.jpg", "speedup": 10, "width": 1280, "height": 360 }, "evidenceImage": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/evidence/enpire-test-0__20260919-233406__ep_0014.jpg", "episodeNumber": 15, "displayName": "Code-as-policy - episode 15" }, { "id": "enpire-test-0__20260919-233406__ep_0015", "run": "20260919-233406", "episode": "ep_0015", "approach": "code-as-policy", "cohort": "frozen-eval", "grasp": { "score": 0, "item": "No purposeful target contact, or only a distractor is grasped.", "evidence": "Arms remain at starting poses and target is untouched.", "timestamps": [ 0, 60, 120, 179.5 ] }, "transport": { "score": 0, "item": "No transport beyond the initial lift, or no retained lift.", "evidence": "No retained lift or transport.", "timestamps": [ 0, 90, 179.5 ] }, "placement": { "score": 0, "item": "Finishes outside the box after a miss or bounce, or placement is not reached.", "evidence": "Target remains outside the box throughout.", "timestamps": [ 0, 177, 179.5 ] }, "complete": false, "attempts": 0, "drops": 0, "placementAttempts": 0, "confidence": "high", "reviewNote": "Topdown and active right-wrist frames reviewed over the full rollout at 10 s intervals, grasp at 0.8 s, carry at 3 s and releases at 0.4 s. Final 177–179.5 s frames establish end state. Scores reflect visible behavior at this sampling resolution; brief unsampled contact cannot be excluded.", "reviewedEvidence": [ "enpire-test-0/yam_red_lego_in_dark_gray_box/20260919-233406/recording.rrd", "analysis/sheets/enpire-test-0__20260919-233406__ep_0015.jpg", "analysis/scoring-review/code-sheets/enpire-test-0__20260919-233406__ep_0015__overview.jpg" ], "total": 0, "reviewStatus": "scored", "source": "enpire-test-0/yam_red_lego_in_dark_gray_box/20260919-233406/recording.rrd", "durationSeconds": 180.012, "rolloutOriginSeconds": 3899.516, "terminalCode": "COMPLETED", "reviewer": "code-policy", "ledgerLabel": "233406 / ep 15", "progress": 0, "ledgerIndex": 16, "modelCostUsd": 0.0, "modelCalls": 0, "activeArm": "right", "tokenUsage": { "total": 0, "uncachedInput": 0, "cachedInput": 0, "output": 0, "reasoning": 0 }, "video": { "mp4": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/episode-enpire-test-0__20260919-233406__ep_0015-10x.mp4", "webm": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/episode-enpire-test-0__20260919-233406__ep_0015-10x.webm", "poster": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/episode-enpire-test-0__20260919-233406__ep_0015-10x-poster.jpg", "speedup": 10, "width": 1280, "height": 360 }, "evidenceImage": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/evidence/enpire-test-0__20260919-233406__ep_0015.jpg", "episodeNumber": 16, "displayName": "Code-as-policy - episode 16" }, { "id": "enpire-test-0__20260919-233406__ep_0016", "run": "20260919-233406", "episode": "ep_0016", "approach": "code-as-policy", "cohort": "frozen-eval", "grasp": { "score": 30, "item": "First attempt: stable grasp and lift, without corrective grip adjustment.", "evidence": "Stable first grasp retains target throughout the lift.", "timestamps": [ 89.5, 92.7, 99.1 ] }, "transport": { "score": 5, "item": "Transfer started but release pose never reached; includes an unrecovered drop.", "evidence": "Carry begins but stops alongside the box's left edge; target is not fully over the opening.", "timestamps": [ 112.7, 121.7, 127.7 ] }, "placement": { "score": 0, "item": "Finishes outside the box after a miss or bounce, or placement is not reached.", "evidence": "Target is held at the outside edge; no release and no completed placement approach.", "timestamps": [ 130, 177, 179.5 ] }, "complete": false, "attempts": 1, "drops": 0, "placementAttempts": 0, "confidence": "medium", "reviewNote": "Topdown and active right-wrist frames reviewed over the full rollout at 10 s intervals, grasp at 0.8 s, carry at 3 s and releases at 0.4 s. Final 177–179.5 s frames establish end state. Scores reflect visible behavior at this sampling resolution; brief unsampled contact cannot be excluded.", "reviewedEvidence": [ "enpire-test-0/yam_red_lego_in_dark_gray_box/20260919-233406/recording.rrd", "analysis/sheets/enpire-test-0__20260919-233406__ep_0016.jpg", "analysis/scoring-review/code-sheets/enpire-test-0__20260919-233406__ep_0016__overview.jpg", "analysis/scoring-review/code-sheets/enpire-test-0__20260919-233406__ep_0016__grasp.jpg", "analysis/scoring-review/code-sheets/enpire-test-0__20260919-233406__ep_0016__transfer.jpg" ], "total": 35, "reviewStatus": "scored", "source": "enpire-test-0/yam_red_lego_in_dark_gray_box/20260919-233406/recording.rrd", "durationSeconds": 180.07, "rolloutOriginSeconds": 4123.554, "terminalCode": "COMPLETED", "reviewer": "code-policy", "ledgerLabel": "233406 / ep 16", "progress": 50, "ledgerIndex": 17, "modelCostUsd": 0.0, "modelCalls": 0, "activeArm": "right", "tokenUsage": { "total": 0, "uncachedInput": 0, "cachedInput": 0, "output": 0, "reasoning": 0 }, "video": { "mp4": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/episode-enpire-test-0__20260919-233406__ep_0016-10x.mp4", "webm": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/episode-enpire-test-0__20260919-233406__ep_0016-10x.webm", "poster": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/episode-enpire-test-0__20260919-233406__ep_0016-10x-poster.jpg", "speedup": 10, "width": 1280, "height": 360 }, "evidenceImage": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/evidence/enpire-test-0__20260919-233406__ep_0016.jpg", "episodeNumber": 17, "displayName": "Code-as-policy - episode 17" }, { "id": "enpire-test-0__20260919-233406__ep_0017", "run": "20260919-233406", "episode": "ep_0017", "approach": "code-as-policy", "cohort": "frozen-eval", "grasp": { "score": 30, "item": "First attempt: stable grasp and lift, without corrective grip adjustment.", "evidence": "First closure secures target; no visible slip during clear lift.", "timestamps": [ 84.5, 87.7, 94.9 ] }, "transport": { "score": 30, "item": "Release pose reached with no slipping, drops, or unintended contact.", "evidence": "Target reaches a release pose fully within the opening while remaining securely held.", "timestamps": [ 110.7, 122.7, 126.3 ] }, "placement": { "score": 40, "item": "Complete on the first placement attempt, without rim contact or correction.", "evidence": "Single release around 127.5 s; target lands inside away from rim, rests, and stays there at episode end.", "timestamps": [ 127.9, 130.7, 177, 179.5 ] }, "complete": true, "attempts": 1, "drops": 0, "placementAttempts": 1, "confidence": "medium", "reviewNote": "Topdown and active right-wrist frames reviewed over the full rollout at 10 s intervals, grasp at 0.8 s, carry at 3 s and releases at 0.4 s. Final 177–179.5 s frames establish end state. Scores reflect visible behavior at this sampling resolution; brief unsampled contact cannot be excluded.", "reviewedEvidence": [ "enpire-test-0/yam_red_lego_in_dark_gray_box/20260919-233406/recording.rrd", "analysis/sheets/enpire-test-0__20260919-233406__ep_0017.jpg", "analysis/scoring-review/code-sheets/enpire-test-0__20260919-233406__ep_0017__overview.jpg", "analysis/scoring-review/code-sheets/enpire-test-0__20260919-233406__ep_0017__grasp.jpg", "analysis/scoring-review/code-sheets/enpire-test-0__20260919-233406__ep_0017__transfer.jpg", "analysis/scoring-review/code-sheets/enpire-test-0__20260919-233406__ep_0017__release.jpg" ], "total": 100, "reviewStatus": "scored", "source": "enpire-test-0/yam_red_lego_in_dark_gray_box/20260919-233406/recording.rrd", "durationSeconds": 180.015, "rolloutOriginSeconds": 4367.022, "terminalCode": "COMPLETED", "reviewer": "code-policy", "ledgerLabel": "233406 / ep 17", "progress": 100, "ledgerIndex": 18, "modelCostUsd": 0.0, "modelCalls": 0, "activeArm": "right", "tokenUsage": { "total": 0, "uncachedInput": 0, "cachedInput": 0, "output": 0, "reasoning": 0 }, "video": { "mp4": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/episode-enpire-test-0__20260919-233406__ep_0017-10x.mp4", "webm": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/episode-enpire-test-0__20260919-233406__ep_0017-10x.webm", "poster": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/episode-enpire-test-0__20260919-233406__ep_0017-10x-poster.jpg", "speedup": 10, "width": 1280, "height": 360 }, "evidenceImage": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/evidence/enpire-test-0__20260919-233406__ep_0017.jpg", "episodeNumber": 18, "displayName": "Code-as-policy - episode 18" }, { "id": "enpire-test-0__20260919-233406__ep_0018", "run": "20260919-233406", "episode": "ep_0018", "approach": "code-as-policy", "cohort": "frozen-eval", "grasp": { "score": 30, "item": "First attempt: stable grasp and lift, without corrective grip adjustment.", "evidence": "First grasp lifts the sideways target securely without changing its orientation.", "timestamps": [ 84.5, 87.7, 94.9 ] }, "transport": { "score": 5, "item": "Transfer started but release pose never reached; includes an unrecovered drop.", "evidence": "Target is moved toward the box but stalls outside the left edge; a release pose fully over the opening is not reached.", "timestamps": [ 110.7, 119.7, 122.7 ] }, "placement": { "score": 0, "item": "Finishes outside the box after a miss or bounce, or placement is not reached.", "evidence": "Target remains held outside/alongside the box edge; no release.", "timestamps": [ 130, 177, 179.5 ] }, "complete": false, "attempts": 1, "drops": 0, "placementAttempts": 0, "confidence": "medium", "reviewNote": "Topdown and active right-wrist frames reviewed over the full rollout at 10 s intervals, grasp at 0.8 s, carry at 3 s and releases at 0.4 s. Final 177–179.5 s frames establish end state. Scores reflect visible behavior at this sampling resolution; brief unsampled contact cannot be excluded.", "reviewedEvidence": [ "enpire-test-0/yam_red_lego_in_dark_gray_box/20260919-233406/recording.rrd", "analysis/sheets/enpire-test-0__20260919-233406__ep_0018.jpg", "analysis/scoring-review/code-sheets/enpire-test-0__20260919-233406__ep_0018__overview.jpg", "analysis/scoring-review/code-sheets/enpire-test-0__20260919-233406__ep_0018__grasp.jpg", "analysis/scoring-review/code-sheets/enpire-test-0__20260919-233406__ep_0018__transfer.jpg" ], "total": 35, "reviewStatus": "scored", "source": "enpire-test-0/yam_red_lego_in_dark_gray_box/20260919-233406/recording.rrd", "durationSeconds": 180.041, "rolloutOriginSeconds": 4575.608, "terminalCode": "COMPLETED", "reviewer": "code-policy", "ledgerLabel": "233406 / ep 18", "progress": 50, "ledgerIndex": 19, "modelCostUsd": 0.0, "modelCalls": 0, "activeArm": "right", "tokenUsage": { "total": 0, "uncachedInput": 0, "cachedInput": 0, "output": 0, "reasoning": 0 }, "video": { "mp4": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/episode-enpire-test-0__20260919-233406__ep_0018-10x.mp4", "webm": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/episode-enpire-test-0__20260919-233406__ep_0018-10x.webm", "poster": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/episode-enpire-test-0__20260919-233406__ep_0018-10x-poster.jpg", "speedup": 10, "width": 1280, "height": 360 }, "evidenceImage": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/evidence/enpire-test-0__20260919-233406__ep_0018.jpg", "episodeNumber": 19, "displayName": "Code-as-policy - episode 19" }, { "id": "enpire-test-0__20260919-233406__ep_0019", "run": "20260919-233406", "episode": "ep_0019", "approach": "code-as-policy", "cohort": "frozen-eval", "grasp": { "score": 0, "item": "No purposeful target contact, or only a distractor is grasped.", "evidence": "Open fingers stop around the target with clear gaps; no purposeful contact or closure observed.", "timestamps": [ 0, 70, 80, 179.5 ] }, "transport": { "score": 0, "item": "No transport beyond the initial lift, or no retained lift.", "evidence": "Target stays on the table, without retained lift.", "timestamps": [ 80, 90, 179.5 ] }, "placement": { "score": 0, "item": "Finishes outside the box after a miss or bounce, or placement is not reached.", "evidence": "Placement is not reached; target remains outside the box.", "timestamps": [ 80, 177, 179.5 ] }, "complete": false, "attempts": 0, "drops": 0, "placementAttempts": 0, "confidence": "high", "reviewNote": "Topdown and active right-wrist frames reviewed over the full rollout at 10 s intervals, grasp at 0.8 s, carry at 3 s and releases at 0.4 s. Final 177–179.5 s frames establish end state. Scores reflect visible behavior at this sampling resolution; brief unsampled contact cannot be excluded.", "reviewedEvidence": [ "enpire-test-0/yam_red_lego_in_dark_gray_box/20260919-233406/recording.rrd", "analysis/sheets/enpire-test-0__20260919-233406__ep_0019.jpg", "analysis/scoring-review/code-sheets/enpire-test-0__20260919-233406__ep_0019__overview.jpg" ], "total": 0, "reviewStatus": "scored", "source": "enpire-test-0/yam_red_lego_in_dark_gray_box/20260919-233406/recording.rrd", "durationSeconds": 180.049, "rolloutOriginSeconds": 4920.752, "terminalCode": "COMPLETED", "reviewer": "code-policy", "ledgerLabel": "233406 / ep 19", "progress": 0, "ledgerIndex": 20, "modelCostUsd": 0.0, "modelCalls": 0, "activeArm": "right", "tokenUsage": { "total": 0, "uncachedInput": 0, "cachedInput": 0, "output": 0, "reasoning": 0 }, "video": { "mp4": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/episode-enpire-test-0__20260919-233406__ep_0019-10x.mp4", "webm": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/episode-enpire-test-0__20260919-233406__ep_0019-10x.webm", "poster": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/episode-enpire-test-0__20260919-233406__ep_0019-10x-poster.jpg", "speedup": 10, "width": 1280, "height": 360 }, "evidenceImage": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/evidence/enpire-test-0__20260919-233406__ep_0019.jpg", "episodeNumber": 20, "displayName": "Code-as-policy - episode 20" }, { "id": "se3labs-eval-rac__20260920-211500__ep_0000", "run": "20260920-211500", "episode": "ep_0000", "approach": "agentic-policy", "cohort": "eval", "grasp": { "score": 30, "item": "First attempt: stable grasp and lift, without corrective grip adjustment.", "evidence": "First closure at 92.2 s captures the red block; the lifted target remains fixed between the jaws.", "timestamps": [ 92.2, 94.2, 96.2 ] }, "transport": { "score": 30, "item": "Release pose reached with no slipping, drops, or unintended contact.", "evidence": "Target remains retained through the traverse to a pose over the box interior.", "timestamps": [ 104.2, 112.2, 126.1 ] }, "placement": { "score": 40, "item": "Complete on the first placement attempt, without rim contact or correction.", "evidence": "Release at about 131.1 s; open jaws withdraw and the block remains fully inside through the final two seconds.", "timestamps": [ 131.1, 132.1, 142.1, 144.1 ] }, "complete": true, "attempts": 1, "drops": 0, "placementAttempts": 1, "confidence": "high", "reviewNote": "Paired topdown and active-wrist imagery reviewed at acquisition, transport, release, and rollout end. Gripper telemetry was used to locate actions, not to assign quality scores.", "reviewedEvidence": [ "se3labs-eval-rac/yam_red_lego_in_dark_gray_box/20260920-211500/recording.rrd", "analysis/extract/se3labs-eval-rac__20260920-211500/series.json", "analysis/sheets/se3labs-eval-rac__20260920-211500__ep_0000.jpg", "analysis/scoring-review/agentic-sheets/20260920-211500_0-144.092_8_0.jpg" ], "total": 100, "reviewStatus": "scored", "source": "se3labs-eval-rac/yam_red_lego_in_dark_gray_box/20260920-211500/recording.rrd", "durationSeconds": 144.092, "rolloutOriginSeconds": 25.716, "terminalCode": "COMPLETED", "reviewer": "agentic", "ledgerLabel": "211500", "progress": 100, "ledgerIndex": 21, "modelCostUsd": 1.76, "modelCalls": 28, "activeArm": "right", "tokenUsage": { "total": 1165654, "uncachedInput": 52234, "cachedInput": 1110912, "output": 2508, "reasoning": 584 }, "video": { "mp4": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/episode-se3labs-eval-rac__20260920-211500__ep_0000-10x.mp4", "webm": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/episode-se3labs-eval-rac__20260920-211500__ep_0000-10x.webm", "poster": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/episode-se3labs-eval-rac__20260920-211500__ep_0000-10x-poster.jpg", "speedup": 10, "width": 1280, "height": 360 }, "evidenceImage": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/evidence/se3labs-eval-rac__20260920-211500__ep_0000.jpg", "episodeNumber": 1, "displayName": "Agentic Policy - episode 1" }, { "id": "se3labs-eval-rac__20260920-211841__ep_0000", "run": "20260920-211841", "episode": "ep_0000", "approach": "agentic-policy", "cohort": "eval", "grasp": { "score": 20, "item": "After one or more failed attempts: stable grasp and lift.", "evidence": "Initial partial closure at 97 s fails to acquire the target. Repositioning and a second closure at 115.5 s yield a retained, stable lift.", "timestamps": [ 97, 105.6, 115.5, 118, 122 ] }, "transport": { "score": 30, "item": "Release pose reached with no slipping, drops, or unintended contact.", "evidence": "The red target remains fixed beneath an accompanying purple block as the arm reaches the box interior; no target drop is visible.", "timestamps": [ 133, 145, 165 ] }, "placement": { "score": 40, "item": "Complete on the first placement attempt, without rim contact or correction.", "evidence": "Jaws open at about 167 s and the red target settles wholly inside; it remains stationary through the rollout end.", "timestamps": [ 167, 169, 173, 176.6 ] }, "complete": true, "attempts": 2, "drops": 0, "placementAttempts": 1, "confidence": "medium", "reviewNote": "A purple distractor is carried with the target. The rubric scores target manipulation and contains no separate extra-object penalty.", "reviewedEvidence": [ "se3labs-eval-rac/yam_red_lego_in_dark_gray_box/20260920-211841/recording.rrd", "analysis/extract/se3labs-eval-rac__20260920-211841/series.json", "analysis/sheets/se3labs-eval-rac__20260920-211841__ep_0000.jpg", "analysis/scoring-review/agentic-sheets/20260920-211841_85-279.412_4_0.jpg", "analysis/scoring-review/agentic-sheets/20260920-211841_85-279.412_4_1.jpg", "analysis/scoring-review/agentic-sheets/20260920-211841_94-122_1_0.jpg", "analysis/scoring-review/agentic-sheets/20260920-211841_94-122_1_1.jpg" ], "total": 90, "reviewStatus": "scored", "source": "se3labs-eval-rac/yam_red_lego_in_dark_gray_box/20260920-211841/recording.rrd", "durationSeconds": 176.605, "rolloutOriginSeconds": 23.373, "terminalCode": "COMPLETED", "reviewer": "agentic", "ledgerLabel": "211841", "progress": 100, "ledgerIndex": 22, "modelCostUsd": 2.22, "modelCalls": 32, "activeArm": "right", "tokenUsage": { "total": 1487241, "uncachedInput": 59708, "cachedInput": 1423616, "output": 3917, "reasoning": 1108 }, "video": { "mp4": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/episode-se3labs-eval-rac__20260920-211841__ep_0000-10x.mp4", "webm": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/episode-se3labs-eval-rac__20260920-211841__ep_0000-10x.webm", "poster": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/episode-se3labs-eval-rac__20260920-211841__ep_0000-10x-poster.jpg", "speedup": 10, "width": 1280, "height": 360 }, "evidenceImage": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/evidence/se3labs-eval-rac__20260920-211841__ep_0000.jpg", "episodeNumber": 2, "displayName": "Agentic Policy - episode 2" }, { "id": "se3labs-eval-rac__20260920-212446__ep_0000", "run": "20260920-212446", "episode": "ep_0000", "approach": "agentic-policy", "cohort": "eval", "grasp": { "score": 30, "item": "First attempt: stable grasp and lift, without corrective grip adjustment.", "evidence": "The first closure at about 155 s securely captures the target; it follows the wrist upward without slipping.", "timestamps": [ 154.8, 156.8, 158.8 ] }, "transport": { "score": 30, "item": "Release pose reached with no slipping, drops, or unintended contact.", "evidence": "Stable carry from the acquisition area to a pose above the interior.", "timestamps": [ 166.8, 174.8, 191.8 ] }, "placement": { "score": 40, "item": "Complete on the first placement attempt, without rim contact or correction.", "evidence": "First release at about 196.8 s lands inside without visible rim contact; the target is at rest over the final two seconds.", "timestamps": [ 196.8, 198.8, 207, 209 ] }, "complete": true, "attempts": 1, "drops": 0, "placementAttempts": 1, "confidence": "high", "reviewNote": "Paired topdown and active-wrist imagery reviewed at acquisition, transport, release, and rollout end. Gripper telemetry was used to locate actions, not to assign quality scores.", "reviewedEvidence": [ "se3labs-eval-rac/yam_red_lego_in_dark_gray_box/20260920-212446/recording.rrd", "analysis/extract/se3labs-eval-rac__20260920-212446/series.json", "analysis/sheets/se3labs-eval-rac__20260920-212446__ep_0000.jpg", "analysis/scoring-review/agentic-sheets/20260920-212446_0-209.048_8_0.jpg" ], "total": 100, "reviewStatus": "scored", "source": "se3labs-eval-rac/yam_red_lego_in_dark_gray_box/20260920-212446/recording.rrd", "durationSeconds": 209.048, "rolloutOriginSeconds": 12.444, "terminalCode": "COMPLETED", "reviewer": "agentic", "ledgerLabel": "212446", "progress": 100, "ledgerIndex": 23, "modelCostUsd": 2.72, "modelCalls": 38, "activeArm": "right", "tokenUsage": { "total": 1804890, "uncachedInput": 81508, "cachedInput": 1719680, "output": 3702, "reasoning": 1259 }, "video": { "mp4": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/episode-se3labs-eval-rac__20260920-212446__ep_0000-10x.mp4", "webm": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/episode-se3labs-eval-rac__20260920-212446__ep_0000-10x.webm", "poster": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/episode-se3labs-eval-rac__20260920-212446__ep_0000-10x-poster.jpg", "speedup": 10, "width": 1280, "height": 360 }, "evidenceImage": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/evidence/se3labs-eval-rac__20260920-212446__ep_0000.jpg", "episodeNumber": 3, "displayName": "Agentic Policy - episode 3" }, { "id": "se3labs-eval-rac__20260920-213042__ep_0000", "run": "20260920-213042", "episode": "ep_0000", "approach": "agentic-policy", "cohort": "eval", "grasp": { "score": 20, "item": "After one or more failed attempts: stable grasp and lift.", "evidence": "Closure at 171.3 s misses; the target remains on the table. A second closure at 220.3 s grips and lifts it steadily.", "timestamps": [ 171.3, 180, 187.2, 220.3, 223, 225 ] }, "transport": { "score": 30, "item": "Release pose reached with no slipping, drops, or unintended contact.", "evidence": "The retained target moves to the box opening without visible slip or drop.", "timestamps": [ 244, 255, 268 ] }, "placement": { "score": 40, "item": "Complete on the first placement attempt, without rim contact or correction.", "evidence": "Released at about 269.1 s; the target is fully inside and stationary at the end.", "timestamps": [ 269.1, 271, 276.5, 278.5 ] }, "complete": true, "attempts": 2, "drops": 0, "placementAttempts": 1, "confidence": "high", "reviewNote": "A person's hand enters the far edge of the topdown view at the end but does not visibly touch the target or box.", "reviewedEvidence": [ "se3labs-eval-rac/yam_red_lego_in_dark_gray_box/20260920-213042/recording.rrd", "analysis/extract/se3labs-eval-rac__20260920-213042/series.json", "analysis/sheets/se3labs-eval-rac__20260920-213042__ep_0000.jpg", "analysis/scoring-review/agentic-sheets/20260920-213042_160-278.53_5_0.jpg", "analysis/scoring-review/agentic-sheets/20260920-213042_160-278.53_5_1.jpg", "analysis/scoring-review/agentic-sheets/20260920-213042_0-278.53_8_0.jpg" ], "total": 90, "reviewStatus": "scored", "source": "se3labs-eval-rac/yam_red_lego_in_dark_gray_box/20260920-213042/recording.rrd", "durationSeconds": 278.53, "rolloutOriginSeconds": 31.223, "terminalCode": "COMPLETED", "reviewer": "agentic", "ledgerLabel": "213042", "progress": 100, "ledgerIndex": 24, "modelCostUsd": 3.55, "modelCalls": 52, "activeArm": "left", "tokenUsage": { "total": 2613439, "uncachedInput": 75977, "cachedInput": 2532224, "output": 5238, "reasoning": 1523 }, "video": { "mp4": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/episode-se3labs-eval-rac__20260920-213042__ep_0000-10x.mp4", "webm": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/episode-se3labs-eval-rac__20260920-213042__ep_0000-10x.webm", "poster": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/episode-se3labs-eval-rac__20260920-213042__ep_0000-10x-poster.jpg", "speedup": 10, "width": 1280, "height": 360 }, "evidenceImage": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/evidence/se3labs-eval-rac__20260920-213042__ep_0000.jpg", "episodeNumber": 4, "displayName": "Agentic Policy - episode 4" }, { "id": "se3labs-eval-rac__20260920-214242__ep_0000", "run": "20260920-214242", "episode": "ep_0000", "approach": "agentic-policy", "cohort": "eval", "grasp": { "score": 5, "item": "Target contacted, but no retained lift; includes an immediate loss of grip.", "evidence": "Partial closure at 80.1 s, renewed closure at 116.7 s, and a later attempt at 168.6 s do not produce a retained lift. The target is displaced during the failed closures and leaves the visible work area on the last attempt.", "timestamps": [ 80.1, 116.7, 168.6, 169, 171 ] }, "transport": { "score": 0, "item": "No transport beyond the initial lift, or no retained lift.", "evidence": "The robot never transports a retained target.", "timestamps": [ 169, 183, 231.3 ] }, "placement": { "score": 0, "item": "Finishes outside the box after a miss or bounce, or placement is not reached.", "evidence": "No placement is attempted; the target is not in the box at rollout end.", "timestamps": [ 195, 219, 231.3 ] }, "complete": false, "attempts": 3, "drops": 0, "placementAttempts": 0, "confidence": "medium", "reviewNote": "Three purposeful acquisition closures are counted, including the partial initial closure.", "reviewedEvidence": [ "se3labs-eval-rac/yam_red_lego_in_dark_gray_box/20260920-214242/recording.rrd", "analysis/extract/se3labs-eval-rac__20260920-214242/series.json", "analysis/sheets/se3labs-eval-rac__20260920-214242__ep_0000.jpg", "analysis/scoring-review/agentic-sheets/20260920-214242_75-231.281_6_0.jpg", "analysis/scoring-review/agentic-sheets/20260920-214242_75-231.281_6_1.jpg", "analysis/scoring-review/agentic-sheets/20260920-214242_165-173_0.4_0.jpg" ], "total": 5, "reviewStatus": "scored", "source": "se3labs-eval-rac/yam_red_lego_in_dark_gray_box/20260920-214242/recording.rrd", "durationSeconds": 231.281, "rolloutOriginSeconds": 8.821, "terminalCode": "COMPLETED", "reviewer": "agentic", "ledgerLabel": "214242", "progress": 25, "ledgerIndex": 25, "modelCostUsd": 2.79, "modelCalls": 41, "activeArm": "left", "tokenUsage": { "total": 2017358, "uncachedInput": 65657, "cachedInput": 1948032, "output": 3669, "reasoning": 1102 }, "video": { "mp4": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/episode-se3labs-eval-rac__20260920-214242__ep_0000-10x.mp4", "webm": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/episode-se3labs-eval-rac__20260920-214242__ep_0000-10x.webm", "poster": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/episode-se3labs-eval-rac__20260920-214242__ep_0000-10x-poster.jpg", "speedup": 10, "width": 1280, "height": 360 }, "evidenceImage": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/evidence/se3labs-eval-rac__20260920-214242__ep_0000.jpg", "episodeNumber": 5, "displayName": "Agentic Policy - episode 5" }, { "id": "se3labs-eval-rac__20260920-214816__ep_0000", "run": "20260920-214816", "episode": "ep_0000", "approach": "agentic-policy", "cohort": "eval", "grasp": { "score": 20, "item": "After one or more failed attempts: stable grasp and lift.", "evidence": "The first closure near the box at 157.5 s misses. After reopening and repositioning, the closure at 228.6 s produces a stable retained lift.", "timestamps": [ 157.5, 165, 175.4, 228.6, 231, 234 ] }, "transport": { "score": 30, "item": "Release pose reached with no slipping, drops, or unintended contact.", "evidence": "The target remains fixed between the jaws through a slow traverse around the box edge and over its opening.", "timestamps": [ 280, 305, 320, 346 ] }, "placement": { "score": 40, "item": "Complete on the first placement attempt, without rim contact or correction.", "evidence": "First release at about 347.2 s leaves the target fully inside; it is stationary during the final two seconds.", "timestamps": [ 347.2, 349, 357.2, 359.2 ] }, "complete": true, "attempts": 2, "drops": 0, "placementAttempts": 1, "confidence": "medium", "reviewNote": "Long pauses and view changes do not receive a penalty. No visible target-to-rim impact was found.", "reviewedEvidence": [ "se3labs-eval-rac/yam_red_lego_in_dark_gray_box/20260920-214816/recording.rrd", "analysis/extract/se3labs-eval-rac__20260920-214816/series.json", "analysis/sheets/se3labs-eval-rac__20260920-214816__ep_0000.jpg", "analysis/scoring-review/agentic-sheets/20260920-214816_85-359.234_10_0.jpg", "analysis/scoring-review/agentic-sheets/20260920-214816_85-359.234_10_1.jpg", "analysis/scoring-review/agentic-sheets/20260920-214816_0-359.234_8_0.jpg" ], "total": 90, "reviewStatus": "scored", "source": "se3labs-eval-rac/yam_red_lego_in_dark_gray_box/20260920-214816/recording.rrd", "durationSeconds": 359.234, "rolloutOriginSeconds": 20.584, "terminalCode": "COMPLETED", "reviewer": "agentic", "ledgerLabel": "214816", "progress": 100, "ledgerIndex": 26, "modelCostUsd": 3.81, "modelCalls": 57, "activeArm": "left", "tokenUsage": { "total": 2811720, "uncachedInput": 81665, "cachedInput": 2724736, "output": 5319, "reasoning": 1739 }, "video": { "mp4": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/episode-se3labs-eval-rac__20260920-214816__ep_0000-10x.mp4", "webm": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/episode-se3labs-eval-rac__20260920-214816__ep_0000-10x.webm", "poster": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/episode-se3labs-eval-rac__20260920-214816__ep_0000-10x-poster.jpg", "speedup": 10, "width": 1280, "height": 360 }, "evidenceImage": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/evidence/se3labs-eval-rac__20260920-214816__ep_0000.jpg", "episodeNumber": 6, "displayName": "Agentic Policy - episode 6" }, { "id": "se3labs-eval-rac__20260920-215546__ep_0000", "run": "20260920-215546", "episode": "ep_0000", "approach": "agentic-policy", "cohort": "eval", "grasp": { "score": 0, "item": "No purposeful target contact, or only a distractor is grasped.", "evidence": "The gripper remains open throughout. It observes the cluttered group but makes no visible purposeful contact with the partly occluded red target.", "timestamps": [ 0, 80, 220, 260, 380, 394.2 ] }, "transport": { "score": 0, "item": "No transport beyond the initial lift, or no retained lift.", "evidence": "No target is lifted or transported.", "timestamps": [ 260, 380, 394.2 ] }, "placement": { "score": 0, "item": "Finishes outside the box after a miss or bounce, or placement is not reached.", "evidence": "No placement stage is reached; the block remains in the original cluttered group.", "timestamps": [ 380, 394.2 ] }, "complete": false, "attempts": 0, "drops": 0, "placementAttempts": 0, "confidence": "medium", "reviewNote": "The target is partly obscured by distractors. Full gripper telemetry contains no closure event, and the camera sequence shows no target acquisition.", "reviewedEvidence": [ "se3labs-eval-rac/yam_red_lego_in_dark_gray_box/20260920-215546/recording.rrd", "analysis/extract/se3labs-eval-rac__20260920-215546/series.json", "analysis/sheets/se3labs-eval-rac__20260920-215546__ep_0000.jpg", "analysis/scoring-review/agentic-sheets/20260920-215546_0-394.192_20_0.jpg" ], "total": 0, "reviewStatus": "scored", "source": "se3labs-eval-rac/yam_red_lego_in_dark_gray_box/20260920-215546/recording.rrd", "durationSeconds": 394.192, "rolloutOriginSeconds": 7.662, "terminalCode": "WALK_FAILED", "reviewer": "agentic", "ledgerLabel": "215546", "progress": 0, "ledgerIndex": 27, "modelCostUsd": 2.36, "modelCalls": 39, "activeArm": "left", "tokenUsage": { "total": 1660755, "uncachedInput": 54986, "cachedInput": 1601664, "output": 4105, "reasoning": 1364 }, "video": { "mp4": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/episode-se3labs-eval-rac__20260920-215546__ep_0000-10x.mp4", "webm": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/episode-se3labs-eval-rac__20260920-215546__ep_0000-10x.webm", "poster": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/episode-se3labs-eval-rac__20260920-215546__ep_0000-10x-poster.jpg", "speedup": 10, "width": 1280, "height": 360 }, "evidenceImage": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/evidence/se3labs-eval-rac__20260920-215546__ep_0000.jpg", "episodeNumber": 7, "displayName": "Agentic Policy - episode 7" }, { "id": "se3labs-eval-rac__20260920-222019__ep_0000", "run": "20260920-222019", "episode": "ep_0000", "approach": "agentic-policy", "cohort": "eval", "grasp": { "score": 20, "item": "After one or more failed attempts: stable grasp and lift.", "evidence": "Five acquisition attempts fail. The sixth closure at 300.8 s lifts the target and retains it for more than one second, visible at 304–309 s.", "timestamps": [ 100.3, 136.5, 174.8, 216.6, 245.6, 301, 304, 309 ] }, "transport": { "score": 10, "item": "Target dropped, reacquired, and delivered to the release pose.", "evidence": "An additional squeeze at 309.9 s ejects the already lifted target onto the table. It is reacquired on the seventh acquisition attempt at 357.6 s and delivered above the opening.", "timestamps": [ 309.9, 311, 357.6, 361, 378, 402 ] }, "placement": { "score": 40, "item": "Complete on the first placement attempt, without rim contact or correction.", "evidence": "Released once over the box at about 418.7 s; the target settles fully inside and remains at rest through the final two seconds.", "timestamps": [ 418.7, 420, 426.1, 428.1 ] }, "complete": true, "attempts": 7, "drops": 1, "placementAttempts": 1, "confidence": "high", "reviewNote": "Stage boundary: the sixth attempt meets the ≥1 s retained-lift requirement before the drop. The subsequent loss therefore receives the transport recovery row. The additional squeeze while already holding the target is an adjustment, not an eighth acquisition attempt. Native-resolution frames verify the lift and drop.", "reviewedEvidence": [ "se3labs-eval-rac/yam_red_lego_in_dark_gray_box/20260920-222019/recording.rrd", "analysis/extract/se3labs-eval-rac__20260920-222019/series.json", "analysis/sheets/se3labs-eval-rac__20260920-222019__ep_0000.jpg", "analysis/scoring-review/agentic-sheets/20260920-222019_90-428.128_12_0.jpg", "analysis/scoring-review/agentic-sheets/20260920-222019_90-428.128_12_1.jpg", "analysis/scoring-review/agentic-sheets/20260920-222019_300-318_1_0.jpg", "analysis/scoring-review/agentic-sheets/20260920-222019_0-428.128_8_0.jpg" ], "total": 70, "reviewStatus": "scored", "source": "se3labs-eval-rac/yam_red_lego_in_dark_gray_box/20260920-222019/recording.rrd", "durationSeconds": 428.128, "rolloutOriginSeconds": 16.487, "terminalCode": "COMPLETED", "reviewer": "agentic", "ledgerLabel": "222019", "progress": 100, "ledgerIndex": 28, "modelCostUsd": 5.07, "modelCalls": 51, "activeArm": "left", "tokenUsage": { "total": 3592506, "uncachedInput": 125168, "cachedInput": 3460096, "output": 7242, "reasoning": 2298 }, "video": { "mp4": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/recovery-B-222019-10x.mp4", "webm": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/recovery-B-222019-10x.webm", "poster": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/recovery-B-222019-10x-poster.jpg", "speedup": 10, "width": 1280, "height": 360 }, "evidenceImage": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/evidence/se3labs-eval-rac__20260920-222019__ep_0000.jpg", "episodeNumber": 8, "displayName": "Agentic Policy - episode 8" }, { "id": "se3labs-eval-rac__20260920-224459__ep_0000", "run": "20260920-224459", "episode": "ep_0000", "approach": "agentic-policy", "cohort": "eval", "grasp": { "score": 30, "item": "First attempt: stable grasp and lift, without corrective grip adjustment.", "evidence": "A single closure captures the target and holds it steadily through the lift.", "timestamps": [ 88.9, 90.9, 92.9 ] }, "transport": { "score": 30, "item": "Release pose reached with no slipping, drops, or unintended contact.", "evidence": "Target stays fixed in the jaws to the release pose over the box interior.", "timestamps": [ 100.9, 108.9, 125.8 ] }, "placement": { "score": 40, "item": "Complete on the first placement attempt, without rim contact or correction.", "evidence": "First release at about 130.8 s; the open gripper withdraws and the target remains fully inside over the final two seconds.", "timestamps": [ 130.8, 132.8, 149.9, 151.9 ] }, "complete": true, "attempts": 1, "drops": 0, "placementAttempts": 1, "confidence": "high", "reviewNote": "Paired topdown and active-wrist imagery reviewed at acquisition, transport, release, and rollout end. Gripper telemetry was used to locate actions, not to assign quality scores.", "reviewedEvidence": [ "se3labs-eval-rac/yam_red_lego_in_dark_gray_box/20260920-224459/recording.rrd", "analysis/extract/se3labs-eval-rac__20260920-224459/series.json", "analysis/sheets/se3labs-eval-rac__20260920-224459__ep_0000.jpg", "analysis/scoring-review/agentic-sheets/20260920-224459_0-151.926_8_0.jpg" ], "total": 100, "reviewStatus": "scored", "source": "se3labs-eval-rac/yam_red_lego_in_dark_gray_box/20260920-224459/recording.rrd", "durationSeconds": 151.926, "rolloutOriginSeconds": 12.398, "terminalCode": "COMPLETED", "reviewer": "agentic", "ledgerLabel": "224459", "progress": 100, "ledgerIndex": 29, "modelCostUsd": 1.8, "modelCalls": 24, "activeArm": "left", "tokenUsage": { "total": 1105828, "uncachedInput": 60162, "cachedInput": 1042560, "output": 3106, "reasoning": 816 }, "video": { "mp4": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/episode-se3labs-eval-rac__20260920-224459__ep_0000-10x.mp4", "webm": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/episode-se3labs-eval-rac__20260920-224459__ep_0000-10x.webm", "poster": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/episode-se3labs-eval-rac__20260920-224459__ep_0000-10x-poster.jpg", "speedup": 10, "width": 1280, "height": 360 }, "evidenceImage": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/evidence/se3labs-eval-rac__20260920-224459__ep_0000.jpg", "episodeNumber": 9, "displayName": "Agentic Policy - episode 9" }, { "id": "se3labs-eval-rac__20260920-224900__ep_0000", "run": "20260920-224900", "episode": "ep_0000", "approach": "agentic-policy", "cohort": "eval", "grasp": { "score": 30, "item": "First attempt: stable grasp and lift, without corrective grip adjustment.", "evidence": "First closure at about 95.6 s yields a stable lift, with no corrective regrip.", "timestamps": [ 95.6, 97.6, 99.6 ] }, "transport": { "score": 30, "item": "Release pose reached with no slipping, drops, or unintended contact.", "evidence": "The target remains retained through transport to an interior release pose.", "timestamps": [ 107.6, 115.6, 132.4 ] }, "placement": { "score": 40, "item": "Complete on the first placement attempt, without rim contact or correction.", "evidence": "First release at about 137.4 s lands inside; the target remains still at rollout end.", "timestamps": [ 137.4, 139.4, 147.1, 149.1 ] }, "complete": true, "attempts": 1, "drops": 0, "placementAttempts": 1, "confidence": "high", "reviewNote": "Paired topdown and active-wrist imagery reviewed at acquisition, transport, release, and rollout end. Gripper telemetry was used to locate actions, not to assign quality scores.", "reviewedEvidence": [ "se3labs-eval-rac/yam_red_lego_in_dark_gray_box/20260920-224900/recording.rrd", "analysis/extract/se3labs-eval-rac__20260920-224900/series.json", "analysis/sheets/se3labs-eval-rac__20260920-224900__ep_0000.jpg", "analysis/scoring-review/agentic-sheets/20260920-224900_0-149.096_8_0.jpg" ], "total": 100, "reviewStatus": "scored", "source": "se3labs-eval-rac/yam_red_lego_in_dark_gray_box/20260920-224900/recording.rrd", "durationSeconds": 149.096, "rolloutOriginSeconds": 13.422, "terminalCode": "WALK_FAILED", "reviewer": "agentic", "ledgerLabel": "224900", "progress": 100, "ledgerIndex": 30, "modelCostUsd": 1.71, "modelCalls": 32, "activeArm": "left", "tokenUsage": { "total": 1140166, "uncachedInput": 47850, "cachedInput": 1089536, "output": 2780, "reasoning": 652 }, "video": { "mp4": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/episode-se3labs-eval-rac__20260920-224900__ep_0000-10x.mp4", "webm": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/episode-se3labs-eval-rac__20260920-224900__ep_0000-10x.webm", "poster": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/episode-se3labs-eval-rac__20260920-224900__ep_0000-10x-poster.jpg", "speedup": 10, "width": 1280, "height": 360 }, "evidenceImage": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/evidence/se3labs-eval-rac__20260920-224900__ep_0000.jpg", "episodeNumber": 10, "displayName": "Agentic Policy - episode 10" }, { "id": "se3labs-eval-rac__20260920-230058__ep_0000", "run": "20260920-230058", "episode": "ep_0000", "approach": "agentic-policy", "cohort": "eval", "grasp": { "score": 30, "item": "First attempt: stable grasp and lift, without corrective grip adjustment.", "evidence": "First closure securely grips the target; it remains fixed during lift.", "timestamps": [ 125.8, 127.8, 129.8 ] }, "transport": { "score": 30, "item": "Release pose reached with no slipping, drops, or unintended contact.", "evidence": "Stable carry to a pose fully above the box opening.", "timestamps": [ 137.8, 145.8, 168.3 ] }, "placement": { "score": 40, "item": "Complete on the first placement attempt, without rim contact or correction.", "evidence": "Released on the first placement attempt at about 173.3 s; the target remains fully inside and settled over the final two seconds.", "timestamps": [ 173.3, 175.3, 183.2, 185.2 ] }, "complete": true, "attempts": 1, "drops": 0, "placementAttempts": 1, "confidence": "high", "reviewNote": "Paired topdown and active-wrist imagery reviewed at acquisition, transport, release, and rollout end. Gripper telemetry was used to locate actions, not to assign quality scores.", "reviewedEvidence": [ "se3labs-eval-rac/yam_red_lego_in_dark_gray_box/20260920-230058/recording.rrd", "analysis/extract/se3labs-eval-rac__20260920-230058/series.json", "analysis/sheets/se3labs-eval-rac__20260920-230058__ep_0000.jpg", "analysis/scoring-review/agentic-sheets/20260920-230058_0-185.155_8_0.jpg" ], "total": 100, "reviewStatus": "scored", "source": "se3labs-eval-rac/yam_red_lego_in_dark_gray_box/20260920-230058/recording.rrd", "durationSeconds": 185.155, "rolloutOriginSeconds": 10.358, "terminalCode": "COMPLETED", "reviewer": "agentic", "ledgerLabel": "230058", "progress": 100, "ledgerIndex": 31, "modelCostUsd": 2.11, "modelCalls": 32, "activeArm": "right", "tokenUsage": { "total": 1399134, "uncachedInput": 60803, "cachedInput": 1335040, "output": 3291, "reasoning": 1101 }, "video": { "mp4": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/episode-se3labs-eval-rac__20260920-230058__ep_0000-10x.mp4", "webm": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/episode-se3labs-eval-rac__20260920-230058__ep_0000-10x.webm", "poster": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/episode-se3labs-eval-rac__20260920-230058__ep_0000-10x-poster.jpg", "speedup": 10, "width": 1280, "height": 360 }, "evidenceImage": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/evidence/se3labs-eval-rac__20260920-230058__ep_0000.jpg", "episodeNumber": 11, "displayName": "Agentic Policy - episode 11" }, { "id": "se3labs-eval-rac__20260920-230618__ep_0000", "run": "20260920-230618", "episode": "ep_0000", "approach": "agentic-policy", "cohort": "eval", "grasp": { "score": 30, "item": "First attempt: stable grasp and lift, without corrective grip adjustment.", "evidence": "First closure grips and lifts the target without visible slip or grip correction.", "timestamps": [ 96.2, 98.2, 100.2 ] }, "transport": { "score": 30, "item": "Release pose reached with no slipping, drops, or unintended contact.", "evidence": "The target remains retained through transport to the box interior.", "timestamps": [ 108.2, 116.2, 134.9 ] }, "placement": { "score": 40, "item": "Complete on the first placement attempt, without rim contact or correction.", "evidence": "Clean first release at about 139.9 s; the block is fully inside and at rest at the end.", "timestamps": [ 139.9, 141.9, 150.2, 152.2 ] }, "complete": true, "attempts": 1, "drops": 0, "placementAttempts": 1, "confidence": "high", "reviewNote": "Paired topdown and active-wrist imagery reviewed at acquisition, transport, release, and rollout end. Gripper telemetry was used to locate actions, not to assign quality scores.", "reviewedEvidence": [ "se3labs-eval-rac/yam_red_lego_in_dark_gray_box/20260920-230618/recording.rrd", "analysis/extract/se3labs-eval-rac__20260920-230618/series.json", "analysis/sheets/se3labs-eval-rac__20260920-230618__ep_0000.jpg", "analysis/scoring-review/agentic-sheets/20260920-230618_0-152.245_8_0.jpg" ], "total": 100, "reviewStatus": "scored", "source": "se3labs-eval-rac/yam_red_lego_in_dark_gray_box/20260920-230618/recording.rrd", "durationSeconds": 152.245, "rolloutOriginSeconds": 18.903, "terminalCode": "COMPLETED", "reviewer": "agentic", "ledgerLabel": "230618", "progress": 100, "ledgerIndex": 32, "modelCostUsd": 2.05, "modelCalls": 30, "activeArm": "right", "tokenUsage": { "total": 1275515, "uncachedInput": 71654, "cachedInput": 1201280, "output": 2581, "reasoning": 563 }, "video": { "mp4": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/episode-se3labs-eval-rac__20260920-230618__ep_0000-10x.mp4", "webm": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/episode-se3labs-eval-rac__20260920-230618__ep_0000-10x.webm", "poster": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/episode-se3labs-eval-rac__20260920-230618__ep_0000-10x-poster.jpg", "speedup": 10, "width": 1280, "height": 360 }, "evidenceImage": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/evidence/se3labs-eval-rac__20260920-230618__ep_0000.jpg", "episodeNumber": 12, "displayName": "Agentic Policy - episode 12" }, { "id": "se3labs-eval-rac__20260920-230929__ep_0000", "run": "20260920-230929", "episode": "ep_0000", "approach": "agentic-policy", "cohort": "eval", "grasp": { "score": 20, "item": "After one or more failed attempts: stable grasp and lift.", "evidence": "The initial partial closure at 148.9 s and renewed full closure at 161.9 s fail to lift the target. After reopening, the third acquisition closure at 230.9 s produces a stable lift.", "timestamps": [ 148.9, 161.9, 179.5, 230.9, 233, 236 ] }, "transport": { "score": 30, "item": "Release pose reached with no slipping, drops, or unintended contact.", "evidence": "The successful grasp remains stable through the traverse and interior release pose.", "timestamps": [ 242, 255, 270 ] }, "placement": { "score": 40, "item": "Complete on the first placement attempt, without rim contact or correction.", "evidence": "First release at about 273.4 s lands fully inside; the target rests through the last two seconds.", "timestamps": [ 273.4, 275, 284.8, 286.8 ] }, "complete": true, "attempts": 3, "drops": 0, "placementAttempts": 1, "confidence": "medium", "reviewNote": "The first partial acquisition closure is counted separately from the later empty full closure because it is a distinct acquisition action.", "reviewedEvidence": [ "se3labs-eval-rac/yam_red_lego_in_dark_gray_box/20260920-230929/recording.rrd", "analysis/extract/se3labs-eval-rac__20260920-230929/series.json", "analysis/sheets/se3labs-eval-rac__20260920-230929__ep_0000.jpg", "analysis/scoring-review/agentic-sheets/20260920-230929_140-286.881_6_0.jpg", "analysis/scoring-review/agentic-sheets/20260920-230929_140-286.881_6_1.jpg", "analysis/scoring-review/agentic-sheets/20260920-230929_0-286.881_8_0.jpg" ], "total": 90, "reviewStatus": "scored", "source": "se3labs-eval-rac/yam_red_lego_in_dark_gray_box/20260920-230929/recording.rrd", "durationSeconds": 286.881, "rolloutOriginSeconds": 21.803, "terminalCode": "COMPLETED", "reviewer": "agentic", "ledgerLabel": "230929", "progress": 100, "ledgerIndex": 33, "modelCostUsd": 3.12, "modelCalls": 47, "activeArm": "right", "tokenUsage": { "total": 2236377, "uncachedInput": 70178, "cachedInput": 2161024, "output": 5175, "reasoning": 1938 }, "video": { "mp4": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/episode-se3labs-eval-rac__20260920-230929__ep_0000-10x.mp4", "webm": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/episode-se3labs-eval-rac__20260920-230929__ep_0000-10x.webm", "poster": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/episode-se3labs-eval-rac__20260920-230929__ep_0000-10x-poster.jpg", "speedup": 10, "width": 1280, "height": 360 }, "evidenceImage": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/evidence/se3labs-eval-rac__20260920-230929__ep_0000.jpg", "episodeNumber": 13, "displayName": "Agentic Policy - episode 13" }, { "id": "se3labs-eval-rac__20260920-231850__ep_0000", "run": "20260920-231850", "episode": "ep_0000", "approach": "agentic-policy", "cohort": "eval", "grasp": { "score": 30, "item": "First attempt: stable grasp and lift, without corrective grip adjustment.", "evidence": "The first closure separates and lifts the red target from the surrounding distractors; it remains stable in the jaws.", "timestamps": [ 101, 103, 105 ] }, "transport": { "score": 30, "item": "Release pose reached with no slipping, drops, or unintended contact.", "evidence": "The retained target is carried to a release pose over the box interior, without visible slippage or drop.", "timestamps": [ 113, 121, 150.1 ] }, "placement": { "score": 40, "item": "Complete on the first placement attempt, without rim contact or correction.", "evidence": "First release at about 155.1 s leaves the target fully inside and stationary over the final two seconds.", "timestamps": [ 155.1, 157.1, 165.5, 167.5 ] }, "complete": true, "attempts": 1, "drops": 0, "placementAttempts": 1, "confidence": "high", "reviewNote": "Nearby distractors move during acquisition, but no continuing target instability or transport mishap is visible.", "reviewedEvidence": [ "se3labs-eval-rac/yam_red_lego_in_dark_gray_box/20260920-231850/recording.rrd", "analysis/extract/se3labs-eval-rac__20260920-231850/series.json", "analysis/sheets/se3labs-eval-rac__20260920-231850__ep_0000.jpg", "analysis/scoring-review/agentic-sheets/20260920-231850_0-167.452_8_0.jpg" ], "total": 100, "reviewStatus": "scored", "source": "se3labs-eval-rac/yam_red_lego_in_dark_gray_box/20260920-231850/recording.rrd", "durationSeconds": 167.452, "rolloutOriginSeconds": 33.501, "terminalCode": "COMPLETED", "reviewer": "agentic", "ledgerLabel": "231850", "progress": 100, "ledgerIndex": 34, "modelCostUsd": 2.25, "modelCalls": 34, "activeArm": "left", "tokenUsage": { "total": 1507923, "uncachedInput": 66099, "cachedInput": 1438720, "output": 3104, "reasoning": 843 }, "video": { "mp4": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/episode-se3labs-eval-rac__20260920-231850__ep_0000-10x.mp4", "webm": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/episode-se3labs-eval-rac__20260920-231850__ep_0000-10x.webm", "poster": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/episode-se3labs-eval-rac__20260920-231850__ep_0000-10x-poster.jpg", "speedup": 10, "width": 1280, "height": 360 }, "evidenceImage": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/evidence/se3labs-eval-rac__20260920-231850__ep_0000.jpg", "episodeNumber": 14, "displayName": "Agentic Policy - episode 14" }, { "id": "se3labs-eval-rac__20260920-232318__ep_0000", "run": "20260920-232318", "episode": "ep_0000", "approach": "agentic-policy", "cohort": "eval", "grasp": { "score": 30, "item": "First attempt: stable grasp and lift, without corrective grip adjustment.", "evidence": "First closure grips the target in its current orientation and holds it steady through lift.", "timestamps": [ 148.2, 150.2, 152.2 ] }, "transport": { "score": 30, "item": "Release pose reached with no slipping, drops, or unintended contact.", "evidence": "Stable carry to the interior release pose.", "timestamps": [ 160.2, 168.2, 172.9 ] }, "placement": { "score": 40, "item": "Complete on the first placement attempt, without rim contact or correction.", "evidence": "First release at about 177.9 s; the target rests fully inside the box through the last two seconds.", "timestamps": [ 177.9, 179.9, 185.3, 187.3 ] }, "complete": true, "attempts": 1, "drops": 0, "placementAttempts": 1, "confidence": "high", "reviewNote": "The target is acquired with its underside facing the wrist view; orientation alone is not instability.", "reviewedEvidence": [ "se3labs-eval-rac/yam_red_lego_in_dark_gray_box/20260920-232318/recording.rrd", "analysis/extract/se3labs-eval-rac__20260920-232318/series.json", "analysis/sheets/se3labs-eval-rac__20260920-232318__ep_0000.jpg", "analysis/scoring-review/agentic-sheets/20260920-232318_0-187.282_8_0.jpg" ], "total": 100, "reviewStatus": "scored", "source": "se3labs-eval-rac/yam_red_lego_in_dark_gray_box/20260920-232318/recording.rrd", "durationSeconds": 187.282, "rolloutOriginSeconds": 10.321, "terminalCode": "COMPLETED", "reviewer": "agentic", "ledgerLabel": "232318", "progress": 100, "ledgerIndex": 35, "modelCostUsd": 2.33, "modelCalls": 32, "activeArm": "left", "tokenUsage": { "total": 1510694, "uncachedInput": 71272, "cachedInput": 1435776, "output": 3646, "reasoning": 733 }, "video": { "mp4": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/episode-se3labs-eval-rac__20260920-232318__ep_0000-10x.mp4", "webm": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/episode-se3labs-eval-rac__20260920-232318__ep_0000-10x.webm", "poster": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/episode-se3labs-eval-rac__20260920-232318__ep_0000-10x-poster.jpg", "speedup": 10, "width": 1280, "height": 360 }, "evidenceImage": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/evidence/se3labs-eval-rac__20260920-232318__ep_0000.jpg", "episodeNumber": 15, "displayName": "Agentic Policy - episode 15" }, { "id": "se3labs-eval-rac__20260920-232815__ep_0000", "run": "20260920-232815", "episode": "ep_0000", "approach": "agentic-policy", "cohort": "eval", "grasp": { "score": 30, "item": "First attempt: stable grasp and lift, without corrective grip adjustment.", "evidence": "The first closure captures and lifts the red target with no visible corrective adjustment.", "timestamps": [ 122.5, 124.5, 126.5 ] }, "transport": { "score": 30, "item": "Release pose reached with no slipping, drops, or unintended contact.", "evidence": "The target remains stable through the carry and reaches the box interior.", "timestamps": [ 134.5, 142.5, 155.8 ] }, "placement": { "score": 40, "item": "Complete on the first placement attempt, without rim contact or correction.", "evidence": "First release at about 160.8 s completes placement; the block remains still and fully inside over the final two seconds.", "timestamps": [ 160.8, 162.8, 172.2, 174.2 ] }, "complete": true, "attempts": 1, "drops": 0, "placementAttempts": 1, "confidence": "high", "reviewNote": "Paired topdown and active-wrist imagery reviewed at acquisition, transport, release, and rollout end. Gripper telemetry was used to locate actions, not to assign quality scores.", "reviewedEvidence": [ "se3labs-eval-rac/yam_red_lego_in_dark_gray_box/20260920-232815/recording.rrd", "analysis/extract/se3labs-eval-rac__20260920-232815/series.json", "analysis/sheets/se3labs-eval-rac__20260920-232815__ep_0000.jpg", "analysis/scoring-review/agentic-sheets/20260920-232815_0-174.234_8_0.jpg" ], "total": 100, "reviewStatus": "scored", "source": "se3labs-eval-rac/yam_red_lego_in_dark_gray_box/20260920-232815/recording.rrd", "durationSeconds": 174.234, "rolloutOriginSeconds": 84.348, "terminalCode": "COMPLETED", "reviewer": "agentic", "ledgerLabel": "232815", "progress": 100, "ledgerIndex": 36, "modelCostUsd": 2.37, "modelCalls": 41, "activeArm": "right", "tokenUsage": { "total": 1550287, "uncachedInput": 71968, "cachedInput": 1474816, "output": 3503, "reasoning": 948 }, "video": { "mp4": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/episode-se3labs-eval-rac__20260920-232815__ep_0000-10x.mp4", "webm": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/episode-se3labs-eval-rac__20260920-232815__ep_0000-10x.webm", "poster": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/episode-se3labs-eval-rac__20260920-232815__ep_0000-10x-poster.jpg", "speedup": 10, "width": 1280, "height": 360 }, "evidenceImage": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/evidence/se3labs-eval-rac__20260920-232815__ep_0000.jpg", "episodeNumber": 16, "displayName": "Agentic Policy - episode 16" }, { "id": "se3labs-eval-rac__20260920-233401__ep_0000", "run": "20260920-233401", "episode": "ep_0000", "approach": "agentic-policy", "cohort": "eval", "grasp": { "score": 20, "item": "After one or more failed attempts: stable grasp and lift.", "evidence": "The partial closure at 85.6 s is followed by an empty lift. The robot repositions and closes further at 129.2 s, acquiring and retaining the target steadily.", "timestamps": [ 85.6, 104, 116, 129.2, 132, 136 ] }, "transport": { "score": 30, "item": "Release pose reached with no slipping, drops, or unintended contact.", "evidence": "After acquisition the target remains stable during the carry to the box interior.", "timestamps": [ 144, 156, 171.3 ] }, "placement": { "score": 40, "item": "Complete on the first placement attempt, without rim contact or correction.", "evidence": "First placement release at about 176.3 s leaves the target fully inside and stationary at the end.", "timestamps": [ 176.3, 178.3, 181.3, 183.3 ] }, "complete": true, "attempts": 2, "drops": 0, "placementAttempts": 1, "confidence": "medium", "reviewNote": "Two distinct acquisition actions are visible even though the gripper does not reopen fully between them. The first action never retains the target; the second is not an adjustment of an existing grasp.", "reviewedEvidence": [ "se3labs-eval-rac/yam_red_lego_in_dark_gray_box/20260920-233401/recording.rrd", "analysis/extract/se3labs-eval-rac__20260920-233401/series.json", "analysis/sheets/se3labs-eval-rac__20260920-233401__ep_0000.jpg", "analysis/scoring-review/agentic-sheets/20260920-233401_0-183.29_8_0.jpg", "analysis/scoring-review/agentic-sheets/20260920-233401_96-176_4_0.jpg" ], "total": 90, "reviewStatus": "scored", "source": "se3labs-eval-rac/yam_red_lego_in_dark_gray_box/20260920-233401/recording.rrd", "durationSeconds": 183.29, "rolloutOriginSeconds": 22.73, "terminalCode": "COMPLETED", "reviewer": "agentic", "ledgerLabel": "233401", "progress": 100, "ledgerIndex": 37, "modelCostUsd": 2.44, "modelCalls": 31, "activeArm": "right", "tokenUsage": { "total": 1586145, "uncachedInput": 73095, "cachedInput": 1508992, "output": 4058, "reasoning": 1007 }, "video": { "mp4": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/episode-se3labs-eval-rac__20260920-233401__ep_0000-10x.mp4", "webm": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/episode-se3labs-eval-rac__20260920-233401__ep_0000-10x.webm", "poster": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/episode-se3labs-eval-rac__20260920-233401__ep_0000-10x-poster.jpg", "speedup": 10, "width": 1280, "height": 360 }, "evidenceImage": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/evidence/se3labs-eval-rac__20260920-233401__ep_0000.jpg", "episodeNumber": 17, "displayName": "Agentic Policy - episode 17" }, { "id": "se3labs-eval-rac__20260920-234503__ep_0000", "run": "20260920-234503", "episode": "ep_0000", "approach": "agentic-policy", "cohort": "eval", "grasp": { "score": 20, "item": "After one or more failed attempts: stable grasp and lift.", "evidence": "Initial closure misses the target; a second closure at 116.2 s yields a steady retained lift.", "timestamps": [ 85.8, 92, 97.7, 116.2, 120, 124 ] }, "transport": { "score": 30, "item": "Release pose reached with no slipping, drops, or unintended contact.", "evidence": "The target stays fixed in the jaws through the traverse to the box opening.", "timestamps": [ 132, 148, 160 ] }, "placement": { "score": 40, "item": "Complete on the first placement attempt, without rim contact or correction.", "evidence": "Release at about 163.4 s falls into the interior without visible rim contact. The target remains inside at rest through rollout end.", "timestamps": [ 163.4, 164.4, 168, 172, 175.2 ] }, "complete": true, "attempts": 2, "drops": 0, "placementAttempts": 1, "confidence": "medium", "reviewNote": "A dense 0.4 s paired-camera review of 158–165 s resolves the release close to the box edge.", "reviewedEvidence": [ "se3labs-eval-rac/yam_red_lego_in_dark_gray_box/20260920-234503/recording.rrd", "analysis/extract/se3labs-eval-rac__20260920-234503/series.json", "analysis/sheets/se3labs-eval-rac__20260920-234503__ep_0000.jpg", "analysis/scoring-review/agentic-sheets/20260920-234503_80-175.215_4_0.jpg", "analysis/scoring-review/agentic-sheets/20260920-234503_80-175.215_4_1.jpg", "analysis/scoring-review/agentic-sheets/20260920-234503_158-165_0.4_0.jpg" ], "total": 90, "reviewStatus": "scored", "source": "se3labs-eval-rac/yam_red_lego_in_dark_gray_box/20260920-234503/recording.rrd", "durationSeconds": 175.215, "rolloutOriginSeconds": 158.726, "terminalCode": "COMPLETED", "reviewer": "agentic", "ledgerLabel": "234503", "progress": 100, "ledgerIndex": 38, "modelCostUsd": 2.27, "modelCalls": 36, "activeArm": "right", "tokenUsage": { "total": 1602436, "uncachedInput": 59100, "cachedInput": 1540480, "output": 2856, "reasoning": 581 }, "video": { "mp4": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/episode-se3labs-eval-rac__20260920-234503__ep_0000-10x.mp4", "webm": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/episode-se3labs-eval-rac__20260920-234503__ep_0000-10x.webm", "poster": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/episode-se3labs-eval-rac__20260920-234503__ep_0000-10x-poster.jpg", "speedup": 10, "width": 1280, "height": 360 }, "evidenceImage": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/evidence/se3labs-eval-rac__20260920-234503__ep_0000.jpg", "episodeNumber": 18, "displayName": "Agentic Policy - episode 18" }, { "id": "se3labs-eval-rac__20260920-235124__ep_0000", "run": "20260920-235124", "episode": "ep_0000", "approach": "agentic-policy", "cohort": "eval", "grasp": { "score": 20, "item": "After one or more failed attempts: stable grasp and lift.", "evidence": "The first attempt knocks the target away, and the second closes without retaining it. The third closure at 197.8 s securely lifts the target.", "timestamps": [ 99.8, 106, 164.5, 176, 197.8, 201, 204 ] }, "transport": { "score": 30, "item": "Release pose reached with no slipping, drops, or unintended contact.", "evidence": "After the third pickup, stable carry reaches an interior release pose without dropping the target.", "timestamps": [ 218, 232, 243 ] }, "placement": { "score": 40, "item": "Complete on the first placement attempt, without rim contact or correction.", "evidence": "First placement release at about 244 s lands fully inside; the target is stationary throughout the final two seconds.", "timestamps": [ 244, 246, 254.7, 256.7 ] }, "complete": true, "attempts": 3, "drops": 0, "placementAttempts": 1, "confidence": "high", "reviewNote": "The first displacement is a failed acquisition, not a transport drop: no retained lift precedes it.", "reviewedEvidence": [ "se3labs-eval-rac/yam_red_lego_in_dark_gray_box/20260920-235124/recording.rrd", "analysis/extract/se3labs-eval-rac__20260920-235124/series.json", "analysis/sheets/se3labs-eval-rac__20260920-235124__ep_0000.jpg", "analysis/scoring-review/agentic-sheets/20260920-235124_85-256.705_7_0.jpg", "analysis/scoring-review/agentic-sheets/20260920-235124_85-256.705_7_1.jpg", "analysis/scoring-review/agentic-sheets/20260920-235124_0-256.705_8_0.jpg" ], "total": 90, "reviewStatus": "scored", "source": "se3labs-eval-rac/yam_red_lego_in_dark_gray_box/20260920-235124/recording.rrd", "durationSeconds": 256.705, "rolloutOriginSeconds": 44.716, "terminalCode": "COMPLETED", "reviewer": "agentic", "ledgerLabel": "235124", "progress": 100, "ledgerIndex": 39, "modelCostUsd": 3.36, "modelCalls": 49, "activeArm": "right", "tokenUsage": { "total": 2373709, "uncachedInput": 83536, "cachedInput": 2285440, "output": 4733, "reasoning": 1377 }, "video": { "mp4": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/episode-se3labs-eval-rac__20260920-235124__ep_0000-10x.mp4", "webm": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/episode-se3labs-eval-rac__20260920-235124__ep_0000-10x.webm", "poster": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/episode-se3labs-eval-rac__20260920-235124__ep_0000-10x-poster.jpg", "speedup": 10, "width": 1280, "height": 360 }, "evidenceImage": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/evidence/se3labs-eval-rac__20260920-235124__ep_0000.jpg", "episodeNumber": 19, "displayName": "Agentic Policy - episode 19" }, { "id": "se3labs-eval-rac__20260921-000118__ep_0000", "run": "20260921-000118", "episode": "ep_0000", "approach": "agentic-policy", "cohort": "eval", "grasp": { "score": 20, "item": "After one or more failed attempts: stable grasp and lift.", "evidence": "The first closure misses. The second at 202.8 s captures and lifts the target steadily.", "timestamps": [ 161.6, 170, 178.7, 202.8, 206, 210 ] }, "transport": { "score": 30, "item": "Release pose reached with no slipping, drops, or unintended contact.", "evidence": "Stable retained carry to the release pose over the box interior.", "timestamps": [ 225, 235, 252 ] }, "placement": { "score": 40, "item": "Complete on the first placement attempt, without rim contact or correction.", "evidence": "First release at about 253.4 s; the block remains fully inside and at rest over the final two seconds.", "timestamps": [ 253.4, 255, 265.2, 267.2 ] }, "complete": true, "attempts": 2, "drops": 0, "placementAttempts": 1, "confidence": "high", "reviewNote": "Paired topdown and active-wrist imagery reviewed at acquisition, transport, release, and rollout end. Gripper telemetry was used to locate actions, not to assign quality scores.", "reviewedEvidence": [ "se3labs-eval-rac/yam_red_lego_in_dark_gray_box/20260921-000118/recording.rrd", "analysis/extract/se3labs-eval-rac__20260921-000118/series.json", "analysis/sheets/se3labs-eval-rac__20260921-000118__ep_0000.jpg", "analysis/scoring-review/agentic-sheets/20260921-000118_150-267.285_5_0.jpg", "analysis/scoring-review/agentic-sheets/20260921-000118_150-267.285_5_1.jpg", "analysis/scoring-review/agentic-sheets/20260921-000118_0-267.285_8_0.jpg" ], "total": 90, "reviewStatus": "scored", "source": "se3labs-eval-rac/yam_red_lego_in_dark_gray_box/20260921-000118/recording.rrd", "durationSeconds": 267.285, "rolloutOriginSeconds": 11.757, "terminalCode": "COMPLETED", "reviewer": "agentic", "ledgerLabel": "000118", "progress": 100, "ledgerIndex": 40, "modelCostUsd": 3.51, "modelCalls": 48, "activeArm": "right", "tokenUsage": { "total": 2583275, "uncachedInput": 77516, "cachedInput": 2500992, "output": 4767, "reasoning": 1601 }, "video": { "mp4": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/episode-se3labs-eval-rac__20260921-000118__ep_0000-10x.mp4", "webm": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/episode-se3labs-eval-rac__20260921-000118__ep_0000-10x.webm", "poster": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/episode-se3labs-eval-rac__20260921-000118__ep_0000-10x-poster.jpg", "speedup": 10, "width": 1280, "height": 360 }, "evidenceImage": "https://data.se3labs.ai/samples_v1/blog_post/agentic-vs-code-as-policy/evidence/se3labs-eval-rac__20260921-000118__ep_0000.jpg", "episodeNumber": 20, "displayName": "Agentic Policy - episode 20" } ] }