RoboDojo: task videos from agent trials, replayed and path traced (masters)

masters/robodojo-agent/align_blocks/meta.json

@@ -0,0 +1,64 @@
+{
+ "suite": "robodojo",
+ "task": "align_blocks",
+ "task_dir": "robodojo-align-blocks-i00-standard",
+ "task_version": "mll-lab-nu/embodied-first-exam@d95805a999",
+ "source": "agent-success:codex-gpt6_astra-high/robodojo-align-blocks-i00-standa__NqbzJyu",
+ "renderer": "Isaac Sim 5.1 RTX PathTracing (rendermode PathTracing): 8 spp per frame, max 6 bounces, OptiX denoiser, firefly filter; native 2560x1440; one dedicated showcase camera (the robots' own cameras are not used)",
+ "rendering": "path-tracing",
+ "width": 2560,
+ "height": 1440,
+ "fps": 30,
+ "frames": 412,
+ "seconds": 13.73,
+ "speed": 2.4,
+ "success_check": true,
+ "success_check_how": "the task's own check list (RoboDojo's reward manager through rcb_robodojo.Session) at the end of this render's replay: every recorded action row from the task's frozen instance, as the grader replays it; rendering never steps physics",
+ "end_check": {
+  "success": true,
+  "end_reason": "success",
+  "actions": 701,
+  "state_sha": "3d2e7ad33fb1d38c",
+  "start_state_sha": "4922a5532cb00539",
+  "start_matches_instance": true
+ },
+ "replay_hash_check": "end-state hash 3d2e7ad33fb1d38c = the live episode's (final.json live_sha)",
+ "replay": {
+  "trajectory": "verifier/trajectory.npz (recorded by the simulator service during the live episode)",
+  "n_actions": 701,
+  "control_hz": 25,
+  "every_control_steps": 2,
+  "original_verdict": "verifier reward 1.0"
+ },
+ "camera": {
+  "az_deg": 35,
+  "el_deg": 40,
+  "dist_m": 1.2,
+  "focal_mm": 24,
+  "h_aperture_mm": 36.0,
+  "position": [
+   0.5025,
+   0.478,
+   1.5488
+  ],
+  "target": [
+   -0.0248,
+   -0.275,
+   0.7774
+  ]
+ },
+ "holds": {
+  "start_s": 0.5,
+  "end_s": 1.5,
+  "note": "first/last frame held, no extra simulation"
+ },
+ "image": "ghcr.io/mll-lab-nu/efe-private:robodojo-0.2.1 + rcb_robodojo.py of efe:robodojo-files-0.5.0",
+ "agent": "Codex + GPT-6 Astra (high)",
+ "run": "codex-gpt6_astra-high",
+ "mode": "standard",
+ "trial_choice": "a GPT-6 Astra success in standard mode (first choice)",
+ "poster": "last frame (the episode's end state)",
+ "encode": "libx264 preset slow CRF 16, yuv420p, bt709, +faststart",
+ "replaces": "scene-orbit / earlier agent render at masters/robodojo/align_blocks",
+ "created": "2026-11-02T07:40:09-06:00"
+}

masters/robodojo-agent/align_blocks/video.mp4

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:14b2fbc57b441a7abc7f99366d2b08013f383c185bd3d3dcdb82d93b4485a68c
+size 8285080

masters/robodojo-agent/arrange_largest_number/meta.json

@@ -0,0 +1,64 @@
+{
+ "suite": "robodojo",
+ "task": "arrange_largest_number",
+ "task_dir": "robodojo-arrange-largest-number-i00-privileged",
+ "task_version": "mll-lab-nu/embodied-first-exam@d95805a999",
+ "source": "agent-success:codex-gpt6_luna-xhigh/robodojo-arrange-largest-number__jSAFZjR",
+ "renderer": "Isaac Sim 5.1 RTX PathTracing (rendermode PathTracing): 8 spp per frame, max 6 bounces, OptiX denoiser, firefly filter; native 2560x1440; one dedicated showcase camera (the robots' own cameras are not used)",
+ "rendering": "path-tracing",
+ "width": 2560,
+ "height": 1440,
+ "fps": 30,
+ "frames": 580,
+ "seconds": 19.33,
+ "speed": 2.4,
+ "success_check": true,
+ "success_check_how": "the task's own check list (RoboDojo's reward manager through rcb_robodojo.Session) at the end of this render's replay: every recorded action row from the task's frozen instance, as the grader replays it; rendering never steps physics",
+ "end_check": {
+  "success": true,
+  "end_reason": "success",
+  "actions": 1038,
+  "state_sha": "e359d9259804e2fd",
+  "start_state_sha": "460e7dc4bed8737a",
+  "start_matches_instance": true
+ },
+ "replay_hash_check": "privileged trial: graded by replaying the submitted trajectory; this replay reaches the same verdict",
+ "replay": {
+  "trajectory": "artifacts/app/output/trajectory.npz (the submitted trajectory)",
+  "n_actions": 1038,
+  "control_hz": 25,
+  "every_control_steps": 2,
+  "original_verdict": "verifier reward 1.0"
+ },
+ "camera": {
+  "az_deg": 35,
+  "el_deg": 40,
+  "dist_m": 1.2,
+  "focal_mm": 24,
+  "h_aperture_mm": 36.0,
+  "position": [
+   0.4943,
+   0.5072,
+   1.5408
+  ],
+  "target": [
+   -0.033,
+   -0.2458,
+   0.7694
+  ]
+ },
+ "holds": {
+  "start_s": 0.5,
+  "end_s": 1.5,
+  "note": "first/last frame held, no extra simulation"
+ },
+ "image": "ghcr.io/mll-lab-nu/efe-private:robodojo-0.2.1 + rcb_robodojo.py of efe:robodojo-files-0.5.0",
+ "agent": "Codex + GPT-6 Luna (xhigh)",
+ "run": "codex-gpt6_luna-xhigh",
+ "mode": "privileged",
+ "trial_choice": "GPT-6 Luna's success: the only model that solved this task",
+ "poster": "last frame (the episode's end state)",
+ "encode": "libx264 preset slow CRF 16, yuv420p, bt709, +faststart",
+ "replaces": "scene-orbit / earlier agent render at masters/robodojo/arrange_largest_number",
+ "created": "2026-11-02T06:05:23-06:00"
+}

masters/robodojo-agent/arrange_largest_number/video.mp4

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:492f598be63d1ccec57eed6e67a6c4848f79709903d741e653528a2e206825cf
+size 13286976

masters/robodojo-agent/arrange_largest_number_random/meta.json

@@ -0,0 +1,64 @@
+{
+ "suite": "robodojo",
+ "task": "arrange_largest_number_random",
+ "task_dir": "robodojo-arrange-largest-number-random-i00-standard",
+ "task_version": "mll-lab-nu/embodied-first-exam@d95805a999",
+ "source": "agent-success:codex-gpt6_astra-high/robodojo-arrange-largest-number__tfZ3omN",
+ "renderer": "Isaac Sim 5.1 RTX PathTracing (rendermode PathTracing): 8 spp per frame, max 6 bounces, OptiX denoiser, firefly filter; native 2560x1440; one dedicated showcase camera (the robots' own cameras are not used)",
+ "rendering": "path-tracing",
+ "width": 2560,
+ "height": 1440,
+ "fps": 30,
+ "frames": 641,
+ "seconds": 21.37,
+ "speed": 2.4,
+ "success_check": true,
+ "success_check_how": "the task's own check list (RoboDojo's reward manager through rcb_robodojo.Session) at the end of this render's replay: every recorded action row from the task's frozen instance, as the grader replays it; rendering never steps physics",
+ "end_check": {
+  "success": true,
+  "end_reason": "success",
+  "actions": 1159,
+  "state_sha": "8ef36b83698ecf08",
+  "start_state_sha": "543f4432767b5372",
+  "start_matches_instance": true
+ },
+ "replay_hash_check": "end-state hash 8ef36b83698ecf08 = the live episode's (final.json live_sha)",
+ "replay": {
+  "trajectory": "verifier/trajectory.npz (recorded by the simulator service during the live episode)",
+  "n_actions": 1159,
+  "control_hz": 25,
+  "every_control_steps": 2,
+  "original_verdict": "verifier reward 1.0"
+ },
+ "camera": {
+  "az_deg": 35,
+  "el_deg": 40,
+  "dist_m": 1.2,
+  "focal_mm": 24,
+  "h_aperture_mm": 36.0,
+  "position": [
+   0.5223,
+   0.4976,
+   1.5409
+  ],
+  "target": [
+   -0.005,
+   -0.2554,
+   0.7696
+  ]
+ },
+ "holds": {
+  "start_s": 0.5,
+  "end_s": 1.5,
+  "note": "first/last frame held, no extra simulation"
+ },
+ "image": "ghcr.io/mll-lab-nu/efe-private:robodojo-0.2.1 + rcb_robodojo.py of efe:robodojo-files-0.5.0",
+ "agent": "Codex + GPT-6 Astra (high)",
+ "run": "codex-gpt6_astra-high",
+ "mode": "standard",
+ "trial_choice": "a GPT-6 Astra success in standard mode (first choice)",
+ "poster": "last frame (the episode's end state)",
+ "encode": "libx264 preset slow CRF 16, yuv420p, bt709, +faststart",
+ "replaces": "scene-orbit / earlier agent render at masters/robodojo/arrange_largest_number_random",
+ "created": "2026-11-02T06:44:07-06:00"
+}

masters/robodojo-agent/arrange_largest_number_random/video.mp4

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:574432a1aa687e11fcd8ea57255b4ff50da6f3dae6eebf103e2bd637e082b231
+size 16629079

masters/robodojo-agent/build_tower/meta.json

@@ -0,0 +1,64 @@
+{
+ "suite": "robodojo",
+ "task": "build_tower",
+ "task_dir": "robodojo-build-tower-i00-privileged",
+ "task_version": "mll-lab-nu/embodied-first-exam@d95805a999",
+ "source": "agent-success:codex-gpt6_astra-high/robodojo-build-tower-i00-privile__zBYXxnb",
+ "renderer": "Isaac Sim 5.1 RTX PathTracing (rendermode PathTracing): 8 spp per frame, max 6 bounces, OptiX denoiser, firefly filter; native 2560x1440; one dedicated showcase camera (the robots' own cameras are not used)",
+ "rendering": "path-tracing",
+ "width": 2560,
+ "height": 1440,
+ "fps": 30,
+ "frames": 869,
+ "seconds": 28.97,
+ "speed": 9.6,
+ "success_check": true,
+ "success_check_how": "the task's own check list (RoboDojo's reward manager through rcb_robodojo.Session) at the end of this render's replay: every recorded action row from the task's frozen instance, as the grader replays it; rendering never steps physics",
+ "end_check": {
+  "success": true,
+  "end_reason": "success",
+  "actions": 6464,
+  "state_sha": "860990c6cdff13b3",
+  "start_state_sha": "d362e158a85a88b0",
+  "start_matches_instance": true
+ },
+ "replay_hash_check": "privileged trial: graded by replaying the submitted trajectory; this replay reaches the same verdict",
+ "replay": {
+  "trajectory": "artifacts/app/output/trajectory.npz (the submitted trajectory)",
+  "n_actions": 6464,
+  "control_hz": 25,
+  "every_control_steps": 8,
+  "original_verdict": "verifier reward 1.0"
+ },
+ "camera": {
+  "az_deg": 35,
+  "el_deg": 40,
+  "dist_m": 1.2,
+  "focal_mm": 24,
+  "h_aperture_mm": 36.0,
+  "position": [
+   0.535,
+   0.4882,
+   1.5491
+  ],
+  "target": [
+   0.0077,
+   -0.2648,
+   0.7778
+  ]
+ },
+ "holds": {
+  "start_s": 0.5,
+  "end_s": 1.5,
+  "note": "first/last frame held, no extra simulation"
+ },
+ "image": "ghcr.io/mll-lab-nu/efe-private:robodojo-0.2.1 + rcb_robodojo.py of efe:robodojo-files-0.5.0",
+ "agent": "Codex + GPT-6 Astra (high)",
+ "run": "codex-gpt6_astra-high",
+ "mode": "privileged",
+ "trial_choice": "a GPT-6 Astra success in privileged mode (no Astra standard success)",
+ "poster": "last frame (the episode's end state)",
+ "encode": "libx264 preset slow CRF 16, yuv420p, bt709, +faststart",
+ "replaces": "scene-orbit / earlier agent render at masters/robodojo/build_tower",
+ "created": "2026-11-02T07:40:03-06:00"
+}

masters/robodojo-agent/build_tower/video.mp4

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3a6b3ffe22db34e76d5994b5c3d6ac442015ac9d6dc714711a3b5d4756262a26
+size 20997522

masters/robodojo-agent/classify_objects/meta.json

@@ -0,0 +1,64 @@
+{
+ "suite": "robodojo",
+ "task": "classify_objects",
+ "task_dir": "robodojo-classify-objects-i00-standard",
+ "task_version": "mll-lab-nu/embodied-first-exam@d95805a999",
+ "source": "agent-success:codex-gpt6_astra-high/robodojo-classify-objects-i00-st__xeVPaxk",
+ "renderer": "Isaac Sim 5.1 RTX PathTracing (rendermode PathTracing): 8 spp per frame, max 6 bounces, OptiX denoiser, firefly filter; native 2560x1440; one dedicated showcase camera (the robots' own cameras are not used)",
+ "rendering": "path-tracing",
+ "width": 2560,
+ "height": 1440,
+ "fps": 30,
+ "frames": 795,
+ "seconds": 26.5,
+ "speed": 2.4,
+ "success_check": true,
+ "success_check_how": "the task's own check list (RoboDojo's reward manager through rcb_robodojo.Session) at the end of this render's replay: every recorded action row from the task's frozen instance, as the grader replays it; rendering never steps physics",
+ "end_check": {
+  "success": true,
+  "end_reason": "success",
+  "actions": 1467,
+  "state_sha": "46900121ef82ee4d",
+  "start_state_sha": "5b24ec93c8218187",
+  "start_matches_instance": true
+ },
+ "replay_hash_check": "end-state hash 46900121ef82ee4d = the live episode's (final.json live_sha)",
+ "replay": {
+  "trajectory": "verifier/trajectory.npz (recorded by the simulator service during the live episode)",
+  "n_actions": 1467,
+  "control_hz": 25,
+  "every_control_steps": 2,
+  "original_verdict": "verifier reward 1.0"
+ },
+ "camera": {
+  "az_deg": 35,
+  "el_deg": 40,
+  "dist_m": 1.2,
+  "focal_mm": 24,
+  "h_aperture_mm": 36.0,
+  "position": [
+   0.5379,
+   0.5166,
+   1.5579
+  ],
+  "target": [
+   0.0106,
+   -0.2365,
+   0.7866
+  ]
+ },
+ "holds": {
+  "start_s": 0.5,
+  "end_s": 1.5,
+  "note": "first/last frame held, no extra simulation"
+ },
+ "image": "ghcr.io/mll-lab-nu/efe-private:robodojo-0.2.1 + rcb_robodojo.py of efe:robodojo-files-0.5.0",
+ "agent": "Codex + GPT-6 Astra (high)",
+ "run": "codex-gpt6_astra-high",
+ "mode": "standard",
+ "trial_choice": "a GPT-6 Astra success in standard mode (first choice)",
+ "poster": "last frame (the episode's end state)",
+ "encode": "libx264 preset slow CRF 16, yuv420p, bt709, +faststart",
+ "replaces": "scene-orbit / earlier agent render at masters/robodojo/classify_objects",
+ "created": "2026-11-02T05:32:14-06:00"
+}

masters/robodojo-agent/classify_objects/video.mp4

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:24e02619bfcf5ed8eccd458a29ca05ef7a293668abc72bc142a80eab09a9a6ca
+size 19766292

masters/robodojo-agent/classify_objects_by_language/meta.json

@@ -0,0 +1,64 @@
+{
+ "suite": "robodojo",
+ "task": "classify_objects_by_language",
+ "task_dir": "robodojo-classify-objects-by-language-i00-standard",
+ "task_version": "mll-lab-nu/embodied-first-exam@d95805a999",
+ "source": "agent-success:codex-gpt6_astra-high/robodojo-classify-objects-by-lan__fkNrKtE",
+ "renderer": "Isaac Sim 5.1 RTX PathTracing (rendermode PathTracing): 8 spp per frame, max 6 bounces, OptiX denoiser, firefly filter; native 2560x1440; one dedicated showcase camera (the robots' own cameras are not used)",
+ "rendering": "path-tracing",
+ "width": 2560,
+ "height": 1440,
+ "fps": 30,
+ "frames": 757,
+ "seconds": 25.23,
+ "speed": 4.8,
+ "success_check": true,
+ "success_check_how": "the task's own check list (RoboDojo's reward manager through rcb_robodojo.Session) at the end of this render's replay: every recorded action row from the task's frozen instance, as the grader replays it; rendering never steps physics",
+ "end_check": {
+  "success": true,
+  "end_reason": "success",
+  "actions": 2781,
+  "state_sha": "e7d36da73e360292",
+  "start_state_sha": "72e1f4184e093df6",
+  "start_matches_instance": true
+ },
+ "replay_hash_check": "end-state hash e7d36da73e360292 = the live episode's (final.json live_sha)",
+ "replay": {
+  "trajectory": "verifier/trajectory.npz (recorded by the simulator service during the live episode)",
+  "n_actions": 2781,
+  "control_hz": 25,
+  "every_control_steps": 4,
+  "original_verdict": "verifier reward 1.0"
+ },
+ "camera": {
+  "az_deg": 35,
+  "el_deg": 40,
+  "dist_m": 1.2,
+  "focal_mm": 24,
+  "h_aperture_mm": 36.0,
+  "position": [
+   0.514,
+   0.5205,
+   1.5597
+  ],
+  "target": [
+   -0.0132,
+   -0.2326,
+   0.7884
+  ]
+ },
+ "holds": {
+  "start_s": 0.5,
+  "end_s": 1.5,
+  "note": "first/last frame held, no extra simulation"
+ },
+ "image": "ghcr.io/mll-lab-nu/efe-private:robodojo-0.2.1 + rcb_robodojo.py of efe:robodojo-files-0.5.0",
+ "agent": "Codex + GPT-6 Astra (high)",
+ "run": "codex-gpt6_astra-high",
+ "mode": "standard",
+ "trial_choice": "a GPT-6 Astra success in standard mode (first choice)",
+ "poster": "last frame (the episode's end state)",
+ "encode": "libx264 preset slow CRF 16, yuv420p, bt709, +faststart",
+ "replaces": "scene-orbit / earlier agent render at masters/robodojo/classify_objects_by_language",
+ "created": "2026-11-02T06:05:24-06:00"
+}

masters/robodojo-agent/classify_objects_by_language/video.mp4

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:81a9f776693ead0d226b78dffa00982c599843f6ba4363f75e8ee30a355aa961
+size 15360417

masters/robodojo-agent/cover_blocks/meta.json

@@ -0,0 +1,64 @@
+{
+ "suite": "robodojo",
+ "task": "cover_blocks",
+ "task_dir": "robodojo-cover-blocks-i00-standard",
+ "task_version": "mll-lab-nu/embodied-first-exam@d95805a999",
+ "source": "agent-success:codex-gpt6_astra-high/robodojo-cover-blocks-i00-standa__CRdyEwk",
+ "renderer": "Isaac Sim 5.1 RTX PathTracing (rendermode PathTracing): 8 spp per frame, max 6 bounces, OptiX denoiser, firefly filter; native 2560x1440; one dedicated showcase camera (the robots' own cameras are not used)",
+ "rendering": "path-tracing",
+ "width": 2560,
+ "height": 1440,
+ "fps": 30,
+ "frames": 413,
+ "seconds": 13.77,
+ "speed": 2.4,
+ "success_check": true,
+ "success_check_how": "the task's own check list (RoboDojo's reward manager through rcb_robodojo.Session) at the end of this render's replay: every recorded action row from the task's frozen instance, as the grader replays it; rendering never steps physics",
+ "end_check": {
+  "success": true,
+  "end_reason": "success",
+  "actions": 704,
+  "state_sha": "36a3afa3c206a7de",
+  "start_state_sha": "2cb4f02d2568b28e",
+  "start_matches_instance": true
+ },
+ "replay_hash_check": "end-state hash 36a3afa3c206a7de = the live episode's (final.json live_sha)",
+ "replay": {
+  "trajectory": "verifier/trajectory.npz (recorded by the simulator service during the live episode)",
+  "n_actions": 704,
+  "control_hz": 25,
+  "every_control_steps": 2,
+  "original_verdict": "verifier reward 1.0"
+ },
+ "camera": {
+  "az_deg": 35,
+  "el_deg": 40,
+  "dist_m": 1.2,
+  "focal_mm": 24,
+  "h_aperture_mm": 36.0,
+  "position": [
+   0.5273,
+   0.4673,
+   1.5571
+  ],
+  "target": [
+   0.0,
+   -0.2857,
+   0.7858
+  ]
+ },
+ "holds": {
+  "start_s": 0.5,
+  "end_s": 1.5,
+  "note": "first/last frame held, no extra simulation"
+ },
+ "image": "ghcr.io/mll-lab-nu/efe-private:robodojo-0.2.1 + rcb_robodojo.py of efe:robodojo-files-0.5.0",
+ "agent": "Codex + GPT-6 Astra (high)",
+ "run": "codex-gpt6_astra-high",
+ "mode": "standard",
+ "trial_choice": "a GPT-6 Astra success in standard mode (first choice)",
+ "poster": "last frame (the episode's end state)",
+ "encode": "libx264 preset slow CRF 16, yuv420p, bt709, +faststart",
+ "replaces": "scene-orbit / earlier agent render at masters/robodojo/cover_blocks",
+ "created": "2026-11-02T06:23:38-06:00"
+}

masters/robodojo-agent/cover_blocks/video.mp4

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:70c97f40c070306c5751c695b578c64f0e6671a3cdfef7c9ed6f2704f6939a7e
+size 9364013

masters/robodojo-agent/deposit_coin/meta.json

@@ -0,0 +1,64 @@
+{
+ "suite": "robodojo",
+ "task": "deposit_coin",
+ "task_dir": "robodojo-deposit-coin-i00-standard",
+ "task_version": "mll-lab-nu/embodied-first-exam@d95805a999",
+ "source": "agent-success:codex-gpt6_astra-high/robodojo-deposit-coin-i00-standa__iRusWQf",
+ "renderer": "Isaac Sim 5.1 RTX PathTracing (rendermode PathTracing): 8 spp per frame, max 6 bounces, OptiX denoiser, firefly filter; native 2560x1440; one dedicated showcase camera (the robots' own cameras are not used)",
+ "rendering": "path-tracing",
+ "width": 2560,
+ "height": 1440,
+ "fps": 30,
+ "frames": 747,
+ "seconds": 24.9,
+ "speed": 4.8,
+ "success_check": true,
+ "success_check_how": "the task's own check list (RoboDojo's reward manager through rcb_robodojo.Session) at the end of this render's replay: every recorded action row from the task's frozen instance, as the grader replays it; rendering never steps physics",
+ "end_check": {
+  "success": true,
+  "end_reason": "success",
+  "actions": 2743,
+  "state_sha": "4015c1c36a8b9e49",
+  "start_state_sha": "4edc772310e2debd",
+  "start_matches_instance": true
+ },
+ "replay_hash_check": "end-state hash 4015c1c36a8b9e49 = the live episode's (final.json live_sha)",
+ "replay": {
+  "trajectory": "verifier/trajectory.npz (recorded by the simulator service during the live episode)",
+  "n_actions": 2743,
+  "control_hz": 25,
+  "every_control_steps": 4,
+  "original_verdict": "verifier reward 1.0"
+ },
+ "camera": {
+  "az_deg": 35,
+  "el_deg": 40,
+  "dist_m": 1.2,
+  "focal_mm": 24,
+  "h_aperture_mm": 36.0,
+  "position": [
+   0.4897,
+   0.4676,
+   1.5425
+  ],
+  "target": [
+   -0.0376,
+   -0.2854,
+   0.7711
+  ]
+ },
+ "holds": {
+  "start_s": 0.5,
+  "end_s": 1.5,
+  "note": "first/last frame held, no extra simulation"
+ },
+ "image": "ghcr.io/mll-lab-nu/efe-private:robodojo-0.2.1 + rcb_robodojo.py of efe:robodojo-files-0.5.0",
+ "agent": "Codex + GPT-6 Astra (high)",
+ "run": "codex-gpt6_astra-high",
+ "mode": "standard",
+ "trial_choice": "a GPT-6 Astra success in standard mode (first choice)",
+ "poster": "last frame (the episode's end state)",
+ "encode": "libx264 preset slow CRF 16, yuv420p, bt709, +faststart",
+ "replaces": "scene-orbit / earlier agent render at masters/robodojo/deposit_coin",
+ "created": "2026-11-02T06:23:39-06:00"
+}

masters/robodojo-agent/deposit_coin/video.mp4

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:be8d7d6d60794ac13cb766e0340771bafb9541062c281337646a8ee0194d0c8b
+size 18619347

masters/robodojo-agent/fasten_screws/meta.json

@@ -0,0 +1,64 @@
+{
+ "suite": "robodojo",
+ "task": "fasten_screws",
+ "task_dir": "robodojo-fasten-screws-i00-privileged",
+ "task_version": "mll-lab-nu/embodied-first-exam@d95805a999",
+ "source": "agent-success:codex-gpt6_astra-high/robodojo-fasten-screws-i00-privi__85zupS2",
+ "renderer": "Isaac Sim 5.1 RTX PathTracing (rendermode PathTracing): 8 spp per frame, max 6 bounces, OptiX denoiser, firefly filter; native 2560x1440; one dedicated showcase camera (the robots' own cameras are not used)",
+ "rendering": "path-tracing",
+ "width": 2560,
+ "height": 1440,
+ "fps": 30,
+ "frames": 930,
+ "seconds": 31.0,
+ "speed": 13.2,
+ "success_check": true,
+ "success_check_how": "the task's own check list (RoboDojo's reward manager through rcb_robodojo.Session) at the end of this render's replay: every recorded action row from the task's frozen instance, as the grader replays it; rendering never steps physics",
+ "end_check": {
+  "success": true,
+  "end_reason": "success",
+  "actions": 9552,
+  "state_sha": "1f2e9b75a36d2b97",
+  "start_state_sha": "0c2c81d31576b578",
+  "start_matches_instance": true
+ },
+ "replay_hash_check": "privileged trial: graded by replaying the submitted trajectory; this replay reaches the same verdict",
+ "replay": {
+  "trajectory": "artifacts/app/output/trajectory.npz (the submitted trajectory)",
+  "n_actions": 9552,
+  "control_hz": 25,
+  "every_control_steps": 11,
+  "original_verdict": "verifier reward 1.0"
+ },
+ "camera": {
+  "az_deg": 35,
+  "el_deg": 40,
+  "dist_m": 1.2,
+  "focal_mm": 24,
+  "h_aperture_mm": 36.0,
+  "position": [
+   0.5143,
+   0.4692,
+   1.54
+  ],
+  "target": [
+   -0.013,
+   -0.2839,
+   0.7687
+  ]
+ },
+ "holds": {
+  "start_s": 0.5,
+  "end_s": 1.5,
+  "note": "first/last frame held, no extra simulation"
+ },
+ "image": "ghcr.io/mll-lab-nu/efe-private:robodojo-0.2.1 + rcb_robodojo.py of efe:robodojo-files-0.5.0",
+ "agent": "Codex + GPT-6 Astra (high)",
+ "run": "codex-gpt6_astra-high",
+ "mode": "privileged",
+ "trial_choice": "a GPT-6 Astra success in privileged mode (no Astra standard success)",
+ "poster": "last frame (the episode's end state)",
+ "encode": "libx264 preset slow CRF 16, yuv420p, bt709, +faststart",
+ "replaces": "scene-orbit / earlier agent render at masters/robodojo/fasten_screws",
+ "created": "2026-11-02T05:09:00-06:00"
+}

masters/robodojo-agent/fasten_screws/video.mp4

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:88d47af6adb9656ad44344550177a8db6ca1f8cee5be37019320dfb13fcdbef6
+size 15938695

masters/robodojo-agent/fill_egg_holder/meta.json

@@ -0,0 +1,64 @@
+{
+ "suite": "robodojo",
+ "task": "fill_egg_holder",
+ "task_dir": "robodojo-fill-egg-holder-i00-privileged",
+ "task_version": "mll-lab-nu/embodied-first-exam@d95805a999",
+ "source": "agent-success:codex-gpt6_astra-high/robodojo-fill-egg-holder-i00-pri__pFAwzya",
+ "renderer": "Isaac Sim 5.1 RTX PathTracing (rendermode PathTracing): 8 spp per frame, max 6 bounces, OptiX denoiser, firefly filter; native 2560x1440; one dedicated showcase camera (the robots' own cameras are not used)",
+ "rendering": "path-tracing",
+ "width": 2560,
+ "height": 1440,
+ "fps": 30,
+ "frames": 922,
+ "seconds": 30.73,
+ "speed": 6.0,
+ "success_check": true,
+ "success_check_how": "the task's own check list (RoboDojo's reward manager through rcb_robodojo.Session) at the end of this render's replay: every recorded action row from the task's frozen instance, as the grader replays it; rendering never steps physics",
+ "end_check": {
+  "success": true,
+  "end_reason": "success",
+  "actions": 4302,
+  "state_sha": "e6a9bea1b10f0b99",
+  "start_state_sha": "862bab8c9ad00bf6",
+  "start_matches_instance": true
+ },
+ "replay_hash_check": "privileged trial: graded by replaying the submitted trajectory; this replay reaches the same verdict",
+ "replay": {
+  "trajectory": "artifacts/app/output/trajectory.npz (the submitted trajectory)",
+  "n_actions": 4302,
+  "control_hz": 25,
+  "every_control_steps": 5,
+  "original_verdict": "verifier reward 1.0"
+ },
+ "camera": {
+  "az_deg": 35,
+  "el_deg": 40,
+  "dist_m": 1.2,
+  "focal_mm": 24,
+  "h_aperture_mm": 36.0,
+  "position": [
+   0.626,
+   0.4732,
+   1.566
+  ],
+  "target": [
+   0.0987,
+   -0.2798,
+   0.7947
+  ]
+ },
+ "holds": {
+  "start_s": 0.5,
+  "end_s": 1.5,
+  "note": "first/last frame held, no extra simulation"
+ },
+ "image": "ghcr.io/mll-lab-nu/efe-private:robodojo-0.2.1 + rcb_robodojo.py of efe:robodojo-files-0.5.0",
+ "agent": "Codex + GPT-6 Astra (high)",
+ "run": "codex-gpt6_astra-high",
+ "mode": "privileged",
+ "trial_choice": "a GPT-6 Astra success in privileged mode (no Astra standard success)",
+ "poster": "last frame (the episode's end state)",
+ "encode": "libx264 preset slow CRF 16, yuv420p, bt709, +faststart",
+ "replaces": "scene-orbit / earlier agent render at masters/robodojo/fill_egg_holder",
+ "created": "2026-11-02T05:09:01-06:00"
+}

masters/robodojo-agent/fill_egg_holder/video.mp4

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c76ec0d5259e008e79522cece4d2b2d4c926ab54bab8727330a6e1c2d31f6c54
+size 22274054

masters/robodojo-agent/fill_pen_holder/meta.json

@@ -0,0 +1,64 @@
+{
+ "suite": "robodojo",
+ "task": "fill_pen_holder",
+ "task_dir": "robodojo-fill-pen-holder-i00-standard",
+ "task_version": "mll-lab-nu/embodied-first-exam@d95805a999",
+ "source": "agent-success:codex-gpt6_astra-high/robodojo-fill-pen-holder-i00-sta__2KQS5Ne",
+ "renderer": "Isaac Sim 5.1 RTX PathTracing (rendermode PathTracing): 8 spp per frame, max 6 bounces, OptiX denoiser, firefly filter; native 2560x1440; one dedicated showcase camera (the robots' own cameras are not used)",
+ "rendering": "path-tracing",
+ "width": 2560,
+ "height": 1440,
+ "fps": 30,
+ "frames": 924,
+ "seconds": 30.8,
+ "speed": 3.6,
+ "success_check": true,
+ "success_check_how": "the task's own check list (RoboDojo's reward manager through rcb_robodojo.Session) at the end of this render's replay: every recorded action row from the task's frozen instance, as the grader replays it; rendering never steps physics",
+ "end_check": {
+  "success": true,
+  "end_reason": "success",
+  "actions": 2588,
+  "state_sha": "4ebc94efa77dc21f",
+  "start_state_sha": "5a23081b3ebbd05b",
+  "start_matches_instance": true
+ },
+ "replay_hash_check": "end-state hash 4ebc94efa77dc21f = the live episode's (final.json live_sha)",
+ "replay": {
+  "trajectory": "verifier/trajectory.npz (recorded by the simulator service during the live episode)",
+  "n_actions": 2588,
+  "control_hz": 25,
+  "every_control_steps": 3,
+  "original_verdict": "verifier reward 1.0"
+ },
+ "camera": {
+  "az_deg": 35,
+  "el_deg": 40,
+  "dist_m": 1.2,
+  "focal_mm": 24,
+  "h_aperture_mm": 36.0,
+  "position": [
+   0.4642,
+   0.518,
+   1.5416
+  ],
+  "target": [
+   -0.0631,
+   -0.235,
+   0.7702
+  ]
+ },
+ "holds": {
+  "start_s": 0.5,
+  "end_s": 1.5,
+  "note": "first/last frame held, no extra simulation"
+ },
+ "image": "ghcr.io/mll-lab-nu/efe-private:robodojo-0.2.1 + rcb_robodojo.py of efe:robodojo-files-0.5.0",
+ "agent": "Codex + GPT-6 Astra (high)",
+ "run": "codex-gpt6_astra-high",
+ "mode": "standard",
+ "trial_choice": "a GPT-6 Astra success in standard mode (first choice)",
+ "poster": "last frame (the episode's end state)",
+ "encode": "libx264 preset slow CRF 16, yuv420p, bt709, +faststart",
+ "replaces": "scene-orbit / earlier agent render at masters/robodojo/fill_pen_holder",
+ "created": "2026-11-02T05:09:02-06:00"
+}

masters/robodojo-agent/fill_pen_holder/video.mp4

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d419c6653615aaf5f381966c38bbb9828a5b9b0fb5d9b328fd54dcaf5cae20ff
+size 22882610

masters/robodojo-agent/fold_clothes/meta.json

@@ -0,0 +1,65 @@
+{
+ "suite": "robodojo",
+ "task": "fold_clothes",
+ "task_dir": "robodojo-fold-clothes-i00-standard",
+ "task_version": "mll-lab-nu/embodied-first-exam@d95805a999",
+ "source": "agent-success:codex-gpt6_astra-high/robodojo-fold-clothes-i00-standa__8nriFt4",
+ "renderer": "Isaac Sim 5.1 RTX PathTracing (rendermode PathTracing): 8 spp per frame, max 6 bounces, OptiX denoiser, firefly filter; native 2560x1440; one dedicated showcase camera (the robots' own cameras are not used)",
+ "rendering": "path-tracing",
+ "width": 2560,
+ "height": 1440,
+ "fps": 30,
+ "frames": 365,
+ "seconds": 12.17,
+ "speed": 2.4,
+ "success_check": true,
+ "success_check_how": "the task's own check list (RoboDojo's reward manager through rcb_robodojo.Session) at the end of this render's replay: every recorded action row from the task's frozen instance, as the grader replays it; rendering never steps physics",
+ "end_check": {
+  "success": true,
+  "end_reason": "success",
+  "actions": 607,
+  "state_sha": "cb428325d829e338",
+  "start_state_sha": "5fe7775799792961",
+  "start_matches_instance": true
+ },
+ "replay_hash_check": "end-state hash cb428325d829e338 vs live 72389cf0ca193ddb: this task is graded by verdict (its particle physics does not replay bit for bit); the verdict matches",
+ "replay": {
+  "trajectory": "verifier/trajectory.npz (recorded by the simulator service during the live episode)",
+  "n_actions": 607,
+  "control_hz": 25,
+  "every_control_steps": 2,
+  "original_verdict": "verifier reward 1.0"
+ },
+ "camera": {
+  "az_deg": 35,
+  "el_deg": 40,
+  "dist_m": 1.2,
+  "focal_mm": 24,
+  "h_aperture_mm": 36.0,
+  "position": [
+   0.5338,
+   0.4254,
+   1.546
+  ],
+  "target": [
+   0.0066,
+   -0.3276,
+   0.7747
+  ]
+ },
+ "holds": {
+  "start_s": 0.5,
+  "end_s": 1.5,
+  "note": "first/last frame held, no extra simulation"
+ },
+ "image": "ghcr.io/mll-lab-nu/efe-private:robodojo-0.2.1 + rcb_robodojo.py of efe:robodojo-files-0.5.0",
+ "agent": "Codex + GPT-6 Astra (high)",
+ "run": "codex-gpt6_astra-high",
+ "mode": "standard",
+ "trial_choice": "a GPT-6 Astra success in standard mode (first choice)",
+ "poster": "last frame (the episode's end state)",
+ "encode": "libx264 preset slow CRF 16, yuv420p, bt709, +faststart",
+ "replaces": "scene-orbit / earlier agent render at masters/robodojo/fold_clothes",
+ "created": "2026-11-02T07:09:59-06:00",
+ "replay_note": "cloth physics does not replay bit for bit (the task is graded by verdict): a first replay of this trial ended without success; this second replay reaches the live verdict (success) and is the one shown"
+}

masters/robodojo-agent/fold_clothes/video.mp4

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c5a91a52a656dee8435b63214c1b3404377457dfd4d35fb377ff67f7219bff2e
+size 11865770

masters/robodojo-agent/fold_clothes_random/meta.json

@@ -0,0 +1,64 @@
+{
+ "suite": "robodojo",
+ "task": "fold_clothes_random",
+ "task_dir": "robodojo-fold-clothes-random-i00-standard",
+ "task_version": "mll-lab-nu/embodied-first-exam@d95805a999",
+ "source": "agent-success:codex-gpt6_astra-high/robodojo-fold-clothes-random-i00__iiXf67q",
+ "renderer": "Isaac Sim 5.1 RTX PathTracing (rendermode PathTracing): 8 spp per frame, max 6 bounces, OptiX denoiser, firefly filter; native 2560x1440; one dedicated showcase camera (the robots' own cameras are not used)",
+ "rendering": "path-tracing",
+ "width": 2560,
+ "height": 1440,
+ "fps": 30,
+ "frames": 363,
+ "seconds": 12.1,
+ "speed": 2.4,
+ "success_check": true,
+ "success_check_how": "the task's own check list (RoboDojo's reward manager through rcb_robodojo.Session) at the end of this render's replay: every recorded action row from the task's frozen instance, as the grader replays it; rendering never steps physics",
+ "end_check": {
+  "success": true,
+  "end_reason": "success",
+  "actions": 603,
+  "state_sha": "aecac4ca314072fa",
+  "start_state_sha": "f204bd8800c9fc2a",
+  "start_matches_instance": true
+ },
+ "replay_hash_check": "end-state hash aecac4ca314072fa vs live 0062b4cbd8cf3ed4: this task is graded by verdict (its particle physics does not replay bit for bit); the verdict matches",
+ "replay": {
+  "trajectory": "verifier/trajectory.npz (recorded by the simulator service during the live episode)",
+  "n_actions": 603,
+  "control_hz": 25,
+  "every_control_steps": 2,
+  "original_verdict": "verifier reward 1.0"
+ },
+ "camera": {
+  "az_deg": 35,
+  "el_deg": 40,
+  "dist_m": 1.2,
+  "focal_mm": 24,
+  "h_aperture_mm": 36.0,
+  "position": [
+   0.5273,
+   0.423,
+   1.5471
+  ],
+  "target": [
+   0.0001,
+   -0.33,
+   0.7757
+  ]
+ },
+ "holds": {
+  "start_s": 0.5,
+  "end_s": 1.5,
+  "note": "first/last frame held, no extra simulation"
+ },
+ "image": "ghcr.io/mll-lab-nu/efe-private:robodojo-0.2.1 + rcb_robodojo.py of efe:robodojo-files-0.5.0",
+ "agent": "Codex + GPT-6 Astra (high)",
+ "run": "codex-gpt6_astra-high",
+ "mode": "standard",
+ "trial_choice": "a GPT-6 Astra success in standard mode (first choice)",
+ "poster": "last frame (the episode's end state)",
+ "encode": "libx264 preset slow CRF 16, yuv420p, bt709, +faststart",
+ "replaces": "scene-orbit / earlier agent render at masters/robodojo/fold_clothes_random",
+ "created": "2026-11-02T07:40:10-06:00"
+}

masters/robodojo-agent/fold_clothes_random/video.mp4

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7640fd88e6da0f271a3571965a67f54760ee49f249ddd44df906327dcd31a4e8
+size 11010865

masters/robodojo-agent/general_pickup/meta.json

@@ -0,0 +1,64 @@
+{
+ "suite": "robodojo",
+ "task": "general_pickup",
+ "task_dir": "robodojo-general-pickup-i00-privileged",
+ "task_version": "mll-lab-nu/embodied-first-exam@d95805a999",
+ "source": "agent-success:codex-gpt6_luna-xhigh/robodojo-general-pickup-i00-priv__a2wHLF7",
+ "renderer": "Isaac Sim 5.1 RTX PathTracing (rendermode PathTracing): 8 spp per frame, max 6 bounces, OptiX denoiser, firefly filter; native 2560x1440; one dedicated showcase camera (the robots' own cameras are not used)",
+ "rendering": "path-tracing",
+ "width": 2560,
+ "height": 1440,
+ "fps": 30,
+ "frames": 231,
+ "seconds": 7.7,
+ "speed": 1.2,
+ "success_check": true,
+ "success_check_how": "the task's own check list (RoboDojo's reward manager through rcb_robodojo.Session) at the end of this render's replay: every recorded action row from the task's frozen instance, as the grader replays it; rendering never steps physics",
+ "end_check": {
+  "success": true,
+  "end_reason": "success",
+  "actions": 170,
+  "state_sha": "ca9c7908d186d6b1",
+  "start_state_sha": "3d8f511ee7e4b021",
+  "start_matches_instance": true
+ },
+ "replay_hash_check": "privileged trial: graded by replaying the submitted trajectory; this replay reaches the same verdict",
+ "replay": {
+  "trajectory": "artifacts/app/output/trajectory.npz (the submitted trajectory)",
+  "n_actions": 170,
+  "control_hz": 25,
+  "every_control_steps": 1,
+  "original_verdict": "verifier reward 1.0"
+ },
+ "camera": {
+  "az_deg": 35,
+  "el_deg": 40,
+  "dist_m": 1.2,
+  "focal_mm": 24,
+  "h_aperture_mm": 36.0,
+  "position": [
+   0.6264,
+   0.4513,
+   1.5392
+  ],
+  "target": [
+   0.0992,
+   -0.3017,
+   0.7679
+  ]
+ },
+ "holds": {
+  "start_s": 0.5,
+  "end_s": 1.5,
+  "note": "first/last frame held, no extra simulation"
+ },
+ "image": "ghcr.io/mll-lab-nu/efe-private:robodojo-0.2.1 + rcb_robodojo.py of efe:robodojo-files-0.5.0",
+ "agent": "Codex + GPT-6 Luna (xhigh)",
+ "run": "codex-gpt6_luna-xhigh",
+ "mode": "privileged",
+ "trial_choice": "GPT-6 Luna's success: the only model that solved this task",
+ "poster": "last frame (the episode's end state)",
+ "encode": "libx264 preset slow CRF 16, yuv420p, bt709, +faststart",
+ "replaces": "scene-orbit / earlier agent render at masters/robodojo/general_pickup",
+ "created": "2026-11-02T07:40:11-06:00"
+}

masters/robodojo-agent/general_pickup/video.mp4

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1cbc3d5eb7905da15049c41e5c55201319904cc5d75386f8dc6acaf161c0a163
+size 4289338

masters/robodojo-agent/hang_mugs/meta.json

@@ -0,0 +1,64 @@
+{
+ "suite": "robodojo",
+ "task": "hang_mugs",
+ "task_dir": "robodojo-hang-mugs-i00-privileged",
+ "task_version": "mll-lab-nu/embodied-first-exam@d95805a999",
+ "source": "agent-success:codex-gpt6_astra-high/robodojo-hang-mugs-i00-privilege__PabFRXN",
+ "renderer": "Isaac Sim 5.1 RTX PathTracing (rendermode PathTracing): 8 spp per frame, max 6 bounces, OptiX denoiser, firefly filter; native 2560x1440; one dedicated showcase camera (the robots' own cameras are not used)",
+ "rendering": "path-tracing",
+ "width": 2560,
+ "height": 1440,
+ "fps": 30,
+ "frames": 802,
+ "seconds": 26.73,
+ "speed": 3.6,
+ "success_check": true,
+ "success_check_how": "the task's own check list (RoboDojo's reward manager through rcb_robodojo.Session) at the end of this render's replay: every recorded action row from the task's frozen instance, as the grader replays it; rendering never steps physics",
+ "end_check": {
+  "success": true,
+  "end_reason": "success",
+  "actions": 2223,
+  "state_sha": "bde6970495ed34ea",
+  "start_state_sha": "caba229120e87e95",
+  "start_matches_instance": true
+ },
+ "replay_hash_check": "privileged trial: graded by replaying the submitted trajectory; this replay reaches the same verdict",
+ "replay": {
+  "trajectory": "artifacts/app/output/trajectory.npz (the submitted trajectory)",
+  "n_actions": 2270,
+  "control_hz": 25,
+  "every_control_steps": 3,
+  "original_verdict": "verifier reward 1.0"
+ },
+ "camera": {
+  "az_deg": 35,
+  "el_deg": 40,
+  "dist_m": 1.2,
+  "focal_mm": 24,
+  "h_aperture_mm": 36.0,
+  "position": [
+   0.5198,
+   0.4989,
+   1.5918
+  ],
+  "target": [
+   -0.0075,
+   -0.2541,
+   0.8205
+  ]
+ },
+ "holds": {
+  "start_s": 0.5,
+  "end_s": 1.5,
+  "note": "first/last frame held, no extra simulation"
+ },
+ "image": "ghcr.io/mll-lab-nu/efe-private:robodojo-0.2.1 + rcb_robodojo.py of efe:robodojo-files-0.5.0",
+ "agent": "Codex + GPT-6 Astra (high)",
+ "run": "codex-gpt6_astra-high",
+ "mode": "privileged",
+ "trial_choice": "a GPT-6 Astra success in privileged mode (no Astra standard success)",
+ "poster": "last frame (the episode's end state)",
+ "encode": "libx264 preset slow CRF 16, yuv420p, bt709, +faststart",
+ "replaces": "scene-orbit / earlier agent render at masters/robodojo/hang_mugs",
+ "created": "2026-11-02T05:09:03-06:00"
+}

masters/robodojo-agent/hang_mugs/video.mp4

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:56390886b17dc0594224afdb8e37a94648dd69f26822195775234b803c472572
+size 16162900

masters/robodojo-agent/hang_mugs_random/meta.json

@@ -0,0 +1,64 @@
+{
+ "suite": "robodojo",
+ "task": "hang_mugs_random",
+ "task_dir": "robodojo-hang-mugs-random-i00-standard",
+ "task_version": "mll-lab-nu/embodied-first-exam@d95805a999",
+ "source": "agent-success:antigravity_cli-gemini_4_argon-high/robodojo-hang-mugs-random-i00-st__JaVoykK",
+ "renderer": "Isaac Sim 5.1 RTX PathTracing (rendermode PathTracing): 8 spp per frame, max 6 bounces, OptiX denoiser, firefly filter; native 2560x1440; one dedicated showcase camera (the robots' own cameras are not used)",
+ "rendering": "path-tracing",
+ "width": 2560,
+ "height": 1440,
+ "fps": 30,
+ "frames": 758,
+ "seconds": 25.27,
+ "speed": 2.4,
+ "success_check": true,
+ "success_check_how": "the task's own check list (RoboDojo's reward manager through rcb_robodojo.Session) at the end of this render's replay: every recorded action row from the task's frozen instance, as the grader replays it; rendering never steps physics",
+ "end_check": {
+  "success": true,
+  "end_reason": "success",
+  "actions": 1394,
+  "state_sha": "92dc37e85ecfa139",
+  "start_state_sha": "67eafbb6c4f5692b",
+  "start_matches_instance": true
+ },
+ "replay_hash_check": "end-state hash 92dc37e85ecfa139 = the live episode's (final.json live_sha)",
+ "replay": {
+  "trajectory": "verifier/trajectory.npz (recorded by the simulator service during the live episode)",
+  "n_actions": 1394,
+  "control_hz": 25,
+  "every_control_steps": 2,
+  "original_verdict": "verifier reward 1.0"
+ },
+ "camera": {
+  "az_deg": 35,
+  "el_deg": 40,
+  "dist_m": 1.2,
+  "focal_mm": 24,
+  "h_aperture_mm": 36.0,
+  "position": [
+   0.5232,
+   0.4914,
+   1.5883
+  ],
+  "target": [
+   -0.004,
+   -0.2616,
+   0.817
+  ]
+ },
+ "holds": {
+  "start_s": 0.5,
+  "end_s": 1.5,
+  "note": "first/last frame held, no extra simulation"
+ },
+ "image": "ghcr.io/mll-lab-nu/efe-private:robodojo-0.2.1 + rcb_robodojo.py of efe:robodojo-files-0.5.0",
+ "agent": "Antigravity + Gemini 4 Argon (high)",
+ "run": "antigravity_cli-gemini_4_argon-high",
+ "mode": "standard",
+ "trial_choice": "a success by another model (no GPT-6 Astra success)",
+ "poster": "last frame (the episode's end state)",
+ "encode": "libx264 preset slow CRF 16, yuv420p, bt709, +faststart",
+ "replaces": "scene-orbit / earlier agent render at masters/robodojo/hang_mugs_random",
+ "created": "2026-11-02T06:23:41-06:00"
+}

masters/robodojo-agent/hang_mugs_random/video.mp4

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f6a0d2c3a49badc90f24c1ec86ecaf0c7b02f9e8126d0530d4c8b965ae6f41e7
+size 14219618

masters/robodojo-agent/highlights.json

@@ -0,0 +1 @@
+{"teaser": "hang_mugs_random", "highlights": ["hang_mugs_random", "stack_blocks_by_language", "arrange_largest_number", "match_and_pick_from_conveyor", "pour_by_language", "make_toast_random"], "why_fewer": ""}

masters/robodojo-agent/imitate_sorting_sequence/meta.json

@@ -0,0 +1,64 @@
+{
+ "suite": "robodojo",
+ "task": "imitate_sorting_sequence",
+ "task_dir": "robodojo-imitate-sorting-sequence-i00-standard",
+ "task_version": "mll-lab-nu/embodied-first-exam@d95805a999",
+ "source": "agent-success:codex-gpt6_astra-high/robodojo-imitate-sorting-sequenc__4UmiC2c",
+ "renderer": "Isaac Sim 5.1 RTX PathTracing (rendermode PathTracing): 8 spp per frame, max 6 bounces, OptiX denoiser, firefly filter; native 2560x1440; one dedicated showcase camera (the robots' own cameras are not used)",
+ "rendering": "path-tracing",
+ "width": 2560,
+ "height": 1440,
+ "fps": 30,
+ "frames": 930,
+ "seconds": 31.0,
+ "speed": 2.4,
+ "success_check": true,
+ "success_check_how": "the task's own check list (RoboDojo's reward manager through rcb_robodojo.Session) at the end of this render's replay: every recorded action row from the task's frozen instance, as the grader replays it; rendering never steps physics",
+ "end_check": {
+  "success": true,
+  "end_reason": "success",
+  "actions": 1738,
+  "state_sha": "d2ab0510e21b9a6a",
+  "start_state_sha": "e28b78099f6e102d",
+  "start_matches_instance": true
+ },
+ "replay_hash_check": "end-state hash d2ab0510e21b9a6a = the live episode's (final.json live_sha)",
+ "replay": {
+  "trajectory": "verifier/trajectory.npz (recorded by the simulator service during the live episode)",
+  "n_actions": 1738,
+  "control_hz": 25,
+  "every_control_steps": 2,
+  "original_verdict": "verifier reward 1.0"
+ },
+ "camera": {
+  "az_deg": 80,
+  "el_deg": 45,
+  "dist_m": 1.4,
+  "focal_mm": 24,
+  "h_aperture_mm": 36.0,
+  "position": [
+   0.9658,
+   -0.0051,
+   1.7727
+  ],
+  "target": [
+   -0.0091,
+   -0.177,
+   0.7827
+  ]
+ },
+ "holds": {
+  "start_s": 0.5,
+  "end_s": 1.5,
+  "note": "first/last frame held, no extra simulation"
+ },
+ "image": "ghcr.io/mll-lab-nu/efe-private:robodojo-0.2.1 + rcb_robodojo.py of efe:robodojo-files-0.5.0",
+ "agent": "Codex + GPT-6 Astra (high)",
+ "run": "codex-gpt6_astra-high",
+ "mode": "standard",
+ "trial_choice": "a GPT-6 Astra success in standard mode (first choice)",
+ "poster": "last frame (the episode's end state)",
+ "encode": "libx264 preset slow CRF 16, yuv420p, bt709, +faststart",
+ "replaces": "scene-orbit / earlier agent render at masters/robodojo/imitate_sorting_sequence",
+ "created": "2026-11-02T07:40:05-06:00"
+}