masters: 33 tasks rendered 2026-10-19 (RoboOlympics-Fun, RoboEngineer AnyCar, YAMLab)

2560x1440 path-traced masters, checked with check_media.py; folder = the site task id.

masters/roboengineer/anycar_drift_park/meta.json

@@ -0,0 +1,41 @@
+{
+ "suite": "roboengineer",
+ "task": "anycar_drift_park",
+ "source": "scene-orbit",
+ "renderer": "Blender 4.5.3 LTS Cycles OptiX, 64+ spp, adaptive 0.01, OIDN albedo+normal",
+ "rendering": "path-tracing",
+ "width": 2560,
+ "height": 1440,
+ "fps": 30,
+ "speed": 1.0,
+ "success_check": false,
+ "image": "ghcr.io/mll-lab-nu/efe-private:control-transfer-0.2.0",
+ "text_overlays": false,
+ "physics_in_blender": false,
+ "source_record_sha256": null,
+ "timing": {
+  "physical_time_range_s": [
+   0.0,
+   0.0
+  ],
+  "video_pts_range_s": [
+   0.0,
+   6.966666666666667
+  ],
+  "final_state_grid_delay_s": 0.0,
+  "final_frame_hold_s": 0.03333333333333333,
+  "policy": "Frozen initial state for all 210 frames; camera-only 7-second orbit; no physical time progression",
+  "frozen_scene_display_duration_s": 7.0
+ },
+ "why_no_video": "No successful exam trial in the supplied 2026-10-19 inventory; current task metadata declares oracle=none and frozen reference has zero successful held-out cases. Showing this exact task held-out case 0 frozen initial scene only.",
+ "presentation_note": "overhead luminaire hidden from the camera; lighting unchanged; owner-approved 2026-10-19",
+ "asset_attribution": [
+  {
+   "creator": "LeCAR-Lab AnyCar contributors",
+   "source": "https://github.com/LeCAR-Lab/anycar/tree/20fc50a60d3dfff5e6e002a9cdbbb02d604d1b1f",
+   "license": "Apache-2.0 (upstream README declaration)",
+   "license_url": "https://www.apache.org/licenses/LICENSE-2.0",
+   "changes": "Wheel, suspension and steering mechanism adapted to procedural bodywork and benchmark terrain; no upstream mesh or checkpoint included."
+  }
+ ]
+}

masters/roboengineer/anycar_drift_park/video.mp4

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c00092b8a079bc9a13854349818ab357739a5f149f2f4cc20ac8f973ec1399c1
+size 3395056

masters/roboengineer/anycar_drift_park_i01/meta.json

@@ -0,0 +1,41 @@
+{
+ "suite": "roboengineer",
+ "task": "anycar_drift_park_i01",
+ "source": "scene-orbit",
+ "renderer": "Blender 4.5.3 LTS Cycles OptiX, 64+ spp, adaptive 0.01, OIDN albedo+normal",
+ "rendering": "path-tracing",
+ "width": 2560,
+ "height": 1440,
+ "fps": 30,
+ "speed": 1.0,
+ "success_check": false,
+ "image": "ghcr.io/mll-lab-nu/efe-private:control-transfer-0.2.0",
+ "text_overlays": false,
+ "physics_in_blender": false,
+ "source_record_sha256": null,
+ "timing": {
+  "physical_time_range_s": [
+   0.0,
+   0.0
+  ],
+  "video_pts_range_s": [
+   0.0,
+   6.966666666666667
+  ],
+  "final_state_grid_delay_s": 0.0,
+  "final_frame_hold_s": 0.03333333333333333,
+  "policy": "Frozen initial state for all 210 frames; camera-only 7-second orbit; no physical time progression",
+  "frozen_scene_display_duration_s": 7.0
+ },
+ "why_no_video": "No successful exam trial in the supplied 2026-10-19 inventory; current task metadata declares oracle=none and frozen reference has zero successful held-out cases. Showing this exact task held-out case 0 frozen initial scene only.",
+ "presentation_note": "overhead luminaire hidden from the camera; lighting unchanged; owner-approved 2026-10-19",
+ "asset_attribution": [
+  {
+   "creator": "LeCAR-Lab AnyCar contributors",
+   "source": "https://github.com/LeCAR-Lab/anycar/tree/20fc50a60d3dfff5e6e002a9cdbbb02d604d1b1f",
+   "license": "Apache-2.0 (upstream README declaration)",
+   "license_url": "https://www.apache.org/licenses/LICENSE-2.0",
+   "changes": "Wheel, suspension and steering mechanism adapted to procedural bodywork and benchmark terrain; no upstream mesh or checkpoint included."
+  }
+ ]
+}

masters/roboengineer/anycar_drift_park_i01/video.mp4

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:491a0819385a9d37bfc8bffbc830a99fcfea885bc68153dccbce18dacd2554cf
+size 3399862

masters/roboengineer/anycar_tight_drift_park/meta.json

@@ -0,0 +1,41 @@
+{
+ "suite": "roboengineer",
+ "task": "anycar_tight_drift_park",
+ "source": "scene-orbit",
+ "renderer": "Blender 4.5.3 LTS Cycles OptiX, 64+ spp, adaptive 0.01, OIDN albedo+normal",
+ "rendering": "path-tracing",
+ "width": 2560,
+ "height": 1440,
+ "fps": 30,
+ "speed": 1.0,
+ "success_check": false,
+ "image": "ghcr.io/mll-lab-nu/efe-private:control-transfer-0.2.0",
+ "text_overlays": false,
+ "physics_in_blender": false,
+ "source_record_sha256": null,
+ "timing": {
+  "physical_time_range_s": [
+   0.0,
+   0.0
+  ],
+  "video_pts_range_s": [
+   0.0,
+   6.966666666666667
+  ],
+  "final_state_grid_delay_s": 0.0,
+  "final_frame_hold_s": 0.03333333333333333,
+  "policy": "Frozen initial state for all 210 frames; camera-only 7-second orbit; no physical time progression",
+  "frozen_scene_display_duration_s": 7.0
+ },
+ "why_no_video": "No successful exam trial in the supplied 2026-10-19 inventory; current task metadata declares oracle=none and frozen reference has zero successful held-out cases. Showing this exact task held-out case 0 frozen initial scene only.",
+ "presentation_note": "overhead luminaire hidden from the camera; lighting unchanged; owner-approved 2026-10-19",
+ "asset_attribution": [
+  {
+   "creator": "LeCAR-Lab AnyCar contributors",
+   "source": "https://github.com/LeCAR-Lab/anycar/tree/20fc50a60d3dfff5e6e002a9cdbbb02d604d1b1f",
+   "license": "Apache-2.0 (upstream README declaration)",
+   "license_url": "https://www.apache.org/licenses/LICENSE-2.0",
+   "changes": "Wheel, suspension and steering mechanism adapted to procedural bodywork and benchmark terrain; no upstream mesh or checkpoint included."
+  }
+ ]
+}

masters/roboengineer/anycar_tight_drift_park/video.mp4

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ee292ca364b3ec8fb9105e8ac3e61ff2d2f8ae49dcbe2cdb477d09ed1e93b52d
+size 3473684

masters/roboolympics-fun/badminton_serve/meta.json

@@ -0,0 +1,103 @@
+{
+ "suite": "roboolympics-fun",
+ "task": "badminton_serve",
+ "source": "agent-success:codex-gpt6_astra-high-roboolympics-fun-badminton-serve-i00-privileged/roboolympics-badminton-serve-i00__D5pDUDF",
+ "renderer": "Blender 4.2.23 LTS Cycles path tracing (GPU OptiX), adaptive sampling (threshold 0.01, 64-128 spp), OpenImageDenoise with albedo + normal passes; kinematic playback of the exact MuJoCo replay poses (MuJoCo's own compiled meshes, primitives and colours); daylight: a sun lamp (2 degree disc) and a Nishita sky (no sun disc), the venue's own ground colour on a large ground plane with a faint world-space variation; AgX view transform, Medium High Contrast look; a tracking camera (fixed 22 degree elevation, per-frame fit of the robot and the object in play, smoothed in time) with a slow eased 10-degree orbit",
+ "rendering": "path-tracing",
+ "width": 2560,
+ "height": 1440,
+ "fps": 30,
+ "frames": 279,
+ "seconds": 9.3,
+ "speed": 0.3,
+ "success_check": true,
+ "success_check_how": "the task's own verdict (rcb_rfo.py Episode.verdict(), the referee, from the task dir's tests/ (privileged verifier copy)) at the end of this re-render's own replay of every recorded row from the task's frozen instance (t = 0 state hash checked); a second plain replay with the task's own replay() agrees and ends in the same final state hash (9afedd98535665a1)",
+ "end_check": {
+  "success": true,
+  "stages": [
+   "release",
+   "hit",
+   "over the net",
+   "in"
+  ],
+  "stages_total": 4,
+  "progress": 1.0,
+  "metrics": {
+   "hit_height_m": 0.4881,
+   "shuttle_speed_m_s": 11.9426,
+   "shuttle_angle_deg": 45.48,
+   "landing_x_m": 5.788,
+   "landing_y_m": 1.716
+  },
+  "time_s": 2.18
+ },
+ "recorded_result": {
+  "verifier_reward_json": {
+   "success": 1,
+   "n_actions": 218,
+   "progress": 1.0,
+   "stages_reached": 4,
+   "stages_total": 4,
+   "metric_hit_height_m": 0.4881,
+   "metric_landing_x_m": 5.788,
+   "metric_landing_y_m": 1.716,
+   "metric_shuttle_angle_deg": 45.48,
+   "metric_shuttle_speed_m_s": 11.9426
+  }
+ },
+ "replay": {
+  "task_dir": "roboolympics-fun-badminton-serve-i00-privileged",
+  "trajectory": "jobs/codex/gpt6_astra/high/codex-gpt6_astra-high-roboolympics-fun-badminton-serve-i00-privileged/roboolympics-badminton-serve-i00__D5pDUDF/artifacts/app/output/trajectory.npz",
+  "n_actions": 218,
+  "control_steps_replayed": 218,
+  "control_hz": 100.0,
+  "every_control_steps": 1,
+  "final_state_sha": "9afedd98535665a1dee70b866bc1073569264d702c09337714ba6088628908f6",
+  "replays_agree": true,
+  "mujoco": "3.3.7",
+  "original_verdict": "counted success (verifier reward 1)"
+ },
+ "visual_changes": [
+  "daylight sky and a plain ground plane in the venue's ground colour instead of the venue's painted sky, hills, trees, grandstands and perimeter boards (drawn-only scenery)"
+ ],
+ "camera": {
+  "mode": "director",
+  "az_deg": -99.04,
+  "az_rule": "object-travel-115",
+  "el_deg": 22.0,
+  "fovy_deg": 34.0,
+  "fill": 0.72,
+  "distance_m": [
+   4.504,
+   7.156
+  ],
+  "robot_root": "pelvis",
+  "travel_m": 0.006,
+  "focus": {
+   "robot+object": 212,
+   "robot": 7
+  },
+  "objects": {
+   "shuttle": 5.62,
+   "racket": 0.31
+  }
+ },
+ "orbit_deg": 10.0,
+ "holds": {
+  "start_s": 0.5,
+  "end_s": 1.5,
+  "note": "first/last frame held, no extra simulation"
+ },
+ "seconds_per_frame": 3.43,
+ "image": "ghcr.io/mll-lab-nu/efe:robofunolympics-0.2.1 (simulator) + ghcr.io/mll-lab-nu/efe:roboolympics-fun-files-0.1.0@sha256:49a33342a72b57b56742cc754e427205a30f39dfcd69a99d1d7275c0eae4a619 (rcb_rfo.py, rfo/)",
+ "task_dir": "roboolympics-fun-badminton-serve-i00-privileged",
+ "agent": "codex + GPT-6 Astra (high)",
+ "run": "codex-gpt6_astra-high",
+ "mode": "privileged",
+ "trial": "codex-gpt6_astra-high-roboolympics-fun-badminton-serve-i00-privileged/roboolympics-badminton-serve-i00__D5pDUDF",
+ "trial_choice": "a counted GPT-6 Astra success in privileged mode",
+ "poster": "last frame (solved state)",
+ "host": "vast.ai RTX 4090 gpu0",
+ "encode": "libx264 preset slow CRF 16, yuv420p, bt709, faststart",
+ "created": "2026-10-19T06:14:57-0500"
+}

masters/roboolympics-fun/badminton_serve/video.mp4

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:53d49e8d1404649fbd1331dffae3745cae61bbd1f82ff29e816261746aa4860a
+size 5400934

masters/roboolympics-fun/basketball_dribble_realgrasp/meta.json

@@ -0,0 +1,98 @@
+{
+ "suite": "roboolympics-fun",
+ "task": "basketball_dribble_realgrasp",
+ "source": "agent-success:codex-gpt6_astra-high-roboolympics-fun-basketball-dribble-realgrasp-i00-privileged/roboolympics-fun-basketball-drib__Lr9SVep",
+ "renderer": "Blender 4.2.23 LTS Cycles path tracing (GPU OptiX), adaptive sampling (threshold 0.01, 64-128 spp), OpenImageDenoise with albedo + normal passes; kinematic playback of the exact MuJoCo replay poses (MuJoCo's own compiled meshes, primitives and colours); daylight: a sun lamp (2 degree disc) and a Nishita sky (no sun disc), the venue's own ground colour on a large ground plane with a faint world-space variation; AgX view transform, Medium High Contrast look; a tracking camera (fixed 22 degree elevation, per-frame fit of the robot and the object in play, smoothed in time) with a slow eased 10-degree orbit",
+ "rendering": "path-tracing",
+ "width": 2560,
+ "height": 1440,
+ "fps": 30,
+ "frames": 719,
+ "seconds": 23.97,
+ "speed": 0.9,
+ "success_check": true,
+ "success_check_how": "the task's own verdict (rcb_rfo.py Episode.verdict(), the referee, from the task dir's tests/ (privileged verifier copy)) at the end of this re-render's own replay of every recorded row from the task's frozen instance (t = 0 state hash checked); a second plain replay with the task's own replay() agrees and ends in the same final state hash (b6ce2d2927f37627)",
+ "end_check": {
+  "success": true,
+  "stages": [
+   "dribble",
+   "past the free-throw line",
+   "in the frontcourt"
+  ],
+  "stages_total": 3,
+  "progress": 1.0,
+  "metrics": {
+   "distance_m": 14.014,
+   "ball_height_m": 0.109,
+   "dribbles": 37
+  },
+  "time_s": 19.72
+ },
+ "recorded_result": {
+  "verifier_reward_json": {
+   "success": 1,
+   "n_actions": 1972,
+   "progress": 1.0,
+   "stages_reached": 3,
+   "stages_total": 3,
+   "metric_ball_height_m": 0.109,
+   "metric_distance_m": 14.014,
+   "metric_dribbles": 37.0
+  }
+ },
+ "replay": {
+  "task_dir": "roboolympics-fun-basketball-dribble-realgrasp-i00-privileged",
+  "trajectory": "jobs/codex/gpt6_astra/high/codex-gpt6_astra-high-roboolympics-fun-basketball-dribble-realgrasp-i00-privileged/roboolympics-fun-basketball-drib__Lr9SVep/artifacts/app/output/trajectory.npz",
+  "n_actions": 1972,
+  "control_steps_replayed": 1972,
+  "control_hz": 100.0,
+  "every_control_steps": 3,
+  "final_state_sha": "b6ce2d2927f37627cd0b83a63b7efead77600ce178e1b6ee685776b4538068e2",
+  "replays_agree": true,
+  "mujoco": "3.3.7",
+  "original_verdict": "counted success (verifier reward 1)"
+ },
+ "visual_changes": [
+  "daylight sky and a plain ground plane in the venue's ground colour instead of the venue's painted sky, hills, trees, grandstands and perimeter boards (drawn-only scenery)",
+  "the ball drawn as a basketball (orange, black channels)"
+ ],
+ "camera": {
+  "mode": "director",
+  "az_deg": -70.04,
+  "az_rule": "robot-travel-70",
+  "el_deg": 22.0,
+  "fovy_deg": 34.0,
+  "fill": 0.72,
+  "distance_m": [
+   4.396,
+   4.459
+  ],
+  "robot_root": "pelvis",
+  "travel_m": 13.514,
+  "focus": {
+   "robot": 3,
+   "robot+object": 656
+  },
+  "objects": {
+   "ball": 13.4
+  }
+ },
+ "orbit_deg": 10.0,
+ "holds": {
+  "start_s": 0.5,
+  "end_s": 1.5,
+  "note": "first/last frame held, no extra simulation"
+ },
+ "seconds_per_frame": 3.09,
+ "image": "ghcr.io/mll-lab-nu/efe:robofunolympics-0.2.1 (simulator) + ghcr.io/mll-lab-nu/efe:roboolympics-fun-files-0.1.0@sha256:49a33342a72b57b56742cc754e427205a30f39dfcd69a99d1d7275c0eae4a619 (rcb_rfo.py, rfo/)",
+ "task_dir": "roboolympics-fun-basketball-dribble-realgrasp-i00-privileged",
+ "agent": "codex + GPT-6 Astra (high)",
+ "run": "codex-gpt6_astra-high",
+ "mode": "privileged",
+ "trial": "codex-gpt6_astra-high-roboolympics-fun-basketball-dribble-realgrasp-i00-privileged/roboolympics-fun-basketball-drib__Lr9SVep",
+ "trial_choice": "a counted GPT-6 Astra success in privileged mode",
+ "poster": "last frame (solved state)",
+ "host": "vast.ai RTX 4090 gpu4",
+ "encode": "libx264 preset slow CRF 16, yuv420p, bt709, faststart",
+ "created": "2026-10-19T06:14:57-0500"
+}

masters/roboolympics-fun/basketball_dribble_realgrasp/video.mp4

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cd73eda06a3a113d7e9b1ef4d7099cd7d517e32ee08ada6c84f1a8052a0ed05b
+size 14060695

masters/roboolympics-fun/bowling_realgrasp/meta.json

@@ -0,0 +1,113 @@
+{
+ "suite": "roboolympics-fun",
+ "task": "bowling_realgrasp",
+ "source": "agent-success:codex-gpt6_astra-high-roboolympics-fun-bowling-realgrasp-i00-privileged/roboolympics-fun-bowling-realgra__pzQLbbU",
+ "renderer": "Blender 4.2.23 LTS Cycles path tracing (GPU OptiX), adaptive sampling (threshold 0.01, 64-128 spp), OpenImageDenoise with albedo + normal passes; kinematic playback of the exact MuJoCo replay poses (MuJoCo's own compiled meshes, primitives and colours); daylight: a sun lamp (2 degree disc) and a Nishita sky (no sun disc), the venue's own ground colour on a large ground plane with a faint world-space variation; AgX view transform, Medium High Contrast look; a tracking camera (fixed 22 degree elevation, per-frame fit of the robot and the object in play, smoothed in time) with a slow eased 10-degree orbit",
+ "rendering": "path-tracing",
+ "width": 2560,
+ "height": 1440,
+ "fps": 30,
+ "frames": 395,
+ "seconds": 13.17,
+ "speed": 0.9,
+ "success_check": true,
+ "success_check_how": "the task's own verdict (rcb_rfo.py Episode.verdict(), the referee, from the task dir's tests/ (privileged verifier copy)) at the end of this re-render's own replay of every recorded row from the task's frozen instance (t = 0 state hash checked); a second plain replay with the task's own replay() agrees and ends in the same final state hash (6f9510f3274e2aed)",
+ "end_check": {
+  "success": true,
+  "stages": [
+   "delivery",
+   "pins",
+   "strike"
+  ],
+  "stages_total": 3,
+  "progress": 1.0,
+  "metrics": {
+   "pins_down": 10,
+   "release_speed_m_s": 6.66,
+   "entry_speed_m_s": 4.14,
+   "entry_y_m": -0.042,
+   "entry_angle_deg": 0.17
+  },
+  "time_s": 10.01
+ },
+ "recorded_result": {
+  "verifier_reward_json": {
+   "success": 1,
+   "n_actions": 1001,
+   "progress": 1.0,
+   "stages_reached": 3,
+   "stages_total": 3,
+   "metric_entry_angle_deg": 0.17,
+   "metric_entry_speed_m_s": 4.14,
+   "metric_entry_y_m": -0.042,
+   "metric_pins_down": 10.0,
+   "metric_release_speed_m_s": 6.66
+  }
+ },
+ "replay": {
+  "task_dir": "roboolympics-fun-bowling-realgrasp-i00-privileged",
+  "trajectory": "jobs/codex/gpt6_astra/high/codex-gpt6_astra-high-roboolympics-fun-bowling-realgrasp-i00-privileged/roboolympics-fun-bowling-realgra__pzQLbbU/artifacts/app/output/trajectory.npz",
+  "n_actions": 1001,
+  "control_steps_replayed": 1001,
+  "control_hz": 100.0,
+  "every_control_steps": 3,
+  "final_state_sha": "6f9510f3274e2aedf403e7de24236b766cbecb9503de7d1d469176ae81ec9a85",
+  "replays_agree": true,
+  "mujoco": "3.3.7",
+  "original_verdict": "counted success (verifier reward 1)"
+ },
+ "visual_changes": [
+  "daylight sky and a plain ground plane in the venue's ground colour instead of the venue's painted sky, hills, trees, grandstands and perimeter boards (drawn-only scenery)"
+ ],
+ "camera": {
+  "mode": "director",
+  "az_deg": -115.29,
+  "az_rule": "object-travel-115",
+  "el_deg": 22.0,
+  "fovy_deg": 34.0,
+  "fill": 0.72,
+  "distance_m": [
+   4.751,
+   30.998
+  ],
+  "robot_root": "pelvis",
+  "travel_m": 0.029,
+  "focus": {
+   "object": 153,
+   "robot": 9,
+   "robot+object": 76,
+   "robot+landed": 97
+  },
+  "objects": {
+   "pin1": 0.61,
+   "pin2": 0.4,
+   "pin3": 0.4,
+   "pin4": 0.41,
+   "pin5": 1.02,
+   "pin6": 0.43,
+   "pin7": 0.18,
+   "pin8": 0.63,
+   "pin9": 0.68,
+   "pin10": 0.33,
+   "ball": 20.35
+  }
+ },
+ "orbit_deg": 10.0,
+ "holds": {
+  "start_s": 0.5,
+  "end_s": 1.5,
+  "note": "first/last frame held, no extra simulation"
+ },
+ "seconds_per_frame": 3.2,
+ "image": "ghcr.io/mll-lab-nu/efe:robofunolympics-0.2.1 (simulator) + ghcr.io/mll-lab-nu/efe:roboolympics-fun-files-0.1.0@sha256:49a33342a72b57b56742cc754e427205a30f39dfcd69a99d1d7275c0eae4a619 (rcb_rfo.py, rfo/)",
+ "task_dir": "roboolympics-fun-bowling-realgrasp-i00-privileged",
+ "agent": "codex + GPT-6 Astra (high)",
+ "run": "codex-gpt6_astra-high",
+ "mode": "privileged",
+ "trial": "codex-gpt6_astra-high-roboolympics-fun-bowling-realgrasp-i00-privileged/roboolympics-fun-bowling-realgra__pzQLbbU",
+ "trial_choice": "a counted GPT-6 Astra success in privileged mode",
+ "poster": "frame 55 of the video's own frames: just after the release / strike, robot and object in view (the last frame no longer shows the robot)",
+ "host": "vast.ai RTX 4090 gpu5",
+ "encode": "libx264 preset slow CRF 16, yuv420p, bt709, faststart",
+ "created": "2026-10-19T06:14:57-0500"
+}

masters/roboolympics-fun/bowling_realgrasp/video.mp4

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c5d81b19a5e433f665ecd609862ba294332df5d114820624622cf6b8c377f97b
+size 6850850

masters/roboolympics-fun/catch_and_shoot_realgrasp/meta.json

@@ -0,0 +1,94 @@
+{
+ "suite": "roboolympics-fun",
+ "task": "catch_and_shoot_realgrasp",
+ "source": "agent-success:codex-gpt6_astra-high-roboolympics-fun-catch-and-shoot-realgrasp-i00-privileged/roboolympics-fun-catch-and-shoot__HyT4GE2",
+ "renderer": "Blender 4.2.23 LTS Cycles path tracing (GPU OptiX), adaptive sampling (threshold 0.01, 64-128 spp), OpenImageDenoise with albedo + normal passes; kinematic playback of the exact MuJoCo replay poses (MuJoCo's own compiled meshes, primitives and colours); daylight: a sun lamp (2 degree disc) and a Nishita sky (no sun disc), the venue's own ground colour on a large ground plane with a faint world-space variation; AgX view transform, Medium High Contrast look; a tracking camera (fixed 22 degree elevation, per-frame fit of the robot and the object in play, smoothed in time) with a slow eased 10-degree orbit",
+ "rendering": "path-tracing",
+ "width": 2560,
+ "height": 1440,
+ "fps": 30,
+ "frames": 271,
+ "seconds": 9.03,
+ "speed": 0.9,
+ "success_check": true,
+ "success_check_how": "the task's own verdict (rcb_rfo.py Episode.verdict(), the referee, from the task dir's tests/ (privileged verifier copy)) at the end of this re-render's own replay of every recorded row from the task's frozen instance (t = 0 state hash checked); a second plain replay with the task's own replay() agrees and ends in the same final state hash (c973e23d370bc98c)",
+ "end_check": {
+  "success": true,
+  "stages": [
+   "catch",
+   "release",
+   "through the ring"
+  ],
+  "stages_total": 3,
+  "progress": 1.0,
+  "metrics": {
+   "ball_height_m": 2.778
+  },
+  "time_s": 6.3
+ },
+ "recorded_result": {
+  "verifier_reward_json": {
+   "success": 1,
+   "n_actions": 630,
+   "progress": 1.0,
+   "stages_reached": 3,
+   "stages_total": 3,
+   "metric_ball_height_m": 2.778
+  }
+ },
+ "replay": {
+  "task_dir": "roboolympics-fun-catch-and-shoot-realgrasp-i00-privileged",
+  "trajectory": "jobs/codex/gpt6_astra/high/codex-gpt6_astra-high-roboolympics-fun-catch-and-shoot-realgrasp-i00-privileged/roboolympics-fun-catch-and-shoot__HyT4GE2/artifacts/app/output/trajectory.npz",
+  "n_actions": 630,
+  "control_steps_replayed": 630,
+  "control_hz": 100.0,
+  "every_control_steps": 3,
+  "final_state_sha": "c973e23d370bc98ce1d4c3f7ef2dfeceacd44d315dbbd10cd54cac10086cc6ce",
+  "replays_agree": true,
+  "mujoco": "3.3.7",
+  "original_verdict": "counted success (verifier reward 1)"
+ },
+ "visual_changes": [
+  "daylight sky and a plain ground plane in the venue's ground colour instead of the venue's painted sky, hills, trees, grandstands and perimeter boards (drawn-only scenery)",
+  "the ball drawn as a basketball (orange, black channels)"
+ ],
+ "camera": {
+  "mode": "director",
+  "az_deg": -159.64,
+  "az_rule": "object-travel-115",
+  "el_deg": 22.0,
+  "fovy_deg": 34.0,
+  "fill": 0.72,
+  "distance_m": [
+   4.501,
+   11.372
+  ],
+  "robot_root": "pelvis",
+  "travel_m": 0.218,
+  "focus": {
+   "robot+object": 111,
+   "robot": 100
+  },
+  "objects": {
+   "ball": 7.0
+  }
+ },
+ "orbit_deg": 10.0,
+ "holds": {
+  "start_s": 0.5,
+  "end_s": 1.5,
+  "note": "first/last frame held, no extra simulation"
+ },
+ "seconds_per_frame": 3.12,
+ "image": "ghcr.io/mll-lab-nu/efe:robofunolympics-0.2.1 (simulator) + ghcr.io/mll-lab-nu/efe:roboolympics-fun-files-0.1.0@sha256:49a33342a72b57b56742cc754e427205a30f39dfcd69a99d1d7275c0eae4a619 (rcb_rfo.py, rfo/)",
+ "task_dir": "roboolympics-fun-catch-and-shoot-realgrasp-i00-privileged",
+ "agent": "codex + GPT-6 Astra (high)",
+ "run": "codex-gpt6_astra-high",
+ "mode": "privileged",
+ "trial": "codex-gpt6_astra-high-roboolympics-fun-catch-and-shoot-realgrasp-i00-privileged/roboolympics-fun-catch-and-shoot__HyT4GE2",
+ "trial_choice": "a counted GPT-6 Astra success in privileged mode",
+ "poster": "last frame (solved state)",
+ "host": "vast.ai RTX 4090 gpu2",
+ "encode": "libx264 preset slow CRF 16, yuv420p, bt709, faststart",
+ "created": "2026-10-19T06:14:58-0500"
+}

masters/roboolympics-fun/catch_and_shoot_realgrasp/video.mp4

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a7f9fb8d37c12968c6ceabf79af2a00344f549f720f953d71e7cca80ab0ea89f
+size 5430197

masters/roboolympics-fun/catch_realgrasp/meta.json

@@ -0,0 +1,95 @@
+{
+ "suite": "roboolympics-fun",
+ "task": "catch_realgrasp",
+ "source": "agent-success:codex-gpt6_astra-high-roboolympics-fun-catch-realgrasp-i00-privileged/roboolympics-fun-catch-realgrasp__VA6Tbng",
+ "renderer": "Blender 4.2.23 LTS Cycles path tracing (GPU OptiX), adaptive sampling (threshold 0.01, 64-128 spp), OpenImageDenoise with albedo + normal passes; kinematic playback of the exact MuJoCo replay poses (MuJoCo's own compiled meshes, primitives and colours); daylight: a sun lamp (2 degree disc) and a Nishita sky (no sun disc), the venue's own ground colour on a large ground plane with a faint world-space variation; AgX view transform, Medium High Contrast look; a tracking camera (fixed 22 degree elevation, per-frame fit of the robot and the object in play, smoothed in time) with a slow eased 10-degree orbit",
+ "rendering": "path-tracing",
+ "width": 2560,
+ "height": 1440,
+ "fps": 30,
+ "frames": 304,
+ "seconds": 10.13,
+ "speed": 0.6,
+ "success_check": true,
+ "success_check_how": "the task's own verdict (rcb_rfo.py Episode.verdict(), the referee, from the task dir's tests/ (privileged verifier copy)) at the end of this re-render's own replay of every recorded row from the task's frozen instance (t = 0 state hash checked); a second plain replay with the task's own replay() agrees and ends in the same final state hash (76522d054933e3be)",
+ "end_check": {
+  "success": true,
+  "stages": [
+   "catch",
+   "hold"
+  ],
+  "stages_total": 2,
+  "progress": 1.0,
+  "metrics": {
+   "catch_height_m": 1.703,
+   "catch_speed_m_s": 2.32
+  },
+  "time_s": 4.86
+ },
+ "recorded_result": {
+  "verifier_reward_json": {
+   "success": 1,
+   "n_actions": 486,
+   "progress": 1.0,
+   "stages_reached": 2,
+   "stages_total": 2,
+   "metric_catch_height_m": 1.703,
+   "metric_catch_speed_m_s": 2.32
+  }
+ },
+ "replay": {
+  "task_dir": "roboolympics-fun-catch-realgrasp-i00-privileged",
+  "trajectory": "jobs/codex/gpt6_astra/high/codex-gpt6_astra-high-roboolympics-fun-catch-realgrasp-i00-privileged/roboolympics-fun-catch-realgrasp__VA6Tbng/artifacts/app/output/trajectory.npz",
+  "n_actions": 486,
+  "control_steps_replayed": 486,
+  "control_hz": 100.0,
+  "every_control_steps": 2,
+  "final_state_sha": "76522d054933e3bed2f00644a141c7a6824ea040a1a0c02f9ed2eddf1d879d84",
+  "replays_agree": true,
+  "mujoco": "3.3.7",
+  "original_verdict": "counted success (verifier reward 1)"
+ },
+ "visual_changes": [
+  "daylight sky and a plain ground plane in the venue's ground colour instead of the venue's painted sky, hills, trees, grandstands and perimeter boards (drawn-only scenery)"
+ ],
+ "camera": {
+  "mode": "director",
+  "az_deg": -87.34,
+  "az_rule": "robot-travel-70",
+  "el_deg": 22.0,
+  "fovy_deg": 34.0,
+  "fill": 0.72,
+  "distance_m": [
+   4.451,
+   17.065
+  ],
+  "robot_root": "pelvis",
+  "travel_m": 2.255,
+  "focus": {
+   "robot": 51,
+   "robot+object": 69,
+   "robot(incoming)": 124
+  },
+  "objects": {
+   "ball": 37.21
+  }
+ },
+ "orbit_deg": 10.0,
+ "holds": {
+  "start_s": 0.5,
+  "end_s": 1.5,
+  "note": "first/last frame held, no extra simulation"
+ },
+ "seconds_per_frame": 4.71,
+ "image": "ghcr.io/mll-lab-nu/efe:robofunolympics-0.2.1 (simulator) + ghcr.io/mll-lab-nu/efe:roboolympics-fun-files-0.1.0@sha256:49a33342a72b57b56742cc754e427205a30f39dfcd69a99d1d7275c0eae4a619 (rcb_rfo.py, rfo/)",
+ "task_dir": "roboolympics-fun-catch-realgrasp-i00-privileged",
+ "agent": "codex + GPT-6 Astra (high)",
+ "run": "codex-gpt6_astra-high",
+ "mode": "privileged",
+ "trial": "codex-gpt6_astra-high-roboolympics-fun-catch-realgrasp-i00-privileged/roboolympics-fun-catch-realgrasp__VA6Tbng",
+ "trial_choice": "a counted GPT-6 Astra success in privileged mode",
+ "poster": "last frame (solved state)",
+ "host": "vast.ai RTX 4090 gpu5",
+ "encode": "libx264 preset slow CRF 16, yuv420p, bt709, faststart",
+ "created": "2026-10-19T06:14:58-0500"
+}

masters/roboolympics-fun/catch_realgrasp/video.mp4

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:aba9d020f2b589a3a5cf16f49b2b1bcf2ad9f59da7774e3694cf0f580d03c196
+size 4737361

masters/roboolympics-fun/discus_realgrasp/meta.json

@@ -0,0 +1,106 @@
+{
+ "suite": "roboolympics-fun",
+ "task": "discus_realgrasp",
+ "source": "agent-success:codex-gpt6_astra-high-roboolympics-fun-discus-realgrasp-i00-privileged/roboolympics-fun-discus-realgras__KRGHtXU",
+ "renderer": "Blender 4.2.23 LTS Cycles path tracing (GPU OptiX), adaptive sampling (threshold 0.01, 64-128 spp), OpenImageDenoise with albedo + normal passes; kinematic playback of the exact MuJoCo replay poses (MuJoCo's own compiled meshes, primitives and colours); daylight: a sun lamp (2 degree disc) and a Nishita sky (no sun disc), the venue's own ground colour on a large ground plane with a faint world-space variation; AgX view transform, Medium High Contrast look; a tracking camera (fixed 22 degree elevation, per-frame fit of the robot and the object in play, smoothed in time) with a slow eased 10-degree orbit",
+ "rendering": "path-tracing",
+ "width": 2560,
+ "height": 1440,
+ "fps": 30,
+ "frames": 375,
+ "seconds": 12.5,
+ "speed": 0.9,
+ "success_check": true,
+ "success_check_how": "the task's own verdict (rcb_rfo.py Episode.verdict(), the referee, from the task dir's tests/ (privileged verifier copy)) at the end of this re-render's own replay of every recorded row from the task's frozen instance (t = 0 state hash checked); a second plain replay with the task's own replay() agrees and ends in the same final state hash (6fbebc682aca570c)",
+ "end_check": {
+  "success": true,
+  "stages": [
+   "release",
+   "landing",
+   "exit"
+  ],
+  "stages_total": 3,
+  "progress": 1.0,
+  "metrics": {
+   "release_time_s": 5.728,
+   "release_speed_m_s": 12.52,
+   "release_height_m": 2.083,
+   "release_tilt_deg": 38.8,
+   "release_spin_rad_s": 39.2,
+   "distance_m": 16.445
+  },
+  "time_s": 9.4
+ },
+ "recorded_result": {
+  "verifier_reward_json": {
+   "success": 1,
+   "n_actions": 940,
+   "progress": 1.0,
+   "stages_reached": 3,
+   "stages_total": 3,
+   "metric_distance_m": 16.445,
+   "metric_release_height_m": 2.083,
+   "metric_release_speed_m_s": 12.52,
+   "metric_release_spin_rad_s": 39.2,
+   "metric_release_tilt_deg": 38.8,
+   "metric_release_time_s": 5.728
+  }
+ },
+ "replay": {
+  "task_dir": "roboolympics-fun-discus-realgrasp-i00-privileged",
+  "trajectory": "jobs/codex/gpt6_astra/high/codex-gpt6_astra-high-roboolympics-fun-discus-realgrasp-i00-privileged/roboolympics-fun-discus-realgras__KRGHtXU/artifacts/app/output/trajectory.npz",
+  "n_actions": 940,
+  "control_steps_replayed": 940,
+  "control_hz": 100.0,
+  "every_control_steps": 3,
+  "final_state_sha": "6fbebc682aca570ce5b507e06fb9a5c45d4fec34f401696d19828c189b7e3604",
+  "replays_agree": true,
+  "mujoco": "3.3.7",
+  "original_verdict": "counted success (verifier reward 1)"
+ },
+ "visual_changes": [
+  "daylight sky and a plain ground plane in the venue's ground colour instead of the venue's painted sky, hills, trees, grandstands and perimeter boards (drawn-only scenery)"
+ ],
+ "camera": {
+  "mode": "director",
+  "az_deg": -109.25,
+  "az_rule": "object-travel-115",
+  "el_deg": 22.0,
+  "fovy_deg": 34.0,
+  "fill": 0.72,
+  "distance_m": [
+   4.478,
+   31.33
+  ],
+  "robot_root": "pelvis",
+  "travel_m": 1.123,
+  "focus": {
+   "robot+landed": 6,
+   "robot+object": 60,
+   "object": 88,
+   "robot": 150,
+   "robot(incoming)": 11
+  },
+  "objects": {
+   "discus": 18.6
+  }
+ },
+ "orbit_deg": 10.0,
+ "holds": {
+  "start_s": 0.5,
+  "end_s": 1.5,
+  "note": "first/last frame held, no extra simulation"
+ },
+ "seconds_per_frame": 2.97,
+ "image": "ghcr.io/mll-lab-nu/efe:robofunolympics-0.2.1 (simulator) + ghcr.io/mll-lab-nu/efe:roboolympics-fun-files-0.1.0@sha256:49a33342a72b57b56742cc754e427205a30f39dfcd69a99d1d7275c0eae4a619 (rcb_rfo.py, rfo/)",
+ "task_dir": "roboolympics-fun-discus-realgrasp-i00-privileged",
+ "agent": "codex + GPT-6 Astra (high)",
+ "run": "codex-gpt6_astra-high",
+ "mode": "privileged",
+ "trial": "codex-gpt6_astra-high-roboolympics-fun-discus-realgrasp-i00-privileged/roboolympics-fun-discus-realgras__KRGHtXU",
+ "trial_choice": "a counted GPT-6 Astra success in privileged mode",
+ "poster": "frame 193 of the video's own frames: just after the release / strike, robot and object in view (the last frame no longer shows the robot)",
+ "host": "vast.ai RTX 4090 gpu0",
+ "encode": "libx264 preset slow CRF 16, yuv420p, bt709, faststart",
+ "created": "2026-10-19T06:14:58-0500"
+}

masters/roboolympics-fun/discus_realgrasp/video.mp4

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:92dbad023beb6dda2d5f6079fafb019c4aa8718f1889c3dc4f9c198ec1bd1ca8
+size 6564168

masters/roboolympics-fun/dunk_between_legs/meta.json

@@ -0,0 +1,102 @@
+{
+ "suite": "roboolympics-fun",
+ "task": "dunk_between_legs",
+ "source": "scene-orbit",
+ "why_no_video": "No evaluated agent has solved this task (no counted success in any run when this was made). As for the other unsolved RoboOlympics-Fun tasks on the site (board 0161 rule), this video shows the best failed attempt instead, replayed exactly (not a solved run): the task's own check at the end says it failed, as the verifier did.",
+ "shows": "agent-attempt (failed):codex-gpt6_astra-high-roboolympics-fun-dunk-between-legs-i00-privileged/roboolympics-fun-dunk-between-le__zHY69cc",
+ "renderer": "Blender 4.2.23 LTS Cycles path tracing (GPU OptiX), adaptive sampling (threshold 0.01, 64-128 spp), OpenImageDenoise with albedo + normal passes; kinematic playback of the exact MuJoCo replay poses (MuJoCo's own compiled meshes, primitives and colours); daylight: a sun lamp (2 degree disc) and a Nishita sky (no sun disc), the venue's own ground colour on a large ground plane with a faint world-space variation; AgX view transform, Medium High Contrast look; a tracking camera (fixed 22 degree elevation, per-frame fit of the robot and the object in play, smoothed in time) with a slow eased 10-degree orbit",
+ "rendering": "path-tracing",
+ "width": 2560,
+ "height": 1440,
+ "fps": 30,
+ "frames": 293,
+ "seconds": 9.77,
+ "speed": 0.9,
+ "success_check": false,
+ "success_check_how": "the task's own verdict (rcb_rfo.py Episode.verdict(), the referee, from the task dir's tests/ (privileged verifier copy)) at the end of this re-render's own replay of every recorded row from the task's frozen instance (t = 0 state hash checked); a second plain replay with the task's own replay() agrees and ends in the same final state hash (4bb20c99b755beb2)",
+ "end_check": {
+  "success": false,
+  "stages": [
+   "dribble",
+   "between the legs",
+   "gather"
+  ],
+  "stages_total": 5,
+  "progress": 0.6,
+  "metrics": {
+   "ball_height_m": 1.344,
+   "basket_touch_s": 0.0,
+   "dribbles": 7,
+   "between_the_legs": 2
+  },
+  "time_s": 6.95
+ },
+ "recorded_result": {
+  "verifier_reward_json": {
+   "success": 0,
+   "n_actions": 695,
+   "progress": 0.6,
+   "stages_reached": 3,
+   "stages_total": 5,
+   "metric_ball_height_m": 1.344,
+   "metric_basket_touch_s": 0.0,
+   "metric_between_the_legs": 2.0,
+   "metric_dribbles": 7.0
+  }
+ },
+ "replay": {
+  "task_dir": "roboolympics-fun-dunk-between-legs-i00-privileged",
+  "trajectory": "jobs/codex/gpt6_astra/high/codex-gpt6_astra-high-roboolympics-fun-dunk-between-legs-i00-privileged/roboolympics-fun-dunk-between-le__zHY69cc/artifacts/app/output/trajectory.npz",
+  "n_actions": 695,
+  "control_steps_replayed": 695,
+  "control_hz": 100.0,
+  "every_control_steps": 3,
+  "final_state_sha": "4bb20c99b755beb27d523b60b987ed763ecfa8c4009abc34a793eaaf4e98c523",
+  "replays_agree": true,
+  "mujoco": "3.3.7",
+  "original_verdict": "counted failure (verifier reward 0)"
+ },
+ "visual_changes": [
+  "daylight sky and a plain ground plane in the venue's ground colour instead of the venue's painted sky, hills, trees, grandstands and perimeter boards (drawn-only scenery)",
+  "the ball drawn as a basketball (orange, black channels)"
+ ],
+ "camera": {
+  "mode": "director",
+  "az_deg": -75.18,
+  "az_rule": "robot-travel-70",
+  "el_deg": 22.0,
+  "fovy_deg": 34.0,
+  "fill": 0.72,
+  "distance_m": [
+   4.203,
+   4.539
+  ],
+  "robot_root": "pelvis",
+  "travel_m": 2.88,
+  "focus": {
+   "robot": 9,
+   "robot+object": 224
+  },
+  "objects": {
+   "ball": 2.89
+  }
+ },
+ "orbit_deg": 10.0,
+ "holds": {
+  "start_s": 0.5,
+  "end_s": 1.5,
+  "note": "first/last frame held, no extra simulation"
+ },
+ "seconds_per_frame": 3.11,
+ "image": "ghcr.io/mll-lab-nu/efe:robofunolympics-0.2.1 (simulator) + ghcr.io/mll-lab-nu/efe:roboolympics-fun-files-0.1.0@sha256:49a33342a72b57b56742cc754e427205a30f39dfcd69a99d1d7275c0eae4a619 (rcb_rfo.py, rfo/)",
+ "task_dir": "roboolympics-fun-dunk-between-legs-i00-privileged",
+ "agent": "codex + GPT-6 Astra (high)",
+ "run": "codex-gpt6_astra-high",
+ "mode": "privileged",
+ "trial": "codex-gpt6_astra-high-roboolympics-fun-dunk-between-legs-i00-privileged/roboolympics-fun-dunk-between-le__zHY69cc",
+ "trial_choice": "no counted success in any run; GPT-6 Astra's privileged attempt with the most progress",
+ "poster": "last frame (end of the failed attempt)",
+ "host": "vast.ai RTX 4090 gpu2",
+ "encode": "libx264 preset slow CRF 16, yuv420p, bt709, faststart",
+ "created": "2026-10-19T06:14:58-0500"
+}

masters/roboolympics-fun/dunk_between_legs/video.mp4

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4616f481c1104f326f8c19c9bbd99addef171e86b228678abd5aee0189b384be
+size 6603886

masters/roboolympics-fun/dunk_realgrasp/meta.json

@@ -0,0 +1,94 @@
+{
+ "suite": "roboolympics-fun",
+ "task": "dunk_realgrasp",
+ "source": "scene-orbit",
+ "why_no_video": "No evaluated agent has solved this task (no counted success in any run when this was made). As for the other unsolved RoboOlympics-Fun tasks on the site (board 0161 rule), this video shows the best failed attempt instead, replayed exactly (not a solved run): the task's own check at the end says it failed, as the verifier did.",
+ "shows": "agent-attempt (failed):codex-gpt6_astra-high-roboolympics-fun-dunk-realgrasp-i00-privileged/roboolympics-dunk-realgrasp-i00__ZWR3UwT",
+ "renderer": "Blender 4.2.23 LTS Cycles path tracing (GPU OptiX), adaptive sampling (threshold 0.01, 64-128 spp), OpenImageDenoise with albedo + normal passes; kinematic playback of the exact MuJoCo replay poses (MuJoCo's own compiled meshes, primitives and colours); daylight: a sun lamp (2 degree disc) and a Nishita sky (no sun disc), the venue's own ground colour on a large ground plane with a faint world-space variation; AgX view transform, Medium High Contrast look; a tracking camera (fixed 22 degree elevation, per-frame fit of the robot and the object in play, smoothed in time) with a slow eased 10-degree orbit",
+ "rendering": "path-tracing",
+ "width": 2560,
+ "height": 1440,
+ "fps": 30,
+ "frames": 353,
+ "seconds": 11.77,
+ "speed": 0.3,
+ "success_check": false,
+ "success_check_how": "the task's own verdict (rcb_rfo.py Episode.verdict(), the referee, from the task dir's tests/ (privileged verifier copy)) at the end of this re-render's own replay of every recorded row from the task's frozen instance (t = 0 state hash checked); a second plain replay with the task's own replay() agrees and ends in the same final state hash (ad8cf0efa21c1183)",
+ "end_check": {
+  "success": false,
+  "stages": [],
+  "stages_total": 2,
+  "progress": 0.0,
+  "metrics": {
+   "ball_height_m": 3.073,
+   "basket_touch_s": 0.0
+  },
+  "time_s": 2.92
+ },
+ "recorded_result": {
+  "verifier_reward_json": {
+   "success": 0,
+   "n_actions": 292,
+   "progress": 0.0,
+   "stages_reached": 0,
+   "stages_total": 2,
+   "metric_ball_height_m": 3.073,
+   "metric_basket_touch_s": 0.0
+  }
+ },
+ "replay": {
+  "task_dir": "roboolympics-fun-dunk-realgrasp-i00-privileged",
+  "trajectory": "jobs/codex/gpt6_astra/high/codex-gpt6_astra-high-roboolympics-fun-dunk-realgrasp-i00-privileged/roboolympics-dunk-realgrasp-i00__ZWR3UwT/artifacts/app/output/trajectory.npz",
+  "n_actions": 292,
+  "control_steps_replayed": 292,
+  "control_hz": 100.0,
+  "every_control_steps": 1,
+  "final_state_sha": "ad8cf0efa21c118372d37a33acc3b2cdc11848f220203d49c9d898f82877dcd1",
+  "replays_agree": true,
+  "mujoco": "3.3.7",
+  "original_verdict": "counted failure (verifier reward 0)"
+ },
+ "visual_changes": [
+  "daylight sky and a plain ground plane in the venue's ground colour instead of the venue's painted sky, hills, trees, grandstands and perimeter boards (drawn-only scenery)",
+  "the ball drawn as a basketball (orange, black channels)"
+ ],
+ "camera": {
+  "mode": "director",
+  "az_deg": -50.0,
+  "az_rule": "facing-50",
+  "el_deg": 22.0,
+  "fovy_deg": 34.0,
+  "fill": 0.72,
+  "distance_m": [
+   4.205,
+   6.348
+  ],
+  "robot_root": "pelvis",
+  "travel_m": 0.109,
+  "focus": {
+   "robot+object": 190,
+   "robot": 103
+  },
+  "objects": {
+   "ball": 0.48
+  }
+ },
+ "orbit_deg": 10.0,
+ "holds": {
+  "start_s": 0.5,
+  "end_s": 1.5,
+  "note": "first/last frame held, no extra simulation"
+ },
+ "seconds_per_frame": 3.13,
+ "image": "ghcr.io/mll-lab-nu/efe:robofunolympics-0.2.1 (simulator) + ghcr.io/mll-lab-nu/efe:roboolympics-fun-files-0.1.0@sha256:49a33342a72b57b56742cc754e427205a30f39dfcd69a99d1d7275c0eae4a619 (rcb_rfo.py, rfo/)",
+ "task_dir": "roboolympics-fun-dunk-realgrasp-i00-privileged",
+ "agent": "codex + GPT-6 Astra (high)",
+ "run": "codex-gpt6_astra-high",
+ "mode": "privileged",
+ "trial": "codex-gpt6_astra-high-roboolympics-fun-dunk-realgrasp-i00-privileged/roboolympics-dunk-realgrasp-i00__ZWR3UwT",
+ "trial_choice": "no counted success in any run; GPT-6 Astra's privileged attempt with the most progress",
+ "poster": "last frame (end of the failed attempt)",
+ "host": "vast.ai RTX 4090 gpu2",
+ "encode": "libx264 preset slow CRF 16, yuv420p, bt709, faststart",
+ "created": "2026-10-19T06:14:58-0500"
+}

masters/roboolympics-fun/dunk_realgrasp/video.mp4

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5ca48ccd789afd1f88b9b3eeeeaaeab135c13948dbc1c83cee432a2aff092320
+size 6537624

masters/roboolympics-fun/free_throw_realgrasp/cover.mp4

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:584d136c094fcfe5974c90104906c72f2f01581b206d4db2f58d972f83b4af35
+size 1444422

masters/roboolympics-fun/free_throw_realgrasp/meta.json

@@ -0,0 +1,127 @@
+{
+ "suite": "roboolympics-fun",
+ "task": "free_throw_realgrasp",
+ "source": "agent-success:codex-gpt6_astra-high-roboolympics-fun-free-throw-realgrasp-i00-privileged/roboolympics-fun-free-throw-real__Xefgnw6",
+ "renderer": "Blender 4.2.23 LTS Cycles path tracing (GPU OptiX), adaptive sampling (threshold 0.01, 64-128 spp), OpenImageDenoise with albedo + normal passes; kinematic playback of the exact MuJoCo replay poses (MuJoCo's own compiled meshes, primitives and colours); daylight: a sun lamp (2 degree disc) and a Nishita sky (no sun disc), the venue's own ground colour on a large ground plane with a faint world-space variation; AgX view transform, Medium High Contrast look; a tracking camera (fixed 22 degree elevation, per-frame fit of the robot and the object in play, smoothed in time) with a slow eased 10-degree orbit",
+ "rendering": "path-tracing",
+ "width": 2560,
+ "height": 1440,
+ "fps": 30,
+ "frames": 329,
+ "seconds": 10.97,
+ "speed": 0.3,
+ "success_check": true,
+ "success_check_how": "the task's own verdict (rcb_rfo.py Episode.verdict(), the referee, from the task dir's tests/ (privileged verifier copy)) at the end of this re-render's own replay of every recorded row from the task's frozen instance (t = 0 state hash checked); a second plain replay with the task's own replay() agrees and ends in the same final state hash (f32ff0267bba07e3)",
+ "end_check": {
+  "success": true,
+  "stages": [
+   "release",
+   "through the ring"
+  ],
+  "stages_total": 2,
+  "progress": 1.0,
+  "metrics": {
+   "ball_height_m": 2.782
+  },
+  "time_s": 2.68
+ },
+ "recorded_result": {
+  "verifier_reward_json": {
+   "success": 1,
+   "n_actions": 268,
+   "progress": 1.0,
+   "stages_reached": 2,
+   "stages_total": 2,
+   "metric_ball_height_m": 2.782
+  }
+ },
+ "replay": {
+  "task_dir": "roboolympics-fun-free-throw-realgrasp-i00-privileged",
+  "trajectory": "jobs/codex/gpt6_astra/high/codex-gpt6_astra-high-roboolympics-fun-free-throw-realgrasp-i00-privileged/roboolympics-fun-free-throw-real__Xefgnw6/artifacts/app/output/trajectory.npz",
+  "n_actions": 268,
+  "control_steps_replayed": 268,
+  "control_hz": 100.0,
+  "every_control_steps": 1,
+  "final_state_sha": "f32ff0267bba07e3eced2a23f566e6e12fb2d6456fa7fc6465b9cf634a485905",
+  "replays_agree": true,
+  "mujoco": "3.3.7",
+  "original_verdict": "counted success (verifier reward 1)"
+ },
+ "visual_changes": [
+  "daylight sky and a plain ground plane in the venue's ground colour instead of the venue's painted sky, hills, trees, grandstands and perimeter boards (drawn-only scenery)",
+  "the ball drawn as a basketball (orange, black channels)"
+ ],
+ "camera": {
+  "mode": "director",
+  "az_deg": -114.95,
+  "az_rule": "object-travel-115",
+  "el_deg": 22.0,
+  "fovy_deg": 34.0,
+  "fill": 0.72,
+  "distance_m": [
+   4.928,
+   10.497
+  ],
+  "robot_root": "pelvis",
+  "travel_m": 0.283,
+  "focus": {
+   "robot": 8,
+   "robot+object": 261
+  },
+  "objects": {
+   "ball": 4.35
+  }
+ },
+ "orbit_deg": 10.0,
+ "holds": {
+  "start_s": 0.5,
+  "end_s": 1.5,
+  "note": "first/last frame held, no extra simulation"
+ },
+ "seconds_per_frame": 3.15,
+ "image": "ghcr.io/mll-lab-nu/efe:robofunolympics-0.2.1 (simulator) + ghcr.io/mll-lab-nu/efe:roboolympics-fun-files-0.1.0@sha256:49a33342a72b57b56742cc754e427205a30f39dfcd69a99d1d7275c0eae4a619 (rcb_rfo.py, rfo/)",
+ "task_dir": "roboolympics-fun-free-throw-realgrasp-i00-privileged",
+ "agent": "codex + GPT-6 Astra (high)",
+ "run": "codex-gpt6_astra-high",
+ "mode": "privileged",
+ "trial": "codex-gpt6_astra-high-roboolympics-fun-free-throw-realgrasp-i00-privileged/roboolympics-fun-free-throw-real__Xefgnw6",
+ "trial_choice": "a counted GPT-6 Astra success in privileged mode",
+ "poster": "last frame (solved state)",
+ "host": "vast.ai RTX 4090 gpu0",
+ "encode": "libx264 preset slow CRF 16, yuv420p, bt709, faststart",
+ "created": "2026-10-19T06:14:59-0500",
+ "cover": {
+  "file": "cover.mp4",
+  "role": "RoboOlympics-Fun suite cover clip on the website (owner's choice)",
+  "width": 2560,
+  "height": 1440,
+  "fps": 30,
+  "frames": 150,
+  "seconds": 5.0,
+  "loop": "seamless",
+  "how": "a second Cycles render (same renderer, settings and look) of the same replay with a fixed camera (no orbit) that keeps the robot, the ball's whole flight and the hoop in frame: every 2nd control step (0.6x), all 135 frames; cut = frames 15..end + a 1.0 s hold of the last frame (ball through the ring), whose last 0.5 s cross-fades into frames 0..14, so the clip loops without a jump",
+  "replay_check": {
+   "reproduced": true,
+   "final_state_sha": "f32ff0267bba07e3eced2a23f566e6e12fb2d6456fa7fc6465b9cf634a485905"
+  },
+  "camera": {
+   "mode": "fixed",
+   "az_deg": -125.0,
+   "az_rule": "fixed",
+   "el_deg": 20.0,
+   "fovy_deg": 34.0,
+   "fill": 0.85,
+   "distance_m": [
+    8.661,
+    8.661
+   ],
+   "robot_root": "pelvis",
+   "travel_m": 0.283,
+   "focus": {},
+   "objects": {
+    "ball": 4.35
+   }
+  },
+  "encode": "libx264 preset slow CRF 16, yuv420p, bt709, faststart"
+ }
+}

masters/roboolympics-fun/free_throw_realgrasp/video.mp4

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3be6f47f0d066ed1120e06bf734f0dab29c5dc2700eee66432c39341e9e72cc0
+size 5790627

masters/roboolympics-fun/golf_putt_realgrasp/meta.json

@@ -0,0 +1,102 @@
+{
+ "suite": "roboolympics-fun",
+ "task": "golf_putt_realgrasp",
+ "source": "agent-success:codex-gpt6_astra-high-roboolympics-fun-golf-putt-realgrasp-i00-privileged/roboolympics-fun-golf-putt-realg__WTrN5e2",
+ "renderer": "Blender 4.2.23 LTS Cycles path tracing (GPU OptiX), adaptive sampling (threshold 0.01, 64-128 spp), OpenImageDenoise with albedo + normal passes; kinematic playback of the exact MuJoCo replay poses (MuJoCo's own compiled meshes, primitives and colours); daylight: a sun lamp (2 degree disc) and a Nishita sky (no sun disc), the venue's own ground colour on a large ground plane with a faint world-space variation; AgX view transform, Medium High Contrast look; a tracking camera (fixed 22 degree elevation, per-frame fit of the robot and the object in play, smoothed in time) with a slow eased 10-degree orbit",
+ "rendering": "path-tracing",
+ "width": 2560,
+ "height": 1440,
+ "fps": 30,
+ "frames": 271,
+ "seconds": 9.03,
+ "speed": 0.9,
+ "success_check": true,
+ "success_check_how": "the task's own verdict (rcb_rfo.py Episode.verdict(), the referee, from the task dir's tests/ (privileged verifier copy)) at the end of this re-render's own replay of every recorded row from the task's frozen instance (t = 0 state hash checked); a second plain replay with the task's own replay() agrees and ends in the same final state hash (cdb8e6c1440e612d)",
+ "end_check": {
+  "success": true,
+  "stages": [
+   "stroke",
+   "holed"
+  ],
+  "stages_total": 2,
+  "progress": 1.0,
+  "metrics": {
+   "ball_to_hole_m": 0.007,
+   "ball_speed_m_s": 0.006,
+   "head_contact_s": 0.013,
+   "head_hits": 1,
+   "grip_break_max_s": 0.002
+  },
+  "time_s": 6.28
+ },
+ "recorded_result": {
+  "verifier_reward_json": {
+   "success": 1,
+   "n_actions": 628,
+   "progress": 1.0,
+   "stages_reached": 2,
+   "stages_total": 2,
+   "metric_ball_speed_m_s": 0.006,
+   "metric_ball_to_hole_m": 0.007,
+   "metric_grip_break_max_s": 0.002,
+   "metric_head_contact_s": 0.013,
+   "metric_head_hits": 1.0
+  }
+ },
+ "replay": {
+  "task_dir": "roboolympics-fun-golf-putt-realgrasp-i00-privileged",
+  "trajectory": "jobs/codex/gpt6_astra/high/codex-gpt6_astra-high-roboolympics-fun-golf-putt-realgrasp-i00-privileged/roboolympics-fun-golf-putt-realg__WTrN5e2/artifacts/app/output/trajectory.npz",
+  "n_actions": 628,
+  "control_steps_replayed": 628,
+  "control_hz": 100.0,
+  "every_control_steps": 3,
+  "final_state_sha": "cdb8e6c1440e612d3cc2ed01eea58f9d6251b8296ce3691aee425c86b67bb0ae",
+  "replays_agree": true,
+  "mujoco": "3.3.7",
+  "original_verdict": "counted success (verifier reward 1)"
+ },
+ "visual_changes": [
+  "daylight sky and a plain ground plane in the venue's ground colour instead of the venue's painted sky, hills, trees, grandstands and perimeter boards (drawn-only scenery)"
+ ],
+ "camera": {
+  "mode": "director",
+  "az_deg": -50.0,
+  "az_rule": "facing-50",
+  "el_deg": 22.0,
+  "fovy_deg": 34.0,
+  "fill": 0.72,
+  "distance_m": [
+   3.269,
+   3.61
+  ],
+  "robot_root": "pelvis",
+  "travel_m": 0.003,
+  "focus": {
+   "robot+object": 110,
+   "robot+landed": 16,
+   "robot": 85
+  },
+  "objects": {
+   "putter": 0.41,
+   "ball": 3.0
+  }
+ },
+ "orbit_deg": 10.0,
+ "holds": {
+  "start_s": 0.5,
+  "end_s": 1.5,
+  "note": "first/last frame held, no extra simulation"
+ },
+ "seconds_per_frame": 2.95,
+ "image": "ghcr.io/mll-lab-nu/efe:robofunolympics-0.2.1 (simulator) + ghcr.io/mll-lab-nu/efe:roboolympics-fun-files-0.1.0@sha256:49a33342a72b57b56742cc754e427205a30f39dfcd69a99d1d7275c0eae4a619 (rcb_rfo.py, rfo/)",
+ "task_dir": "roboolympics-fun-golf-putt-realgrasp-i00-privileged",
+ "agent": "codex + GPT-6 Astra (high)",
+ "run": "codex-gpt6_astra-high",
+ "mode": "privileged",
+ "trial": "codex-gpt6_astra-high-roboolympics-fun-golf-putt-realgrasp-i00-privileged/roboolympics-fun-golf-putt-realg__WTrN5e2",
+ "trial_choice": "a counted GPT-6 Astra success in privileged mode",
+ "poster": "last frame (solved state)",
+ "host": "vast.ai RTX 4090 gpu1",
+ "encode": "libx264 preset slow CRF 16, yuv420p, bt709, faststart",
+ "created": "2026-10-19T06:14:59-0500"
+}

masters/roboolympics-fun/golf_putt_realgrasp/video.mp4

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:28e4c97a70a76b9f58216e73cc73251ed906164eac2b149923a40df44373c942
+size 4783529

masters/roboolympics-fun/hurdles_freestanding/meta.json

@@ -0,0 +1,100 @@
+{
+ "suite": "roboolympics-fun",
+ "task": "hurdles_freestanding",
+ "source": "agent-success:codex-gpt6_astra-high-roboolympics-fun-hurdles-freestanding-i00-privileged/roboolympics-fun-hurdles-freesta__HDyKuzX",
+ "renderer": "Blender 4.2.23 LTS Cycles path tracing (GPU OptiX), adaptive sampling (threshold 0.01, 64-128 spp), OpenImageDenoise with albedo + normal passes; kinematic playback of the exact MuJoCo replay poses (MuJoCo's own compiled meshes, primitives and colours); daylight: a sun lamp (2 degree disc) and a Nishita sky (no sun disc), the venue's own ground colour on a large ground plane with a faint world-space variation; AgX view transform, Medium High Contrast look; a tracking camera (fixed 22 degree elevation, per-frame fit of the robot and the object in play, smoothed in time) with a slow eased 10-degree orbit",
+ "rendering": "path-tracing",
+ "width": 2560,
+ "height": 1440,
+ "fps": 30,
+ "frames": 803,
+ "seconds": 26.77,
+ "speed": 4.5,
+ "success_check": true,
+ "success_check_how": "the task's own verdict (rcb_rfo.py Episode.verdict(), the referee, from the task dir's tests/ (privileged verifier copy)) at the end of this re-render's own replay of every recorded row from the task's frozen instance (t = 0 state hash checked); a second plain replay with the task's own replay() agrees and ends in the same final state hash (d18f85870df6e886)",
+ "end_check": {
+  "success": true,
+  "stages": [
+   "hurdle 1",
+   "hurdle 2",
+   "hurdle 3",
+   "hurdle 4",
+   "finish"
+  ],
+  "stages_total": 5,
+  "progress": 1.0,
+  "metrics": {
+   "hurdles_cleared": 4,
+   "distance_m": 49.36,
+   "hurdles_down": 1,
+   "time_s": 111.217
+  },
+  "time_s": 111.22
+ },
+ "recorded_result": {
+  "verifier_reward_json": {
+   "success": 1,
+   "n_actions": 11122,
+   "progress": 1.0,
+   "stages_reached": 5,
+   "stages_total": 5,
+   "metric_distance_m": 49.36,
+   "metric_hurdles_cleared": 4.0,
+   "metric_hurdles_down": 1.0,
+   "metric_time_s": 111.217
+  }
+ },
+ "replay": {
+  "task_dir": "roboolympics-fun-hurdles-freestanding-i00-privileged",
+  "trajectory": "jobs/codex/gpt6_astra/high/codex-gpt6_astra-high-roboolympics-fun-hurdles-freestanding-i00-privileged/roboolympics-fun-hurdles-freesta__HDyKuzX/artifacts/app/output/trajectory.npz",
+  "n_actions": 11122,
+  "control_steps_replayed": 11122,
+  "control_hz": 100.0,
+  "every_control_steps": 15,
+  "final_state_sha": "d18f85870df6e886cf46bdcfd1e512f6076e58c72cc56c4e33017a8e843309ef",
+  "replays_agree": true,
+  "mujoco": "3.3.7",
+  "original_verdict": "counted success (verifier reward 1)"
+ },
+ "visual_changes": [
+  "daylight sky and a plain ground plane in the venue's ground colour instead of the venue's painted sky, hills, trees, grandstands and perimeter boards (drawn-only scenery)",
+  "the finish posts (drawn-only, at the finish line) not drawn: they hid the runner at the end"
+ ],
+ "camera": {
+  "mode": "director",
+  "az_deg": -69.99,
+  "az_rule": "robot-travel-70",
+  "el_deg": 22.0,
+  "fovy_deg": 34.0,
+  "fill": 0.72,
+  "distance_m": [
+   3.99,
+   5.644
+  ],
+  "robot_root": "pelvis",
+  "travel_m": 49.583,
+  "focus": {
+   "robot": 743
+  },
+  "objects": {},
+  "note": "the hurdles are not followed by the camera (a knocked hurdle would otherwise pull the camera out to a 40 m wide shot)"
+ },
+ "orbit_deg": 10.0,
+ "holds": {
+  "start_s": 0.5,
+  "end_s": 1.5,
+  "note": "first/last frame held, no extra simulation"
+ },
+ "seconds_per_frame": 2.96,
+ "image": "ghcr.io/mll-lab-nu/efe:robofunolympics-0.2.1 (simulator) + ghcr.io/mll-lab-nu/efe:roboolympics-fun-files-0.1.0@sha256:49a33342a72b57b56742cc754e427205a30f39dfcd69a99d1d7275c0eae4a619 (rcb_rfo.py, rfo/)",
+ "task_dir": "roboolympics-fun-hurdles-freestanding-i00-privileged",
+ "agent": "codex + GPT-6 Astra (high)",
+ "run": "codex-gpt6_astra-high",
+ "mode": "privileged",
+ "trial": "codex-gpt6_astra-high-roboolympics-fun-hurdles-freestanding-i00-privileged/roboolympics-fun-hurdles-freesta__HDyKuzX",
+ "trial_choice": "a counted GPT-6 Astra success in privileged mode",
+ "poster": "frame 406 of the video's own frames: clearing hurdle 2 (the last frame shows the robot fallen past the finish line)",
+ "host": "vast.ai RTX 4090 gpu1",
+ "encode": "libx264 preset slow CRF 16, yuv420p, bt709, faststart",
+ "created": "2026-10-19T06:14:59-0500"
+}

masters/roboolympics-fun/hurdles_freestanding/video.mp4

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8b6717d105928a0be34762cc0b8fb363b2e3b0a0fe3d93659cfe83694ad164bd
+size 19964455

masters/roboolympics-fun/javelin_realgrasp/meta.json

@@ -0,0 +1,106 @@
+{
+ "suite": "roboolympics-fun",
+ "task": "javelin_realgrasp",
+ "source": "agent-success:codex-gpt6_astra-high-roboolympics-fun-javelin-realgrasp-i00-privileged/roboolympics-fun-javelin-realgra__A5g8BUH",
+ "renderer": "Blender 4.2.23 LTS Cycles path tracing (GPU OptiX), adaptive sampling (threshold 0.01, 64-128 spp), OpenImageDenoise with albedo + normal passes; kinematic playback of the exact MuJoCo replay poses (MuJoCo's own compiled meshes, primitives and colours); daylight: a sun lamp (2 degree disc) and a Nishita sky (no sun disc), the venue's own ground colour on a large ground plane with a faint world-space variation; AgX view transform, Medium High Contrast look; a tracking camera (fixed 22 degree elevation, per-frame fit of the robot and the object in play, smoothed in time) with a slow eased 10-degree orbit",
+ "rendering": "path-tracing",
+ "width": 2560,
+ "height": 1440,
+ "fps": 30,
+ "frames": 655,
+ "seconds": 21.83,
+ "speed": 0.9,
+ "success_check": true,
+ "success_check_how": "the task's own verdict (rcb_rfo.py Episode.verdict(), the referee, from the task dir's tests/ (privileged verifier copy)) at the end of this re-render's own replay of every recorded row from the task's frozen instance (t = 0 state hash checked); a second plain replay with the task's own replay() agrees and ends in the same final state hash (cc64122ed7b7f287)",
+ "end_check": {
+  "success": true,
+  "stages": [
+   "release",
+   "landing",
+   "exit"
+  ],
+  "stages_total": 3,
+  "progress": 1.0,
+  "metrics": {
+   "release_speed_m_s": 15.05,
+   "release_height_m": 1.548,
+   "release_elevation_deg": 32.3,
+   "release_attitude_deg": -69.0,
+   "distance_m": 26.653,
+   "head_first": 1.0
+  },
+  "time_s": 17.8
+ },
+ "recorded_result": {
+  "verifier_reward_json": {
+   "success": 1,
+   "n_actions": 1780,
+   "progress": 1.0,
+   "stages_reached": 3,
+   "stages_total": 3,
+   "metric_distance_m": 26.653,
+   "metric_head_first": 1.0,
+   "metric_release_attitude_deg": -69.0,
+   "metric_release_elevation_deg": 32.3,
+   "metric_release_height_m": 1.548,
+   "metric_release_speed_m_s": 15.05
+  }
+ },
+ "replay": {
+  "task_dir": "roboolympics-fun-javelin-realgrasp-i00-privileged",
+  "trajectory": "jobs/codex/gpt6_astra/high/codex-gpt6_astra-high-roboolympics-fun-javelin-realgrasp-i00-privileged/roboolympics-fun-javelin-realgra__A5g8BUH/artifacts/app/output/trajectory.npz",
+  "n_actions": 1780,
+  "control_steps_replayed": 1780,
+  "control_hz": 100.0,
+  "every_control_steps": 3,
+  "final_state_sha": "cc64122ed7b7f2879d222b26e078b5635f609776a70f266f5d33a26cf4d64993",
+  "replays_agree": true,
+  "mujoco": "3.3.7",
+  "original_verdict": "counted success (verifier reward 1)"
+ },
+ "visual_changes": [
+  "daylight sky and a plain ground plane in the venue's ground colour instead of the venue's painted sky, hills, trees, grandstands and perimeter boards (drawn-only scenery)"
+ ],
+ "camera": {
+  "mode": "director",
+  "az_deg": 109.64,
+  "az_rule": "robot-travel-70",
+  "el_deg": 22.0,
+  "fovy_deg": 34.0,
+  "fill": 0.72,
+  "distance_m": [
+   5.18,
+   39.821
+  ],
+  "robot_root": "pelvis",
+  "travel_m": 2.98,
+  "focus": {
+   "robot+landed": 359,
+   "object": 82,
+   "robot+object": 22,
+   "robot(incoming)": 19,
+   "robot": 113
+  },
+  "objects": {
+   "javelin": 29.19
+  }
+ },
+ "orbit_deg": 10.0,
+ "holds": {
+  "start_s": 0.5,
+  "end_s": 1.5,
+  "note": "first/last frame held, no extra simulation"
+ },
+ "seconds_per_frame": 2.95,
+ "image": "ghcr.io/mll-lab-nu/efe:robofunolympics-0.2.1 (simulator) + ghcr.io/mll-lab-nu/efe:roboolympics-fun-files-0.1.0@sha256:49a33342a72b57b56742cc754e427205a30f39dfcd69a99d1d7275c0eae4a619 (rcb_rfo.py, rfo/)",
+ "task_dir": "roboolympics-fun-javelin-realgrasp-i00-privileged",
+ "agent": "codex + GPT-6 Astra (high)",
+ "run": "codex-gpt6_astra-high",
+ "mode": "privileged",
+ "trial": "codex-gpt6_astra-high-roboolympics-fun-javelin-realgrasp-i00-privileged/roboolympics-fun-javelin-realgra__A5g8BUH",
+ "trial_choice": "a counted GPT-6 Astra success in privileged mode",
+ "poster": "frame 122 of the video's own frames: just after the release / strike, robot and object in view (the last frame no longer shows the robot)",
+ "host": "vast.ai RTX 4090 gpu3",
+ "encode": "libx264 preset slow CRF 16, yuv420p, bt709, faststart",
+ "created": "2026-10-19T06:14:59-0500"
+}

masters/roboolympics-fun/javelin_realgrasp/video.mp4

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8736c238dab89cf6c0ee73c761ed3eadee03d67fe2c0b14306dbe486b1b79f91
+size 10978009

masters/roboolympics-fun/juggle_three_realgrasp/meta.json

@@ -0,0 +1,96 @@
+{
+ "suite": "roboolympics-fun",
+ "task": "juggle_three_realgrasp",
+ "source": "scene-orbit",
+ "why_no_video": "No evaluated agent has solved this task (no counted success in any run when this was made). As for the other unsolved RoboOlympics-Fun tasks on the site (board 0161 rule), this video shows the best failed attempt instead, replayed exactly (not a solved run): the task's own check at the end says it failed, as the verifier did.",
+ "shows": "agent-attempt (failed):codex-gpt6_astra-high-roboolympics-fun-juggle-three-realgrasp-i00-privileged/roboolympics-fun-juggle-three-re__L74C433",
+ "renderer": "Blender 4.2.23 LTS Cycles path tracing (GPU OptiX), adaptive sampling (threshold 0.01, 64-128 spp), OpenImageDenoise with albedo + normal passes; kinematic playback of the exact MuJoCo replay poses (MuJoCo's own compiled meshes, primitives and colours); daylight: a sun lamp (2 degree disc) and a Nishita sky (no sun disc), the venue's own ground colour on a large ground plane with a faint world-space variation; AgX view transform, Medium High Contrast look; a tracking camera (fixed 22 degree elevation, per-frame fit of the robot and the object in play, smoothed in time) with a slow eased 10-degree orbit",
+ "rendering": "path-tracing",
+ "width": 2560,
+ "height": 1440,
+ "fps": 30,
+ "frames": 268,
+ "seconds": 8.93,
+ "speed": 0.6,
+ "success_check": false,
+ "success_check_how": "the task's own verdict (rcb_rfo.py Episode.verdict(), the referee, from the task dir's tests/ (privileged verifier copy)) at the end of this re-render's own replay of every recorded row from the task's frozen instance (t = 0 state hash checked); a second plain replay with the task's own replay() agrees and ends in the same final state hash (9a43c97a376f5be6)",
+ "end_check": {
+  "success": false,
+  "stages": [
+   "first catch",
+   "half way"
+  ],
+  "stages_total": 3,
+  "progress": 0.6667,
+  "metrics": {
+   "catches": 5
+  },
+  "time_s": 4.14
+ },
+ "recorded_result": {
+  "verifier_reward_json": {
+   "success": 0,
+   "n_actions": 414,
+   "progress": 0.6667,
+   "stages_reached": 2,
+   "stages_total": 3,
+   "metric_catches": 5.0
+  }
+ },
+ "replay": {
+  "task_dir": "roboolympics-fun-juggle-three-realgrasp-i00-privileged",
+  "trajectory": "jobs/codex/gpt6_astra/high/codex-gpt6_astra-high-roboolympics-fun-juggle-three-realgrasp-i00-privileged/roboolympics-fun-juggle-three-re__L74C433/artifacts/app/output/trajectory.npz",
+  "n_actions": 414,
+  "control_steps_replayed": 414,
+  "control_hz": 100.0,
+  "every_control_steps": 2,
+  "final_state_sha": "9a43c97a376f5be63e274a088d2b8ca3d1a845237628b9fa943a10b77c08baaa",
+  "replays_agree": true,
+  "mujoco": "3.3.7",
+  "original_verdict": "counted failure (verifier reward 0)"
+ },
+ "visual_changes": [
+  "daylight sky and a plain ground plane in the venue's ground colour instead of the venue's painted sky, hills, trees, grandstands and perimeter boards (drawn-only scenery)"
+ ],
+ "camera": {
+  "mode": "director",
+  "az_deg": -50.0,
+  "az_rule": "facing-50",
+  "el_deg": 22.0,
+  "fovy_deg": 34.0,
+  "fill": 0.72,
+  "distance_m": [
+   3.186,
+   4.292
+  ],
+  "robot_root": "pelvis",
+  "travel_m": 0.112,
+  "focus": {
+   "robot+object": 107,
+   "robot": 101
+  },
+  "objects": {
+   "ball_a": 0.26,
+   "ball_b": 0.15,
+   "ball_c": 0.34
+  }
+ },
+ "orbit_deg": 10.0,
+ "holds": {
+  "start_s": 0.5,
+  "end_s": 1.5,
+  "note": "first/last frame held, no extra simulation"
+ },
+ "seconds_per_frame": 3.08,
+ "image": "ghcr.io/mll-lab-nu/efe:robofunolympics-0.2.1 (simulator) + ghcr.io/mll-lab-nu/efe:roboolympics-fun-files-0.1.0@sha256:49a33342a72b57b56742cc754e427205a30f39dfcd69a99d1d7275c0eae4a619 (rcb_rfo.py, rfo/)",
+ "task_dir": "roboolympics-fun-juggle-three-realgrasp-i00-privileged",
+ "agent": "codex + GPT-6 Astra (high)",
+ "run": "codex-gpt6_astra-high",
+ "mode": "privileged",
+ "trial": "codex-gpt6_astra-high-roboolympics-fun-juggle-three-realgrasp-i00-privileged/roboolympics-fun-juggle-three-re__L74C433",
+ "trial_choice": "no counted success in any run; GPT-6 Astra's privileged attempt with the most progress",
+ "poster": "last frame (end of the failed attempt)",
+ "host": "vast.ai RTX 4090 gpu3",
+ "encode": "libx264 preset slow CRF 16, yuv420p, bt709, faststart",
+ "created": "2026-10-19T06:20:23-0500"
+}

masters/roboolympics-fun/juggle_three_realgrasp/video.mp4

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:25f7b083a6bb763532af725f05bd1b085f2e8fe09c4ffb93f460d16b4e607601
+size 3831798

masters/roboolympics-fun/juggling_feet/meta.json

@@ -0,0 +1,104 @@
+{
+ "suite": "roboolympics-fun",
+ "task": "juggling_feet",
+ "source": "agent-success:codex-gpt6_astra-high-roboolympics-fun-juggling-feet-i00-privileged/roboolympics-juggling-feet-i00-p__p6ohQgr",
+ "renderer": "Blender 4.2.23 LTS Cycles path tracing (GPU OptiX), adaptive sampling (threshold 0.01, 64-128 spp), OpenImageDenoise with albedo + normal passes; kinematic playback of the exact MuJoCo replay poses (MuJoCo's own compiled meshes, primitives and colours); daylight: a sun lamp (2 degree disc) and a Nishita sky (no sun disc), the venue's own ground colour on a large ground plane with a faint world-space variation; AgX view transform, Medium High Contrast look; a tracking camera (fixed 22 degree elevation, per-frame fit of the robot and the object in play, smoothed in time) with a slow eased 10-degree orbit",
+ "rendering": "path-tracing",
+ "width": 2560,
+ "height": 1440,
+ "fps": 30,
+ "frames": 245,
+ "seconds": 8.17,
+ "speed": 0.9,
+ "success_check": true,
+ "success_check_how": "the task's own verdict (rcb_rfo.py Episode.verdict(), the referee, from the task dir's tests/ (privileged verifier copy)) at the end of this re-render's own replay of every recorded row from the task's frozen instance (t = 0 state hash checked); a second plain replay with the task's own replay() agrees and ends in the same final state hash (77de17d2e73b169a)",
+ "end_check": {
+  "success": true,
+  "stages": [
+   "touch 1",
+   "touch 2",
+   "touch 3",
+   "touch 4",
+   "touch 5",
+   "touch 6",
+   "touch 7",
+   "touch 8",
+   "touch 9",
+   "touch 10"
+  ],
+  "stages_total": 10,
+  "progress": 1.0,
+  "metrics": {
+   "touches": 10,
+   "bounces": 0,
+   "ball_height_m": 0.509
+  },
+  "time_s": 5.52
+ },
+ "recorded_result": {
+  "verifier_reward_json": {
+   "success": 1,
+   "n_actions": 552,
+   "progress": 1.0,
+   "stages_reached": 10,
+   "stages_total": 10,
+   "metric_ball_height_m": 0.509,
+   "metric_bounces": 0.0,
+   "metric_touches": 10.0
+  }
+ },
+ "replay": {
+  "task_dir": "roboolympics-fun-juggling-feet-i00-privileged",
+  "trajectory": "jobs/codex/gpt6_astra/high/codex-gpt6_astra-high-roboolympics-fun-juggling-feet-i00-privileged/roboolympics-juggling-feet-i00-p__p6ohQgr/artifacts/app/output/trajectory.npz",
+  "n_actions": 552,
+  "control_steps_replayed": 552,
+  "control_hz": 100.0,
+  "every_control_steps": 3,
+  "final_state_sha": "77de17d2e73b169ad6ef8ead57ad405b483971baac10f0d5352fdef7f00bf782",
+  "replays_agree": true,
+  "mujoco": "3.3.7",
+  "original_verdict": "counted success (verifier reward 1)"
+ },
+ "visual_changes": [
+  "daylight sky and a plain ground plane in the venue's ground colour instead of the venue's painted sky, hills, trees, grandstands and perimeter boards (drawn-only scenery)"
+ ],
+ "camera": {
+  "mode": "director",
+  "az_deg": -50.0,
+  "az_rule": "facing-50",
+  "el_deg": 22.0,
+  "fovy_deg": 34.0,
+  "fill": 0.72,
+  "distance_m": [
+   2.993,
+   3.327
+  ],
+  "robot_root": "pelvis",
+  "travel_m": 0.158,
+  "focus": {
+   "robot+object": 166,
+   "robot": 19
+  },
+  "objects": {
+   "ball": 0.53
+  }
+ },
+ "orbit_deg": 10.0,
+ "holds": {
+  "start_s": 0.5,
+  "end_s": 1.5,
+  "note": "first/last frame held, no extra simulation"
+ },
+ "seconds_per_frame": 2.97,
+ "image": "ghcr.io/mll-lab-nu/efe:robofunolympics-0.2.1 (simulator) + ghcr.io/mll-lab-nu/efe:roboolympics-fun-files-0.1.0@sha256:49a33342a72b57b56742cc754e427205a30f39dfcd69a99d1d7275c0eae4a619 (rcb_rfo.py, rfo/)",
+ "task_dir": "roboolympics-fun-juggling-feet-i00-privileged",
+ "agent": "codex + GPT-6 Astra (high)",
+ "run": "codex-gpt6_astra-high",
+ "mode": "privileged",
+ "trial": "codex-gpt6_astra-high-roboolympics-fun-juggling-feet-i00-privileged/roboolympics-juggling-feet-i00-p__p6ohQgr",
+ "trial_choice": "a counted GPT-6 Astra success in privileged mode",
+ "poster": "last frame (solved state)",
+ "host": "vast.ai RTX 4090 gpu4",
+ "encode": "libx264 preset slow CRF 16, yuv420p, bt709, faststart",
+ "created": "2026-10-19T06:20:23-0500"
+}