| { |
| "schema": "worldtoken_evaluation_protocol_catalog_v1", |
| "meaning": "Protocol declarations from the current public source/paper snapshot. They must not be represented as recovered historical evaluation configurations.", |
| "source_files": [ |
| { |
| "scope": "code", |
| "path": "environments/README.md" |
| }, |
| { |
| "scope": "code", |
| "path": "experiments/04_robocasa_scaling/baseline/evaluate.py" |
| }, |
| { |
| "scope": "code", |
| "path": "experiments/04_robocasa_scaling/configs/bc_transformer_d300_seed123.json" |
| }, |
| { |
| "scope": "code", |
| "path": "experiments/07_long_history_rmbench/configs/rmbench_ranking_modified_n2_c608_seed1.json" |
| }, |
| { |
| "scope": "code", |
| "path": "experiments/07_long_history_rmbench/evaluate.py" |
| }, |
| { |
| "scope": "code", |
| "path": "experiments/07_long_history_rmbench/seed_selection.json" |
| }, |
| { |
| "scope": "code", |
| "path": "experiments/common/robocasa_horizons.json" |
| }, |
| { |
| "scope": "code", |
| "path": "experiments/common/rollout.py" |
| }, |
| { |
| "scope": "code", |
| "path": "experiments/rmbench_tools/rmbench_long_horizon_common.py" |
| }, |
| { |
| "scope": "code", |
| "path": "experiments/rmbench_tools/rmbench_long_horizon_worker.py" |
| }, |
| { |
| "scope": "code", |
| "path": "experiments/rmbench_tools/rmbench_rollout_worker.py" |
| }, |
| { |
| "scope": "code", |
| "path": "experiments/rmbench_tools/rmbench_seeded_rollout_worker.py" |
| }, |
| { |
| "scope": "code", |
| "path": "worldtoken/eval_rollout.py" |
| }, |
| { |
| "scope": "code", |
| "path": "worldtoken/model.py" |
| }, |
| { |
| "scope": "code", |
| "path": "worldtoken/rmbench_policy.py" |
| }, |
| { |
| "scope": "paper", |
| "path": "appendices/B_robocasa_setup.tex" |
| }, |
| { |
| "scope": "paper", |
| "path": "appendices/C_robocasa_scaling.tex" |
| }, |
| { |
| "scope": "paper", |
| "path": "appendices/F_rmbench_history.tex" |
| } |
| ], |
| "protocols": { |
| "robocasa_worldtoken": { |
| "description": "Declared WorldToken RoboCasa closed-loop protocol; context and checkpoint vary by recipe/execution.", |
| "applicability": { |
| "dataset_sections": [ |
| "04_robocasa_scaling", |
| "05_token_interface", |
| "06_recent_history" |
| ], |
| "excludes_baseline": true |
| }, |
| "settings": { |
| "task_count": { |
| "value": 23, |
| "status": "protocol_declared", |
| "sources": [ |
| { |
| "scope": "code", |
| "path": "experiments/common/robocasa_horizons.json", |
| "pointer": "" |
| }, |
| { |
| "scope": "paper", |
| "path": "appendices/B_robocasa_setup.tex", |
| "line": 115 |
| } |
| ] |
| }, |
| "task_names": { |
| "value": [ |
| "CloseDoubleDoor", |
| "CloseDrawer", |
| "CloseSingleDoor", |
| "CoffeePressButton", |
| "CoffeeServeMug", |
| "CoffeeSetupMug", |
| "OpenDrawer", |
| "OpenSingleDoor", |
| "PnPCabToCounter", |
| "PnPCounterToCab", |
| "PnPCounterToMicrowave", |
| "PnPCounterToSink", |
| "PnPCounterToStove", |
| "PnPMicrowaveToCounter", |
| "PnPSinkToCounter", |
| "PnPStoveToCounter", |
| "TurnOffMicrowave", |
| "TurnOffSinkFaucet", |
| "TurnOffStove", |
| "TurnOnMicrowave", |
| "TurnOnSinkFaucet", |
| "TurnOnStove", |
| "TurnSinkSpout" |
| ], |
| "status": "protocol_declared", |
| "sources": [ |
| { |
| "scope": "code", |
| "path": "experiments/common/robocasa_horizons.json", |
| "pointer": "" |
| } |
| ] |
| }, |
| "episodes_per_task": { |
| "value": 50, |
| "status": "protocol_declared", |
| "sources": [ |
| { |
| "scope": "code", |
| "path": "experiments/common/rollout.py", |
| "line": 54, |
| "symbol": "main" |
| }, |
| { |
| "scope": "paper", |
| "path": "appendices/B_robocasa_setup.tex", |
| "line": 115 |
| } |
| ] |
| }, |
| "expected_episodes": { |
| "value": 1150, |
| "status": "protocol_declared", |
| "sources": [ |
| { |
| "scope": "paper", |
| "path": "appendices/B_robocasa_setup.tex", |
| "line": 115 |
| } |
| ] |
| }, |
| "rollout_seed": { |
| "value": 1, |
| "status": "protocol_declared", |
| "sources": [ |
| { |
| "scope": "code", |
| "path": "experiments/common/rollout.py", |
| "line": 54, |
| "symbol": "main" |
| } |
| ] |
| }, |
| "episode_seed_scheme": { |
| "value": "rollout_seed*100000 + task_idx*1000 + episode_idx", |
| "status": "protocol_declared", |
| "sources": [ |
| { |
| "scope": "code", |
| "path": "worldtoken/eval_rollout.py", |
| "line": 74, |
| "symbol": "EPISODE_SEED_SCHEME" |
| } |
| ] |
| }, |
| "task_horizons": { |
| "value": { |
| "CloseDoubleDoor": 700, |
| "CloseDrawer": 500, |
| "CloseSingleDoor": 500, |
| "CoffeePressButton": 300, |
| "CoffeeServeMug": 600, |
| "CoffeeSetupMug": 600, |
| "OpenDrawer": 500, |
| "OpenSingleDoor": 500, |
| "PnPCabToCounter": 500, |
| "PnPCounterToCab": 500, |
| "PnPCounterToMicrowave": 600, |
| "PnPCounterToSink": 700, |
| "PnPCounterToStove": 500, |
| "PnPMicrowaveToCounter": 500, |
| "PnPSinkToCounter": 500, |
| "PnPStoveToCounter": 500, |
| "TurnOffMicrowave": 500, |
| "TurnOffSinkFaucet": 500, |
| "TurnOffStove": 500, |
| "TurnOnMicrowave": 500, |
| "TurnOnSinkFaucet": 500, |
| "TurnOnStove": 500, |
| "TurnSinkSpout": 500 |
| }, |
| "status": "protocol_declared", |
| "sources": [ |
| { |
| "scope": "code", |
| "path": "experiments/common/robocasa_horizons.json", |
| "pointer": "" |
| }, |
| { |
| "scope": "paper", |
| "path": "appendices/B_robocasa_setup.tex", |
| "line": 134 |
| } |
| ] |
| }, |
| "action_clip": { |
| "value": "env", |
| "status": "protocol_declared", |
| "sources": [ |
| { |
| "scope": "code", |
| "path": "experiments/common/rollout.py", |
| "line": 57, |
| "symbol": "main" |
| } |
| ] |
| }, |
| "action_bound_margin": { |
| "value": 0.0001, |
| "status": "protocol_declared", |
| "sources": [ |
| { |
| "scope": "code", |
| "path": "experiments/common/rollout.py", |
| "line": 57, |
| "symbol": "main" |
| } |
| ] |
| }, |
| "action_scale": { |
| "value": 1.0, |
| "status": "protocol_declared", |
| "sources": [ |
| { |
| "scope": "code", |
| "path": "experiments/common/rollout.py", |
| "line": 57, |
| "symbol": "main" |
| } |
| ] |
| }, |
| "scene_protocol": { |
| "value": { |
| "generative_textures": null, |
| "scene_split": null, |
| "style_ids": null, |
| "layout_ids": null, |
| "layout_and_style_ids": [ |
| [ |
| 1, |
| 1 |
| ], |
| [ |
| 2, |
| 2 |
| ], |
| [ |
| 4, |
| 4 |
| ], |
| [ |
| 6, |
| 9 |
| ], |
| [ |
| 7, |
| 10 |
| ] |
| ], |
| "randomize_cameras": false, |
| "obj_instance_split": "B" |
| }, |
| "status": "protocol_declared", |
| "sources": [ |
| { |
| "scope": "code", |
| "path": "worldtoken/eval_rollout.py", |
| "line": 62, |
| "symbol": "BC_EVAL_ENV_KWARGS" |
| } |
| ] |
| }, |
| "terminate_on_success": { |
| "value": true, |
| "status": "protocol_declared", |
| "sources": [ |
| { |
| "scope": "code", |
| "path": "worldtoken/eval_rollout.py", |
| "line": 226, |
| "symbol": "parse_args" |
| } |
| ] |
| }, |
| "success_predicate": { |
| "value": "official task success (info.is_success.task), accumulated over trajectory", |
| "status": "protocol_declared", |
| "sources": [ |
| { |
| "scope": "code", |
| "path": "worldtoken/eval_rollout.py", |
| "line": 1617, |
| "symbol": "run_episode" |
| } |
| ] |
| }, |
| "environment_control_hz": { |
| "value": 20, |
| "status": "protocol_declared", |
| "sources": [ |
| { |
| "scope": "paper", |
| "path": "appendices/B_robocasa_setup.tex", |
| "line": 71 |
| } |
| ] |
| }, |
| "runtime": { |
| "value": { |
| "robosuite_version": "1.5.0", |
| "robosuite_commit": "dc7fcf9fa6cd" |
| }, |
| "status": "protocol_declared", |
| "sources": [ |
| { |
| "scope": "paper", |
| "path": "appendices/B_robocasa_setup.tex", |
| "line": 134 |
| } |
| ], |
| "note": "This is the paper's reporting-runtime declaration, not per-execution historical proof. The commit is the abbreviated identifier printed in the paper. Known cross-version repair episodes must be retained as execution exceptions. No historical RoboCasa Git hash is inferred from the compatible public v0.2 installation pin." |
| }, |
| "observation_stride": { |
| "value": 4, |
| "status": "protocol_declared", |
| "sources": [ |
| { |
| "scope": "code", |
| "path": "experiments/common/rollout.py", |
| "line": 66, |
| "symbol": "main" |
| }, |
| { |
| "scope": "paper", |
| "path": "appendices/B_robocasa_setup.tex", |
| "line": 71 |
| } |
| ] |
| }, |
| "execute_horizon": { |
| "value": 4, |
| "status": "protocol_declared", |
| "sources": [ |
| { |
| "scope": "code", |
| "path": "experiments/common/rollout.py", |
| "line": 66, |
| "symbol": "main" |
| } |
| ] |
| }, |
| "action_chunk_len": { |
| "value": 10, |
| "status": "protocol_declared", |
| "sources": [ |
| { |
| "scope": "paper", |
| "path": "appendices/B_robocasa_setup.tex", |
| "line": 72 |
| } |
| ] |
| }, |
| "sampler": { |
| "value": "stochastic_ddpm", |
| "status": "protocol_declared", |
| "sources": [ |
| { |
| "scope": "code", |
| "path": "experiments/common/rollout.py", |
| "line": 66, |
| "symbol": "main" |
| }, |
| { |
| "scope": "paper", |
| "path": "appendices/B_robocasa_setup.tex", |
| "line": 73 |
| } |
| ] |
| }, |
| "deterministic": { |
| "value": false, |
| "status": "protocol_declared", |
| "sources": [ |
| { |
| "scope": "code", |
| "path": "experiments/common/rollout.py", |
| "line": 66, |
| "symbol": "main" |
| } |
| ] |
| }, |
| "denoising_steps": { |
| "value": 20, |
| "status": "protocol_declared", |
| "sources": [ |
| { |
| "scope": "paper", |
| "path": "appendices/B_robocasa_setup.tex", |
| "line": 73 |
| } |
| ] |
| }, |
| "action_mean_samples": { |
| "value": 1, |
| "status": "protocol_declared", |
| "sources": [ |
| { |
| "scope": "code", |
| "path": "worldtoken/eval_rollout.py", |
| "line": 121, |
| "symbol": "parse_args" |
| } |
| ], |
| "note": "The public launcher does not override the evaluator's default of one sample." |
| }, |
| "warmup_padding": { |
| "value": { |
| "mode": "repeat_first_observation", |
| "length": "C_test" |
| }, |
| "status": "protocol_declared", |
| "sources": [ |
| { |
| "scope": "code", |
| "path": "experiments/common/rollout.py", |
| "line": 65, |
| "symbol": "main" |
| }, |
| { |
| "scope": "paper", |
| "path": "appendices/B_robocasa_setup.tex", |
| "line": 115 |
| } |
| ] |
| }, |
| "history_choices": { |
| "value": [ |
| 1, |
| 2, |
| 5, |
| 10 |
| ], |
| "status": "protocol_declared", |
| "sources": [ |
| { |
| "scope": "code", |
| "path": "experiments/common/rollout.py", |
| "line": 37, |
| "symbol": "main" |
| } |
| ], |
| "note": "Scaling checkpoints admit post-hoc C_test=1,2,5,10; token and separately trained context variants use their recipe context." |
| }, |
| "repeat_semantics": { |
| "value": "same checkpoint, episode identities, environment seeds, rollout seed, and protocol; fixed-seed repeatability", |
| "status": "protocol_declared", |
| "sources": [ |
| { |
| "scope": "paper", |
| "path": "appendices/B_robocasa_setup.tex", |
| "line": 118 |
| } |
| ] |
| }, |
| "holdout_eval_seed_semantics": { |
| "value": "config.eval_seed=0 selects offline holdout RMSE crops; it is not the closed-loop rollout_seed", |
| "status": "protocol_declared", |
| "sources": [ |
| { |
| "scope": "paper", |
| "path": "appendices/B_robocasa_setup.tex", |
| "line": 124 |
| } |
| ] |
| } |
| } |
| }, |
| "robocasa_bc_transformer": { |
| "description": "Native BC-Transformer/GMM RoboCasa comparison with one action and observation per replanning step.", |
| "applicability": { |
| "dataset_section": "04_robocasa_scaling", |
| "run": "bc_transformer_d300_seed123" |
| }, |
| "settings": { |
| "task_count": { |
| "value": 23, |
| "status": "protocol_declared", |
| "sources": [ |
| { |
| "scope": "code", |
| "path": "experiments/common/robocasa_horizons.json", |
| "pointer": "" |
| }, |
| { |
| "scope": "paper", |
| "path": "appendices/B_robocasa_setup.tex", |
| "line": 115 |
| } |
| ] |
| }, |
| "task_names": { |
| "value": [ |
| "CloseDoubleDoor", |
| "CloseDrawer", |
| "CloseSingleDoor", |
| "CoffeePressButton", |
| "CoffeeServeMug", |
| "CoffeeSetupMug", |
| "OpenDrawer", |
| "OpenSingleDoor", |
| "PnPCabToCounter", |
| "PnPCounterToCab", |
| "PnPCounterToMicrowave", |
| "PnPCounterToSink", |
| "PnPCounterToStove", |
| "PnPMicrowaveToCounter", |
| "PnPSinkToCounter", |
| "PnPStoveToCounter", |
| "TurnOffMicrowave", |
| "TurnOffSinkFaucet", |
| "TurnOffStove", |
| "TurnOnMicrowave", |
| "TurnOnSinkFaucet", |
| "TurnOnStove", |
| "TurnSinkSpout" |
| ], |
| "status": "protocol_declared", |
| "sources": [ |
| { |
| "scope": "code", |
| "path": "experiments/common/robocasa_horizons.json", |
| "pointer": "" |
| } |
| ] |
| }, |
| "episodes_per_task": { |
| "value": 50, |
| "status": "protocol_declared", |
| "sources": [ |
| { |
| "scope": "code", |
| "path": "experiments/common/rollout.py", |
| "line": 54, |
| "symbol": "main" |
| }, |
| { |
| "scope": "paper", |
| "path": "appendices/B_robocasa_setup.tex", |
| "line": 115 |
| } |
| ] |
| }, |
| "expected_episodes": { |
| "value": 1150, |
| "status": "protocol_declared", |
| "sources": [ |
| { |
| "scope": "paper", |
| "path": "appendices/B_robocasa_setup.tex", |
| "line": 115 |
| } |
| ] |
| }, |
| "rollout_seed": { |
| "value": 1, |
| "status": "protocol_declared", |
| "sources": [ |
| { |
| "scope": "code", |
| "path": "experiments/common/rollout.py", |
| "line": 54, |
| "symbol": "main" |
| } |
| ] |
| }, |
| "episode_seed_scheme": { |
| "value": "rollout_seed*100000 + task_idx*1000 + episode_idx", |
| "status": "protocol_declared", |
| "sources": [ |
| { |
| "scope": "code", |
| "path": "worldtoken/eval_rollout.py", |
| "line": 74, |
| "symbol": "EPISODE_SEED_SCHEME" |
| } |
| ] |
| }, |
| "task_horizons": { |
| "value": { |
| "CloseDoubleDoor": 700, |
| "CloseDrawer": 500, |
| "CloseSingleDoor": 500, |
| "CoffeePressButton": 300, |
| "CoffeeServeMug": 600, |
| "CoffeeSetupMug": 600, |
| "OpenDrawer": 500, |
| "OpenSingleDoor": 500, |
| "PnPCabToCounter": 500, |
| "PnPCounterToCab": 500, |
| "PnPCounterToMicrowave": 600, |
| "PnPCounterToSink": 700, |
| "PnPCounterToStove": 500, |
| "PnPMicrowaveToCounter": 500, |
| "PnPSinkToCounter": 500, |
| "PnPStoveToCounter": 500, |
| "TurnOffMicrowave": 500, |
| "TurnOffSinkFaucet": 500, |
| "TurnOffStove": 500, |
| "TurnOnMicrowave": 500, |
| "TurnOnSinkFaucet": 500, |
| "TurnOnStove": 500, |
| "TurnSinkSpout": 500 |
| }, |
| "status": "protocol_declared", |
| "sources": [ |
| { |
| "scope": "code", |
| "path": "experiments/common/robocasa_horizons.json", |
| "pointer": "" |
| }, |
| { |
| "scope": "paper", |
| "path": "appendices/B_robocasa_setup.tex", |
| "line": 134 |
| } |
| ] |
| }, |
| "action_clip": { |
| "value": "env", |
| "status": "protocol_declared", |
| "sources": [ |
| { |
| "scope": "code", |
| "path": "experiments/common/rollout.py", |
| "line": 57, |
| "symbol": "main" |
| } |
| ] |
| }, |
| "action_bound_margin": { |
| "value": 0.0001, |
| "status": "protocol_declared", |
| "sources": [ |
| { |
| "scope": "code", |
| "path": "experiments/common/rollout.py", |
| "line": 57, |
| "symbol": "main" |
| } |
| ] |
| }, |
| "action_scale": { |
| "value": 1.0, |
| "status": "protocol_declared", |
| "sources": [ |
| { |
| "scope": "code", |
| "path": "experiments/common/rollout.py", |
| "line": 57, |
| "symbol": "main" |
| } |
| ] |
| }, |
| "scene_protocol": { |
| "value": { |
| "generative_textures": null, |
| "scene_split": null, |
| "style_ids": null, |
| "layout_ids": null, |
| "layout_and_style_ids": [ |
| [ |
| 1, |
| 1 |
| ], |
| [ |
| 2, |
| 2 |
| ], |
| [ |
| 4, |
| 4 |
| ], |
| [ |
| 6, |
| 9 |
| ], |
| [ |
| 7, |
| 10 |
| ] |
| ], |
| "randomize_cameras": false, |
| "obj_instance_split": "B" |
| }, |
| "status": "protocol_declared", |
| "sources": [ |
| { |
| "scope": "code", |
| "path": "worldtoken/eval_rollout.py", |
| "line": 62, |
| "symbol": "BC_EVAL_ENV_KWARGS" |
| } |
| ] |
| }, |
| "terminate_on_success": { |
| "value": true, |
| "status": "protocol_declared", |
| "sources": [ |
| { |
| "scope": "code", |
| "path": "experiments/04_robocasa_scaling/baseline/evaluate.py", |
| "line": 69, |
| "symbol": "parse_args" |
| } |
| ] |
| }, |
| "success_predicate": { |
| "value": "official task success (info.is_success.task), accumulated over trajectory", |
| "status": "protocol_declared", |
| "sources": [ |
| { |
| "scope": "code", |
| "path": "worldtoken/eval_rollout.py", |
| "line": 1617, |
| "symbol": "run_episode" |
| } |
| ] |
| }, |
| "environment_control_hz": { |
| "value": 20, |
| "status": "protocol_declared", |
| "sources": [ |
| { |
| "scope": "paper", |
| "path": "appendices/B_robocasa_setup.tex", |
| "line": 71 |
| } |
| ] |
| }, |
| "runtime": { |
| "value": { |
| "robosuite_version": "1.5.0", |
| "robosuite_commit": "dc7fcf9fa6cd" |
| }, |
| "status": "protocol_declared", |
| "sources": [ |
| { |
| "scope": "paper", |
| "path": "appendices/B_robocasa_setup.tex", |
| "line": 134 |
| } |
| ], |
| "note": "This is the paper's reporting-runtime declaration, not per-execution historical proof. The commit is the abbreviated identifier printed in the paper. Known cross-version repair episodes must be retained as execution exceptions. No historical RoboCasa Git hash is inferred from the compatible public v0.2 installation pin." |
| }, |
| "observation_stride": { |
| "value": 1, |
| "status": "protocol_declared", |
| "sources": [ |
| { |
| "scope": "code", |
| "path": "experiments/04_robocasa_scaling/baseline/evaluate.py", |
| "line": 482, |
| "symbol": "main" |
| }, |
| { |
| "scope": "paper", |
| "path": "appendices/C_robocasa_scaling.tex", |
| "line": 29 |
| } |
| ] |
| }, |
| "execute_horizon": { |
| "value": 1, |
| "status": "protocol_declared", |
| "sources": [ |
| { |
| "scope": "code", |
| "path": "experiments/04_robocasa_scaling/baseline/evaluate.py", |
| "line": 343, |
| "symbol": "resolve_policy_protocol" |
| } |
| ] |
| }, |
| "sampler": { |
| "value": "robomimic_bc_transformer_gmm", |
| "status": "protocol_declared", |
| "sources": [ |
| { |
| "scope": "code", |
| "path": "experiments/04_robocasa_scaling/baseline/evaluate.py", |
| "line": 342, |
| "symbol": "resolve_policy_protocol" |
| } |
| ] |
| }, |
| "deterministic": { |
| "value": false, |
| "status": "protocol_declared", |
| "sources": [ |
| { |
| "scope": "code", |
| "path": "experiments/04_robocasa_scaling/baseline/evaluate.py", |
| "line": 351, |
| "symbol": "resolve_policy_protocol" |
| } |
| ], |
| "note": "Native GMM policy sampling; low_noise_eval does not mean a deterministic action." |
| }, |
| "action_mean_samples": { |
| "value": 1, |
| "status": "protocol_declared", |
| "sources": [ |
| { |
| "scope": "code", |
| "path": "experiments/04_robocasa_scaling/baseline/evaluate.py", |
| "line": 105, |
| "symbol": "RobomimicPolicyAdapter.__init__" |
| } |
| ] |
| }, |
| "gmm_num_modes": { |
| "value": 5, |
| "status": "protocol_declared", |
| "sources": [ |
| { |
| "scope": "code", |
| "path": "experiments/04_robocasa_scaling/configs/bc_transformer_d300_seed123.json", |
| "pointer": "/algo/gmm/num_modes" |
| } |
| ] |
| }, |
| "gmm_low_noise_eval": { |
| "value": true, |
| "status": "protocol_declared", |
| "sources": [ |
| { |
| "scope": "code", |
| "path": "experiments/04_robocasa_scaling/configs/bc_transformer_d300_seed123.json", |
| "pointer": "/algo/gmm/low_noise_eval" |
| } |
| ] |
| }, |
| "frame_stack": { |
| "value": 10, |
| "status": "protocol_declared", |
| "sources": [ |
| { |
| "scope": "code", |
| "path": "experiments/04_robocasa_scaling/configs/bc_transformer_d300_seed123.json", |
| "pointer": "/train/frame_stack" |
| } |
| ] |
| }, |
| "warmup_padding": { |
| "value": { |
| "mode": "repeat_first_observation", |
| "length": 10 |
| }, |
| "status": "protocol_declared", |
| "sources": [ |
| { |
| "scope": "code", |
| "path": "experiments/04_robocasa_scaling/baseline/evaluate.py", |
| "line": 225, |
| "symbol": "RobomimicPolicyAdapter._stack_observation" |
| } |
| ] |
| }, |
| "policy_behavior": { |
| "value": "official_native", |
| "status": "protocol_declared", |
| "sources": [ |
| { |
| "scope": "code", |
| "path": "experiments/04_robocasa_scaling/baseline/evaluate.py", |
| "line": 477, |
| "symbol": "main" |
| } |
| ] |
| } |
| } |
| }, |
| "rmbench_formal": { |
| "description": "Formal 100-episode Blocks Ranking history interventions; official success termination and 3500-action horizon.", |
| "applicability": { |
| "dataset_section": "07_long_history_rmbench", |
| "rollout_directory_pattern": "C*_100eps" |
| }, |
| "settings": { |
| "task_count": { |
| "value": 1, |
| "status": "protocol_declared", |
| "sources": [ |
| { |
| "scope": "code", |
| "path": "experiments/07_long_history_rmbench/evaluate.py", |
| "line": 17, |
| "symbol": "TASK" |
| } |
| ] |
| }, |
| "task_names": { |
| "value": [ |
| "blocks_ranking_try" |
| ], |
| "status": "protocol_declared", |
| "sources": [ |
| { |
| "scope": "code", |
| "path": "experiments/07_long_history_rmbench/evaluate.py", |
| "line": 17, |
| "symbol": "TASK" |
| } |
| ] |
| }, |
| "task_config": { |
| "value": "demo_clean", |
| "status": "protocol_declared", |
| "sources": [ |
| { |
| "scope": "code", |
| "path": "experiments/rmbench_tools/rmbench_rollout_worker.py", |
| "line": 101, |
| "symbol": "load_environment_args" |
| }, |
| { |
| "scope": "paper", |
| "path": "appendices/F_rmbench_history.tex", |
| "line": 54 |
| } |
| ] |
| }, |
| "initial_state_selection": { |
| "value": "prevalidated candidate seeds; expert planning and official success checks passed", |
| "status": "protocol_declared", |
| "sources": [ |
| { |
| "scope": "code", |
| "path": "experiments/07_long_history_rmbench/seed_selection.json", |
| "pointer": "/tasks/blocks_ranking_try" |
| }, |
| { |
| "scope": "paper", |
| "path": "appendices/F_rmbench_history.tex", |
| "line": 54 |
| } |
| ] |
| }, |
| "episode_seed_scheme": { |
| "value": "candidate_seed from fixed seed selection; policy diffusion RNG uses the same candidate_seed", |
| "status": "protocol_declared", |
| "sources": [ |
| { |
| "scope": "code", |
| "path": "experiments/rmbench_tools/rmbench_seeded_rollout_worker.py", |
| "line": 391, |
| "symbol": "main" |
| } |
| ] |
| }, |
| "observation_stride": { |
| "value": 4, |
| "status": "protocol_declared", |
| "sources": [ |
| { |
| "scope": "code", |
| "path": "worldtoken/rmbench_policy.py", |
| "line": 120, |
| "symbol": "RMBenchRolloutPolicy.update_obs" |
| }, |
| { |
| "scope": "paper", |
| "path": "appendices/F_rmbench_history.tex", |
| "line": 154 |
| } |
| ] |
| }, |
| "execute_horizon": { |
| "value": 4, |
| "status": "protocol_declared", |
| "sources": [ |
| { |
| "scope": "code", |
| "path": "experiments/07_long_history_rmbench/evaluate.py", |
| "line": 36, |
| "symbol": "run_evaluation" |
| } |
| ] |
| }, |
| "action_chunk_len": { |
| "value": 8, |
| "status": "protocol_declared", |
| "sources": [ |
| { |
| "scope": "code", |
| "path": "experiments/07_long_history_rmbench/configs/rmbench_ranking_modified_n2_c608_seed1.json", |
| "pointer": "/model/action_chunk_len" |
| } |
| ] |
| }, |
| "action_dim": { |
| "value": 14, |
| "status": "protocol_declared", |
| "sources": [ |
| { |
| "scope": "paper", |
| "path": "appendices/F_rmbench_history.tex", |
| "line": 20 |
| } |
| ] |
| }, |
| "action_representation": { |
| "value": "absolute_joint_positions_qpos", |
| "status": "protocol_declared", |
| "sources": [ |
| { |
| "scope": "code", |
| "path": "experiments/rmbench_tools/rmbench_seeded_rollout_worker.py", |
| "line": 422, |
| "symbol": "main" |
| } |
| ] |
| }, |
| "sampler": { |
| "value": "stochastic_ddpm", |
| "status": "protocol_declared", |
| "sources": [ |
| { |
| "scope": "code", |
| "path": "experiments/07_long_history_rmbench/evaluate.py", |
| "line": 36, |
| "symbol": "run_evaluation" |
| }, |
| { |
| "scope": "paper", |
| "path": "appendices/F_rmbench_history.tex", |
| "line": 47 |
| } |
| ] |
| }, |
| "deterministic": { |
| "value": false, |
| "status": "protocol_declared", |
| "sources": [ |
| { |
| "scope": "code", |
| "path": "experiments/07_long_history_rmbench/evaluate.py", |
| "line": 36, |
| "symbol": "run_evaluation" |
| } |
| ] |
| }, |
| "denoising_steps": { |
| "value": 20, |
| "status": "protocol_declared", |
| "sources": [ |
| { |
| "scope": "code", |
| "path": "experiments/07_long_history_rmbench/configs/rmbench_ranking_modified_n2_c608_seed1.json", |
| "pointer": "/action_head/params/denoising_steps" |
| }, |
| { |
| "scope": "paper", |
| "path": "appendices/F_rmbench_history.tex", |
| "line": 46 |
| } |
| ] |
| }, |
| "action_mean_samples": { |
| "value": 1, |
| "status": "protocol_declared", |
| "sources": [ |
| { |
| "scope": "code", |
| "path": "worldtoken/model.py", |
| "line": 168, |
| "symbol": "RoboCasaDiffusionActionModel.sample_action_chunk" |
| }, |
| { |
| "scope": "code", |
| "path": "worldtoken/rmbench_policy.py", |
| "line": 184, |
| "symbol": "RMBenchRolloutPolicy.get_action" |
| } |
| ] |
| }, |
| "denoised_sample_clip": { |
| "value": [ |
| -1, |
| 1 |
| ], |
| "status": "protocol_declared", |
| "sources": [ |
| { |
| "scope": "paper", |
| "path": "appendices/F_rmbench_history.tex", |
| "line": 46 |
| } |
| ], |
| "note": "Clipping is applied in normalized DDPM reconstruction, before inherited action unnormalization; this is not RoboCasa environment-bound clipping." |
| }, |
| "warmup_padding": { |
| "value": { |
| "mode": "none", |
| "length": 0 |
| }, |
| "status": "protocol_declared", |
| "sources": [ |
| { |
| "scope": "code", |
| "path": "worldtoken/rmbench_policy.py", |
| "line": 142, |
| "symbol": "RMBenchRolloutPolicy._history_batch" |
| }, |
| { |
| "scope": "paper", |
| "path": "appendices/B_robocasa_setup.tex", |
| "line": 115 |
| } |
| ] |
| }, |
| "action_step_seconds": { |
| "value": 0.06, |
| "status": "protocol_declared", |
| "sources": [ |
| { |
| "scope": "paper", |
| "path": "appendices/F_rmbench_history.tex", |
| "line": 154 |
| } |
| ] |
| }, |
| "checkpoint_step": { |
| "value": 5500, |
| "status": "protocol_declared", |
| "sources": [ |
| { |
| "scope": "code", |
| "path": "experiments/07_long_history_rmbench/evaluate.py", |
| "line": 34, |
| "symbol": "run_evaluation" |
| } |
| ] |
| }, |
| "precision": { |
| "value": "bf16", |
| "status": "protocol_declared", |
| "sources": [ |
| { |
| "scope": "code", |
| "path": "experiments/07_long_history_rmbench/evaluate.py", |
| "line": 35, |
| "symbol": "run_evaluation" |
| } |
| ] |
| }, |
| "cache_encoded_history": { |
| "value": true, |
| "status": "protocol_declared", |
| "sources": [ |
| { |
| "scope": "code", |
| "path": "experiments/07_long_history_rmbench/evaluate.py", |
| "line": 36, |
| "symbol": "run_evaluation" |
| } |
| ] |
| }, |
| "runtime": { |
| "value": { |
| "rmbench_commit": "87e0498891073d483d330195c0f160709bd92ff5" |
| }, |
| "status": "protocol_declared", |
| "sources": [ |
| { |
| "scope": "code", |
| "path": "environments/README.md", |
| "line": 92 |
| } |
| ], |
| "note": "Public reproduction documentation declares this RMBench revision; per-execution runtime artifacts take precedence. This is not a complete historical simulator lockfile." |
| }, |
| "episodes_per_task": { |
| "value": 100, |
| "status": "protocol_declared", |
| "sources": [ |
| { |
| "scope": "code", |
| "path": "experiments/07_long_history_rmbench/seed_selection.json", |
| "pointer": "/tasks/blocks_ranking_try" |
| }, |
| { |
| "scope": "paper", |
| "path": "appendices/F_rmbench_history.tex", |
| "line": 150 |
| } |
| ] |
| }, |
| "expected_episodes": { |
| "value": 100, |
| "status": "protocol_declared", |
| "sources": [ |
| { |
| "scope": "code", |
| "path": "experiments/07_long_history_rmbench/seed_selection.json", |
| "pointer": "/tasks/blocks_ranking_try" |
| } |
| ] |
| }, |
| "seed_range": { |
| "value": [ |
| 100000, |
| 100099 |
| ], |
| "status": "protocol_declared", |
| "sources": [ |
| { |
| "scope": "code", |
| "path": "experiments/07_long_history_rmbench/seed_selection.json", |
| "pointer": "/tasks/blocks_ranking_try" |
| }, |
| { |
| "scope": "paper", |
| "path": "appendices/F_rmbench_history.tex", |
| "line": 54 |
| } |
| ] |
| }, |
| "history_choices": { |
| "value": [ |
| 608, |
| 288, |
| 128, |
| 64, |
| 32 |
| ], |
| "status": "protocol_declared", |
| "sources": [ |
| { |
| "scope": "code", |
| "path": "experiments/07_long_history_rmbench/evaluate.py", |
| "line": 127, |
| "symbol": "main" |
| } |
| ] |
| }, |
| "action_horizon": { |
| "value": 3500, |
| "status": "protocol_declared", |
| "sources": [ |
| { |
| "scope": "paper", |
| "path": "appendices/F_rmbench_history.tex", |
| "line": 150 |
| } |
| ] |
| }, |
| "terminate_on_success": { |
| "value": true, |
| "status": "protocol_declared", |
| "sources": [ |
| { |
| "scope": "code", |
| "path": "experiments/rmbench_tools/rmbench_seeded_rollout_worker.py", |
| "line": 394, |
| "symbol": "main" |
| } |
| ] |
| }, |
| "success_predicate": { |
| "value": "unmodified official evaluator success", |
| "status": "protocol_declared", |
| "sources": [ |
| { |
| "scope": "paper", |
| "path": "appendices/F_rmbench_history.tex", |
| "line": 150 |
| } |
| ] |
| }, |
| "early_failure_heuristic": { |
| "value": "off", |
| "status": "protocol_declared", |
| "sources": [ |
| { |
| "scope": "code", |
| "path": "experiments/07_long_history_rmbench/evaluate.py", |
| "line": 52, |
| "symbol": "run_evaluation" |
| } |
| ] |
| }, |
| "classification": { |
| "value": "formal_100_episode_history_intervention", |
| "status": "protocol_declared", |
| "sources": [ |
| { |
| "scope": "paper", |
| "path": "appendices/F_rmbench_history.tex", |
| "line": 154 |
| } |
| ] |
| } |
| } |
| }, |
| "rmbench_stress": { |
| "description": "Exploratory single-trajectory extended executions with success termination suppressed and event-based stopping.", |
| "applicability": { |
| "dataset_section": "07_long_history_rmbench", |
| "rollout_directory_pattern": "stress_seed*" |
| }, |
| "settings": { |
| "task_count": { |
| "value": 1, |
| "status": "protocol_declared", |
| "sources": [ |
| { |
| "scope": "code", |
| "path": "experiments/07_long_history_rmbench/evaluate.py", |
| "line": 17, |
| "symbol": "TASK" |
| } |
| ] |
| }, |
| "task_names": { |
| "value": [ |
| "blocks_ranking_try" |
| ], |
| "status": "protocol_declared", |
| "sources": [ |
| { |
| "scope": "code", |
| "path": "experiments/07_long_history_rmbench/evaluate.py", |
| "line": 17, |
| "symbol": "TASK" |
| } |
| ] |
| }, |
| "task_config": { |
| "value": "demo_clean", |
| "status": "protocol_declared", |
| "sources": [ |
| { |
| "scope": "code", |
| "path": "experiments/rmbench_tools/rmbench_rollout_worker.py", |
| "line": 101, |
| "symbol": "load_environment_args" |
| }, |
| { |
| "scope": "paper", |
| "path": "appendices/F_rmbench_history.tex", |
| "line": 54 |
| } |
| ] |
| }, |
| "initial_state_selection": { |
| "value": "prevalidated candidate seeds; expert planning and official success checks passed", |
| "status": "protocol_declared", |
| "sources": [ |
| { |
| "scope": "code", |
| "path": "experiments/07_long_history_rmbench/seed_selection.json", |
| "pointer": "/tasks/blocks_ranking_try" |
| }, |
| { |
| "scope": "paper", |
| "path": "appendices/F_rmbench_history.tex", |
| "line": 54 |
| } |
| ] |
| }, |
| "episode_seed_scheme": { |
| "value": "selected candidate_seed from fixed seed selection; policy diffusion RNG uses the same candidate_seed", |
| "status": "protocol_declared", |
| "sources": [ |
| { |
| "scope": "code", |
| "path": "experiments/rmbench_tools/rmbench_long_horizon_worker.py", |
| "line": 561, |
| "symbol": "main" |
| } |
| ] |
| }, |
| "observation_stride": { |
| "value": 4, |
| "status": "protocol_declared", |
| "sources": [ |
| { |
| "scope": "code", |
| "path": "worldtoken/rmbench_policy.py", |
| "line": 120, |
| "symbol": "RMBenchRolloutPolicy.update_obs" |
| }, |
| { |
| "scope": "paper", |
| "path": "appendices/F_rmbench_history.tex", |
| "line": 154 |
| } |
| ] |
| }, |
| "execute_horizon": { |
| "value": 4, |
| "status": "protocol_declared", |
| "sources": [ |
| { |
| "scope": "code", |
| "path": "experiments/07_long_history_rmbench/evaluate.py", |
| "line": 36, |
| "symbol": "run_evaluation" |
| } |
| ] |
| }, |
| "action_chunk_len": { |
| "value": 8, |
| "status": "protocol_declared", |
| "sources": [ |
| { |
| "scope": "code", |
| "path": "experiments/07_long_history_rmbench/configs/rmbench_ranking_modified_n2_c608_seed1.json", |
| "pointer": "/model/action_chunk_len" |
| } |
| ] |
| }, |
| "action_dim": { |
| "value": 14, |
| "status": "protocol_declared", |
| "sources": [ |
| { |
| "scope": "paper", |
| "path": "appendices/F_rmbench_history.tex", |
| "line": 20 |
| } |
| ] |
| }, |
| "action_representation": { |
| "value": "absolute_joint_positions_qpos", |
| "status": "protocol_declared", |
| "sources": [ |
| { |
| "scope": "code", |
| "path": "experiments/rmbench_tools/rmbench_seeded_rollout_worker.py", |
| "line": 422, |
| "symbol": "main" |
| } |
| ] |
| }, |
| "sampler": { |
| "value": "stochastic_ddpm", |
| "status": "protocol_declared", |
| "sources": [ |
| { |
| "scope": "code", |
| "path": "experiments/07_long_history_rmbench/evaluate.py", |
| "line": 36, |
| "symbol": "run_evaluation" |
| }, |
| { |
| "scope": "paper", |
| "path": "appendices/F_rmbench_history.tex", |
| "line": 47 |
| } |
| ] |
| }, |
| "deterministic": { |
| "value": false, |
| "status": "protocol_declared", |
| "sources": [ |
| { |
| "scope": "code", |
| "path": "experiments/07_long_history_rmbench/evaluate.py", |
| "line": 36, |
| "symbol": "run_evaluation" |
| } |
| ] |
| }, |
| "denoising_steps": { |
| "value": 20, |
| "status": "protocol_declared", |
| "sources": [ |
| { |
| "scope": "code", |
| "path": "experiments/07_long_history_rmbench/configs/rmbench_ranking_modified_n2_c608_seed1.json", |
| "pointer": "/action_head/params/denoising_steps" |
| }, |
| { |
| "scope": "paper", |
| "path": "appendices/F_rmbench_history.tex", |
| "line": 46 |
| } |
| ] |
| }, |
| "action_mean_samples": { |
| "value": 1, |
| "status": "protocol_declared", |
| "sources": [ |
| { |
| "scope": "code", |
| "path": "worldtoken/model.py", |
| "line": 168, |
| "symbol": "RoboCasaDiffusionActionModel.sample_action_chunk" |
| }, |
| { |
| "scope": "code", |
| "path": "worldtoken/rmbench_policy.py", |
| "line": 184, |
| "symbol": "RMBenchRolloutPolicy.get_action" |
| } |
| ] |
| }, |
| "denoised_sample_clip": { |
| "value": [ |
| -1, |
| 1 |
| ], |
| "status": "protocol_declared", |
| "sources": [ |
| { |
| "scope": "paper", |
| "path": "appendices/F_rmbench_history.tex", |
| "line": 46 |
| } |
| ], |
| "note": "Clipping is applied in normalized DDPM reconstruction, before inherited action unnormalization; this is not RoboCasa environment-bound clipping." |
| }, |
| "warmup_padding": { |
| "value": { |
| "mode": "none", |
| "length": 0 |
| }, |
| "status": "protocol_declared", |
| "sources": [ |
| { |
| "scope": "code", |
| "path": "worldtoken/rmbench_policy.py", |
| "line": 142, |
| "symbol": "RMBenchRolloutPolicy._history_batch" |
| }, |
| { |
| "scope": "paper", |
| "path": "appendices/B_robocasa_setup.tex", |
| "line": 115 |
| } |
| ] |
| }, |
| "action_step_seconds": { |
| "value": 0.06, |
| "status": "protocol_declared", |
| "sources": [ |
| { |
| "scope": "paper", |
| "path": "appendices/F_rmbench_history.tex", |
| "line": 154 |
| } |
| ] |
| }, |
| "checkpoint_step": { |
| "value": 5500, |
| "status": "protocol_declared", |
| "sources": [ |
| { |
| "scope": "code", |
| "path": "experiments/07_long_history_rmbench/evaluate.py", |
| "line": 34, |
| "symbol": "run_evaluation" |
| } |
| ] |
| }, |
| "precision": { |
| "value": "bf16", |
| "status": "protocol_declared", |
| "sources": [ |
| { |
| "scope": "code", |
| "path": "experiments/07_long_history_rmbench/evaluate.py", |
| "line": 35, |
| "symbol": "run_evaluation" |
| } |
| ] |
| }, |
| "cache_encoded_history": { |
| "value": true, |
| "status": "protocol_declared", |
| "sources": [ |
| { |
| "scope": "code", |
| "path": "experiments/07_long_history_rmbench/evaluate.py", |
| "line": 36, |
| "symbol": "run_evaluation" |
| } |
| ] |
| }, |
| "runtime": { |
| "value": { |
| "rmbench_commit": "87e0498891073d483d330195c0f160709bd92ff5" |
| }, |
| "status": "protocol_declared", |
| "sources": [ |
| { |
| "scope": "code", |
| "path": "environments/README.md", |
| "line": 92 |
| } |
| ], |
| "note": "Public reproduction documentation declares this RMBench revision; per-execution runtime artifacts take precedence. This is not a complete historical simulator lockfile." |
| }, |
| "episodes_per_task": { |
| "value": 1, |
| "status": "protocol_declared", |
| "sources": [ |
| { |
| "scope": "paper", |
| "path": "appendices/F_rmbench_history.tex", |
| "line": 260 |
| } |
| ], |
| "note": "One trajectory per stress_seed directory; nine trajectories in the declared campaign." |
| }, |
| "expected_episodes": { |
| "value": 1, |
| "status": "protocol_declared", |
| "sources": [ |
| { |
| "scope": "paper", |
| "path": "appendices/F_rmbench_history.tex", |
| "line": 260 |
| } |
| ] |
| }, |
| "campaign_episodes": { |
| "value": 9, |
| "status": "protocol_declared", |
| "sources": [ |
| { |
| "scope": "code", |
| "path": "experiments/07_long_history_rmbench/evaluate.py", |
| "line": 18, |
| "symbol": "STRESS_SEEDS" |
| } |
| ] |
| }, |
| "candidate_seeds": { |
| "value": [ |
| 100000, |
| 100001, |
| 100002, |
| 100003, |
| 100004, |
| 100007, |
| 100008, |
| 100015, |
| 100016 |
| ], |
| "status": "protocol_declared", |
| "sources": [ |
| { |
| "scope": "code", |
| "path": "experiments/07_long_history_rmbench/evaluate.py", |
| "line": 18, |
| "symbol": "STRESS_SEEDS" |
| } |
| ] |
| }, |
| "max_history": { |
| "value": 608, |
| "status": "protocol_declared", |
| "sources": [ |
| { |
| "scope": "code", |
| "path": "experiments/07_long_history_rmbench/evaluate.py", |
| "line": 125, |
| "symbol": "main" |
| } |
| ] |
| }, |
| "action_horizon": { |
| "value": null, |
| "status": "protocol_declared", |
| "sources": [ |
| { |
| "scope": "paper", |
| "path": "appendices/F_rmbench_history.tex", |
| "line": 260 |
| } |
| ], |
| "note": "The implementation has a 2,000,000,000-action guard. Scientific stopping is event based; operational pauses/crashes should be read from each run." |
| }, |
| "terminate_on_success": { |
| "value": false, |
| "status": "protocol_declared", |
| "sources": [ |
| { |
| "scope": "paper", |
| "path": "appendices/F_rmbench_history.tex", |
| "line": 260 |
| }, |
| { |
| "scope": "code", |
| "path": "experiments/rmbench_tools/rmbench_long_horizon_worker.py", |
| "line": 425, |
| "symbol": "main" |
| } |
| ] |
| }, |
| "success_predicate": { |
| "value": "official evaluator success logged as an event, not a stopping criterion", |
| "status": "protocol_declared", |
| "sources": [ |
| { |
| "scope": "paper", |
| "path": "appendices/F_rmbench_history.tex", |
| "line": 262 |
| } |
| ] |
| }, |
| "stable_confirmations": { |
| "value": 8, |
| "status": "protocol_declared", |
| "sources": [ |
| { |
| "scope": "code", |
| "path": "experiments/07_long_history_rmbench/evaluate.py", |
| "line": 41, |
| "symbol": "run_evaluation" |
| } |
| ] |
| }, |
| "stall_action_steps": { |
| "value": 1500, |
| "status": "protocol_declared", |
| "sources": [ |
| { |
| "scope": "code", |
| "path": "experiments/07_long_history_rmbench/evaluate.py", |
| "line": 41, |
| "symbol": "run_evaluation" |
| }, |
| { |
| "scope": "paper", |
| "path": "appendices/F_rmbench_history.tex", |
| "line": 264 |
| } |
| ] |
| }, |
| "periodic_swap_pattern": { |
| "value": [ |
| "middle_right", |
| "left_right", |
| "left_middle" |
| ], |
| "status": "protocol_declared", |
| "sources": [ |
| { |
| "scope": "code", |
| "path": "experiments/rmbench_tools/rmbench_long_horizon_common.py", |
| "line": 30, |
| "symbol": "PERIODIC_SWAP_NAMES" |
| } |
| ] |
| }, |
| "stop_conditions": { |
| "value": [ |
| "first confirmed stable-order deviation", |
| "1500 actions without a new confirmed reference transition" |
| ], |
| "status": "protocol_declared", |
| "sources": [ |
| { |
| "scope": "paper", |
| "path": "appendices/F_rmbench_history.tex", |
| "line": 264 |
| } |
| ] |
| }, |
| "classification": { |
| "value": "exploratory_extended_execution", |
| "status": "protocol_declared", |
| "sources": [ |
| { |
| "scope": "paper", |
| "path": "appendices/F_rmbench_history.tex", |
| "line": 260 |
| } |
| ] |
| } |
| } |
| } |
| }, |
| "schema_version": 1, |
| "aggregation_kind": "release_time_catalogue", |
| "status_definitions": { |
| "recorded": "Extracted from retained record fields; grouped values cite their row sources.", |
| "derived": "Computed from retained records or directory naming; derivation is described.", |
| "protocol_declared": "Declared by cited paper/code, not an original per-execution snapshot.", |
| "unknown": "Not established from retained sources; null/empty is not a proven negative." |
| }, |
| "index_summary": { |
| "evaluations": 365, |
| "sections": { |
| "04_robocasa_scaling": 303, |
| "05_token_interface": 30, |
| "06_recent_history": 18, |
| "07_long_history_rmbench": 14 |
| }, |
| "robocasa_episode_executions": 403650 |
| }, |
| "index_notes": [ |
| "This is a release-time aggregation, not a replacement for historical configs.", |
| "Sources use records/code/paper root-relative scopes; do not resolve paths against a different scope.", |
| "Training/holdout settings remain separate from closed-loop rollout settings.", |
| "Runtime exceptions copy the affected evaluation summary's original IDs and note." |
| ], |
| "index_resolution": "Combine protocol.settings with index.settings; index values take precedence. Unresolved historical runtime remains a declaration in the profile, not a recorded execution fact." |
| } |
|
|
