Index all Figure 2 curves, exact evaluations and unresolved checkpoint bindings

README.md

@@ -3913,3 +3913,7 @@ Training data, episode-level evaluations and paper experiment manifests.
 ## Artifact retention
 
 HAIC and Extreme Parkour releases are retired. Model bundles retain the published evaluation checkpoint, or the latest checkpoint when no evaluation selection exists. Optimizer and trainer recovery state are not release assets. Identical dataset copies use the canonical task paths. Original configurations and experiment evidence remain source records.
+
+## Figure 2 latency degradation
+
+[Latency-degradation catalogue](experiments/latency-degradation/README.md) preserves all 15 curves, exact evaluation records, fixed checkpoint sources and provenance gaps. Six MIKASA checkpoints are already central; copying six original game weights to benchmark-models remains pending.

experiments/README.md

@@ -13,3 +13,5 @@
 Each task page provides settings, fixed asset references and evaluation evidence. Shared assets have one current location.
 
 [Sim2Real calibration: 30 cases](sim2real-calibration/README.md) · [ICLR two-GPU archive](sim2real-calibration/archive/iclr-submission-two-gpu/README.md)
+
+- [Latency degradation / Figure 2](latency-degradation/README.md): 15 curves, original evaluations, fixed model sources and explicit missing bindings.

experiments/latency-degradation/README.md

@@ -0,0 +1,33 @@
+# Figure 2: latency degradation
+
+Figure 2 contains nine game curves and six MIKASA curves. Its CSV stores 95 rows; the main panels plot 72 unique points. Insets repeat selected points.
+
+The [paper input snapshot](https://huggingface.co/datasets/latency-sensitive-bench/benchmark-datasets/tree/efeac490144ca9c94660e2410163415e0a8809a9/experiments/latency-degradation/paper) preserves the builder, plotter, CSV, raw JSONs and original rendered figure. The numerical figure inputs are unchanged.
+
+## Checkpoint identities
+
+| Task | OpenVLA / QwenOFT | π / QwenPI v3 | GR00T / QwenGR00T |
+| --- | --- | --- | --- |
+| Flappy Bird | Transcribed curve; checkpoint binding unconfirmed | step 5000; pi05_fix_latency Flappy weight matches the evaluation source weight | high-learning-rate run, step 3000; exact weight missing |
+| Demon Attack | Transcribed curve; checkpoint binding unconfirmed | personal clean-data exp1, step 5000 | low-learning-rate exp3, step 5000 |
+| Deadly Corridor | personal clean-data BCE exp2, step 500 | personal clean-data exp2, step 500 | personal clean-data BCE exp2, step 500 |
+| MIKASA H1 | original SFT step 5000 | original SFT step 5000 | original SFT step 5000 |
+| MIKASA H8 | naive DAgger selected step 5000 | naive DAgger selected step 4000 | naive DAgger selected step 3000 |
+
+[manifest.json](manifest.json) records the exact source repositories, revisions, checkpoint paths, evidence paths and unresolved bindings. [Model source metadata](https://huggingface.co/latency-sensitive-bench/benchmark-models/tree/66341335f4ead6429bbe8d5df4180a0040f9e0cd/reports/latency-degradation-20261007) preserves the eight original configuration/statistics files available for the six game checkpoints.
+
+The six game checkpoints remain at their fixed source revisions. Copying their weights to benchmark-models is blocked by a Hub LFS/Xet registration error. No central game checkpoint was committed in this batch. Six existing central MIKASA checkpoints are reused.
+
+Flappy GR00T post-training step 3000 evaluation is available. Its source main and all 34 audited commits contain only selected best steps 300, 600, 1800 and 2400. Best step 2400 is not a replacement. The unused `demon_gr00t_step_5000.json` is retained only in the original paper snapshot; the plot consumes `demon_gr00t_action_1e-5_backbone_1e-6_step_5000.json`.
+
+## Evaluation and training data
+
+Seven game JSONs match their pinned remote sources. Each stored latency condition has 20 episodes, seeds 42–61, and verified return means. Two OpenVLA curves are transcriptions from the old plotting script; their original episode counts and checkpoint bindings are not confirmed.
+
+MIKASA has 30 conditions with 200 episodes each, seeds 4242424242–4242424441, simulated 20 Hz execution and a 60-transition limit. The [complete condition records](https://huggingface.co/datasets/latency-sensitive-bench/benchmark-datasets/tree/efeac490144ca9c94660e2410163415e0a8809a9/evaluations/mikasa-intercept-grab-fast/latency-degradation) preserve episode records, action/step traces, latency records, configurations, summaries and completion records. The PI v3 H1 `interleave_self_attention: false` override is retained without changing weights.
+
+Training data and evaluation data are separate. The paper reports 200 zero-latency training episodes for Flappy and Demon, and 1000 for Deadly. Exact original training-data revisions are missing for these old game runs. MIKASA H1 bootstrap data are fixed at benchmark-datasets@5421c9d272d7a939bb04045b047af84519c1fba8, `zero-latency/mikasa-intercept-grab-fast/teacher-rollouts-h1-state-v1`. H8 uses the corresponding suffix-label conversion and student DAgger rollouts labeled by the PPO teacher. Original local DAgger data references remain explicit; no exact Hub revision is invented for those increments.
+
+Game plots retain the source latency labels. Their raw JSONs do not establish FPS or a raw-frame conversion. MIKASA's 20 Hz conversion is explicit. All stored latency conditions are preserved even when a panel plots only a subset.
+
+Runtime model loading was not repeated. No training, reevaluation, source deletion, or checkpoint substitution occurred.

experiments/latency-degradation/deadly-corridor/README.md

@@ -0,0 +1,3 @@
+# Figure 2: deadly-corridor
+
+[Experiment contract and source bindings](../README.md). [Task manifest](manifest.json) records 3 curves and their fixed model/evaluation references.

experiments/latency-degradation/deadly-corridor/manifest.json

@@ -0,0 +1,233 @@
+{
+  "experiment": "latency-degradation",
+  "task": "deadly-corridor",
+  "curves": [
+    {
+      "task": "deadly-corridor",
+      "framework": "qwenoft",
+      "action_horizon": 1,
+      "checkpoint_step": 500,
+      "checkpoint_binding": "confirmed_source",
+      "checkpoint": {
+        "repo": "talha15032/openvla_bridge_deadly_corridor_single_latency_clean_data_bce_exp2",
+        "repo_type": "model",
+        "revision": "31983a233371ba1e755e28b38de3b0b72ded39f7",
+        "path": "steps_500_state/model.safetensors",
+        "sha256": "c6293db2208230c4caa45484b185d417babe948ac5222a600aa9fc62589056a5",
+        "size": 9138230516
+      },
+      "central_checkpoint": {
+        "repo": "latency-sensitive-bench/benchmark-models",
+        "repo_type": "model",
+        "path": "zero-latency/deadly-corridor/vla/starvla-qwenoft-h1/openvla_bridge_deadly_corridor_single_latency_clean_data_bce_exp2/checkpoints/steps_500_model.safetensors",
+        "status": "publication_blocked",
+        "reason": "HF duplicate endpoint reports target already has sha256 stored without xetHash; no central checkpoint was committed."
+      },
+      "original_evaluation": {
+        "repo": "talha15032/openvla_bridge_deadly_corridor_single_latency_clean_data_bce_exp2",
+        "revision": "31983a233371ba1e755e28b38de3b0b72ded39f7",
+        "path": "eval/post_train/step_500.json",
+        "paper_raw_file": "deadly_openvla_step_500.json",
+        "paper_json_equal": true
+      },
+      "evaluation": {
+        "repo": "latency-sensitive-bench/benchmark-datasets",
+        "repo_type": "dataset",
+        "revision": "efeac490144ca9c94660e2410163415e0a8809a9",
+        "path": "evaluations/deadly-corridor/latency-degradation/qwenoft/step_500.json"
+      },
+      "plotted_latency_labels": [
+        0,
+        1,
+        2,
+        3,
+        4,
+        5
+      ],
+      "stored_latency_labels": [
+        0,
+        1,
+        2,
+        3,
+        4,
+        5,
+        6,
+        7,
+        8,
+        9
+      ],
+      "episodes_per_stored_condition": 20,
+      "episode_seeds": [
+        42,
+        61
+      ],
+      "latency_unit_status": "source frame labels retained; game inputs do not establish raw-frame conversion or FPS",
+      "training_data": {
+        "reported_episodes": 1000,
+        "reported_training_latency": 0,
+        "exact_dataset_revision": null,
+        "binding_status": "original_training_dataset_revision_unconfirmed"
+      },
+      "available_metadata": [],
+      "runtime_loading_verified": false,
+      "central_metadata": {
+        "repo": "latency-sensitive-bench/benchmark-models",
+        "repo_type": "model",
+        "path": "reports/latency-degradation-20261007/models/deadly-corridor/openvla_bridge_deadly_corridor_single_latency_clean_data_bce_exp2",
+        "revision": "66341335f4ead6429bbe8d5df4180a0040f9e0cd"
+      }
+    },
+    {
+      "task": "deadly-corridor",
+      "framework": "qwenpi_v3",
+      "action_horizon": 1,
+      "checkpoint_step": 500,
+      "checkpoint_binding": "confirmed_source",
+      "checkpoint": {
+        "repo": "talha15032/pi05_bridge_deadly_corridor_single_latency_clean_data_exp2",
+        "repo_type": "model",
+        "revision": "4d80b52ab5160e20327c843d7737c25ecfcb5ec7",
+        "path": "steps_500_state/model.safetensors",
+        "sha256": "171da5334fbd3a5c9d2b581de32dfd0a06165af73ffca77413fcf70fe96d0945",
+        "size": 11412998004
+      },
+      "central_checkpoint": {
+        "repo": "latency-sensitive-bench/benchmark-models",
+        "repo_type": "model",
+        "path": "zero-latency/deadly-corridor/vla/starvla-qwenpi_v3-h1/pi05_bridge_deadly_corridor_single_latency_clean_data_exp2/checkpoints/steps_500_model.safetensors",
+        "status": "publication_blocked",
+        "reason": "HF duplicate endpoint reports target already has sha256 stored without xetHash; no central checkpoint was committed."
+      },
+      "original_evaluation": {
+        "repo": "talha15032/pi05_bridge_deadly_corridor_single_latency_clean_data_exp2",
+        "revision": "4d80b52ab5160e20327c843d7737c25ecfcb5ec7",
+        "path": "eval/post_train/step_500.json",
+        "paper_raw_file": "deadly_pi05_step_500.json",
+        "paper_json_equal": true
+      },
+      "evaluation": {
+        "repo": "latency-sensitive-bench/benchmark-datasets",
+        "repo_type": "dataset",
+        "revision": "efeac490144ca9c94660e2410163415e0a8809a9",
+        "path": "evaluations/deadly-corridor/latency-degradation/qwenpi_v3/step_500.json"
+      },
+      "plotted_latency_labels": [
+        0,
+        1,
+        2,
+        3,
+        4,
+        5
+      ],
+      "stored_latency_labels": [
+        0,
+        1,
+        2,
+        3,
+        4,
+        5,
+        6,
+        7,
+        8,
+        9,
+        10
+      ],
+      "episodes_per_stored_condition": 20,
+      "episode_seeds": [
+        42,
+        61
+      ],
+      "latency_unit_status": "source frame labels retained; game inputs do not establish raw-frame conversion or FPS",
+      "training_data": {
+        "reported_episodes": 1000,
+        "reported_training_latency": 0,
+        "exact_dataset_revision": null,
+        "binding_status": "original_training_dataset_revision_unconfirmed"
+      },
+      "available_metadata": [],
+      "runtime_loading_verified": false,
+      "central_metadata": {
+        "repo": "latency-sensitive-bench/benchmark-models",
+        "repo_type": "model",
+        "path": "reports/latency-degradation-20261007/models/deadly-corridor/pi05_bridge_deadly_corridor_single_latency_clean_data_exp2",
+        "revision": "66341335f4ead6429bbe8d5df4180a0040f9e0cd"
+      }
+    },
+    {
+      "task": "deadly-corridor",
+      "framework": "qwengr00t",
+      "action_horizon": 1,
+      "checkpoint_step": 500,
+      "checkpoint_binding": "confirmed_source",
+      "checkpoint": {
+        "repo": "talha15032/gr00t_bridge_deadly_corridor_single_latency_clean_data_bce_exp2",
+        "repo_type": "model",
+        "revision": "e93458b10c98dd1f4c712071a74f46dda2fa905d",
+        "path": "steps_500_state/model.safetensors",
+        "sha256": "6e4b8e4858ee32c19303680674afa8dfd22eaf8fab4e90842acf29358102cff6",
+        "size": 9521659380
+      },
+      "central_checkpoint": {
+        "repo": "latency-sensitive-bench/benchmark-models",
+        "repo_type": "model",
+        "path": "zero-latency/deadly-corridor/vla/starvla-qwengr00t-h1/gr00t_bridge_deadly_corridor_single_latency_clean_data_bce_exp2/checkpoints/steps_500_model.safetensors",
+        "status": "publication_blocked",
+        "reason": "HF duplicate endpoint reports target already has sha256 stored without xetHash; no central checkpoint was committed."
+      },
+      "original_evaluation": {
+        "repo": "talha15032/gr00t_bridge_deadly_corridor_single_latency_clean_data_bce_exp2",
+        "revision": "e93458b10c98dd1f4c712071a74f46dda2fa905d",
+        "path": "eval/post_train/step_500.json",
+        "paper_raw_file": "deadly_gr00t_step_500.json",
+        "paper_json_equal": true
+      },
+      "evaluation": {
+        "repo": "latency-sensitive-bench/benchmark-datasets",
+        "repo_type": "dataset",
+        "revision": "efeac490144ca9c94660e2410163415e0a8809a9",
+        "path": "evaluations/deadly-corridor/latency-degradation/qwengr00t/step_500.json"
+      },
+      "plotted_latency_labels": [
+        0,
+        1,
+        2,
+        3,
+        4,
+        5
+      ],
+      "stored_latency_labels": [
+        0,
+        1,
+        2,
+        3,
+        4,
+        5,
+        6,
+        7,
+        8,
+        9,
+        10
+      ],
+      "episodes_per_stored_condition": 20,
+      "episode_seeds": [
+        42,
+        61
+      ],
+      "latency_unit_status": "source frame labels retained; game inputs do not establish raw-frame conversion or FPS",
+      "training_data": {
+        "reported_episodes": 1000,
+        "reported_training_latency": 0,
+        "exact_dataset_revision": null,
+        "binding_status": "original_training_dataset_revision_unconfirmed"
+      },
+      "available_metadata": [],
+      "runtime_loading_verified": false,
+      "central_metadata": {
+        "repo": "latency-sensitive-bench/benchmark-models",
+        "repo_type": "model",
+        "path": "reports/latency-degradation-20261007/models/deadly-corridor/gr00t_bridge_deadly_corridor_single_latency_clean_data_bce_exp2",
+        "revision": "66341335f4ead6429bbe8d5df4180a0040f9e0cd"
+      }
+    }
+  ]
+}

experiments/latency-degradation/demon-attack/README.md

@@ -0,0 +1,3 @@
+# Figure 2: demon-attack
+
+[Experiment contract and source bindings](../README.md). [Task manifest](manifest.json) records 3 curves and their fixed model/evaluation references.

experiments/latency-degradation/demon-attack/manifest.json

@@ -0,0 +1,179 @@
+{
+  "experiment": "latency-degradation",
+  "task": "demon-attack",
+  "curves": [
+    {
+      "task": "demon-attack",
+      "framework": "qwenpi_v3",
+      "action_horizon": 1,
+      "checkpoint_step": 5000,
+      "checkpoint_binding": "confirmed_source",
+      "checkpoint": {
+        "repo": "talha15032/pi05_bridge_demon_attack_single_latency_clean_data_exp1",
+        "repo_type": "model",
+        "revision": "66966cf0ba7889f447b031aa33b865d3ada419da",
+        "path": "steps_5000_state/model.safetensors",
+        "sha256": "168cab931db1793c763bdcf454c8b8bf6817221d150e980c7ec15c42adf64c2f",
+        "size": 11412998004
+      },
+      "central_checkpoint": {
+        "repo": "latency-sensitive-bench/benchmark-models",
+        "repo_type": "model",
+        "path": "zero-latency/demon-attack/vla/starvla-qwenpi_v3-h1/pi05_bridge_demon_attack_single_latency_clean_data_exp1/checkpoints/steps_5000_model.safetensors",
+        "status": "publication_blocked",
+        "reason": "HF duplicate endpoint reports target already has sha256 stored without xetHash; no central checkpoint was committed."
+      },
+      "original_evaluation": {
+        "repo": "talha15032/pi05_bridge_demon_attack_single_latency_clean_data_exp1",
+        "revision": "66966cf0ba7889f447b031aa33b865d3ada419da",
+        "path": "eval/post_train/step_5000.json",
+        "paper_raw_file": "demon_pi05_step_5000.json",
+        "paper_json_equal": true
+      },
+      "evaluation": {
+        "repo": "latency-sensitive-bench/benchmark-datasets",
+        "repo_type": "dataset",
+        "revision": "efeac490144ca9c94660e2410163415e0a8809a9",
+        "path": "evaluations/demon-attack/latency-degradation/qwenpi_v3/step_5000.json"
+      },
+      "plotted_latency_labels": [
+        0,
+        2,
+        4
+      ],
+      "stored_latency_labels": [
+        0,
+        1,
+        2,
+        3,
+        4,
+        5
+      ],
+      "episodes_per_stored_condition": 20,
+      "episode_seeds": [
+        42,
+        61
+      ],
+      "latency_unit_status": "source frame labels retained; game inputs do not establish raw-frame conversion or FPS",
+      "training_data": {
+        "reported_episodes": 200,
+        "reported_training_latency": 0,
+        "exact_dataset_revision": null,
+        "binding_status": "original_training_dataset_revision_unconfirmed"
+      },
+      "available_metadata": [],
+      "runtime_loading_verified": false,
+      "central_metadata": {
+        "repo": "latency-sensitive-bench/benchmark-models",
+        "repo_type": "model",
+        "path": "reports/latency-degradation-20261007/models/demon-attack/pi05_bridge_demon_attack_single_latency_clean_data_exp1",
+        "revision": "66341335f4ead6429bbe8d5df4180a0040f9e0cd"
+      }
+    },
+    {
+      "task": "demon-attack",
+      "framework": "qwengr00t",
+      "action_horizon": 1,
+      "checkpoint_step": 5000,
+      "checkpoint_binding": "confirmed_source",
+      "checkpoint": {
+        "repo": "latency-sensitive-bench/gr00t_bridge_demon_attack_single_latency_clean_data_exp3_action_1e-5_backbone_1e-6",
+        "repo_type": "model",
+        "revision": "fb0b8c4ef4d76b8c7ad97edf4990341f2b93cf70",
+        "path": "checkpoints/steps_5000_model.safetensors",
+        "sha256": "adb7a7b841e0495b5f18a081184a8960a6817c2468645c673f8ac1c090c06cd9",
+        "size": 9521659380
+      },
+      "central_checkpoint": {
+        "repo": "latency-sensitive-bench/benchmark-models",
+        "repo_type": "model",
+        "path": "zero-latency/demon-attack/vla/starvla-qwengr00t-h1/gr00t_bridge_demon_attack_single_latency_clean_data_exp3_action_1e-5_backbone_1e-6/checkpoints/steps_5000_model.safetensors",
+        "status": "publication_blocked",
+        "reason": "HF duplicate endpoint reports target already has sha256 stored without xetHash; no central checkpoint was committed."
+      },
+      "original_evaluation": {
+        "repo": "latency-sensitive-bench/gr00t_bridge_demon_attack_single_latency_clean_data_exp3_action_1e-5_backbone_1e-6",
+        "revision": "fb0b8c4ef4d76b8c7ad97edf4990341f2b93cf70",
+        "path": "eval/post_train/step_5000.json",
+        "paper_raw_file": "demon_gr00t_action_1e-5_backbone_1e-6_step_5000.json",
+        "paper_json_equal": true
+      },
+      "evaluation": {
+        "repo": "latency-sensitive-bench/benchmark-datasets",
+        "repo_type": "dataset",
+        "revision": "efeac490144ca9c94660e2410163415e0a8809a9",
+        "path": "evaluations/demon-attack/latency-degradation/qwengr00t/step_5000.json"
+      },
+      "plotted_latency_labels": [
+        0,
+        2,
+        4
+      ],
+      "stored_latency_labels": [
+        0,
+        2,
+        4,
+        6,
+        8
+      ],
+      "episodes_per_stored_condition": 20,
+      "episode_seeds": [
+        42,
+        61
+      ],
+      "latency_unit_status": "source frame labels retained; game inputs do not establish raw-frame conversion or FPS",
+      "training_data": {
+        "reported_episodes": 200,
+        "reported_training_latency": 0,
+        "exact_dataset_revision": null,
+        "binding_status": "original_training_dataset_revision_unconfirmed"
+      },
+      "available_metadata": [
+        "config.full.yaml",
+        "config.yaml",
+        "dataset_statistics.json",
+        "dataset_statistics_eval.json"
+      ],
+      "runtime_loading_verified": false,
+      "central_metadata": {
+        "repo": "latency-sensitive-bench/benchmark-models",
+        "repo_type": "model",
+        "path": "reports/latency-degradation-20261007/models/demon-attack/gr00t_bridge_demon_attack_single_latency_clean_data_exp3_action_1e-5_backbone_1e-6",
+        "revision": "66341335f4ead6429bbe8d5df4180a0040f9e0cd"
+      }
+    },
+    {
+      "task": "demon-attack",
+      "paper_model_label": "OpenVLA",
+      "framework": "qwenoft",
+      "checkpoint_step": null,
+      "checkpoint_binding": "unconfirmed_transcribed_curve",
+      "checkpoint": null,
+      "evaluation": {
+        "repo": "latency-sensitive-bench/benchmark-datasets",
+        "repo_type": "dataset",
+        "revision": "68f1b2d3bcf42de973e7508dae68461801c501e3",
+        "path": "evaluations/demon-attack/latency-degradation/qwenoft/transcribed_curve.json"
+      },
+      "transcribed_values": {
+        "0": 6557.75,
+        "2": 209.25,
+        "4": 159.25,
+        "6": 92.25,
+        "8": 60.5
+      },
+      "plotted_latency_labels": [
+        0,
+        2,
+        4
+      ],
+      "episodes_per_condition": null,
+      "training_data": {
+        "reported_episodes": 200,
+        "reported_training_latency": 0,
+        "exact_dataset_revision": null,
+        "binding_status": "original_training_dataset_revision_unconfirmed"
+      }
+    }
+  ]
+}

experiments/latency-degradation/flappy/README.md

@@ -0,0 +1,3 @@
+# Figure 2: flappy
+
+[Experiment contract and source bindings](../README.md). [Task manifest](manifest.json) records 3 curves and their fixed model/evaluation references.

experiments/latency-degradation/flappy/manifest.json

@@ -0,0 +1,173 @@
+{
+  "experiment": "latency-degradation",
+  "task": "flappy",
+  "curves": [
+    {
+      "task": "flappy",
+      "framework": "qwenpi_v3",
+      "action_horizon": 1,
+      "checkpoint_step": 5000,
+      "checkpoint_binding": "confirmed_source",
+      "checkpoint": {
+        "repo": "latency-sensitive-bench/pi05_fix_latency",
+        "repo_type": "model",
+        "revision": "d7c75396e88ae9be7f6988869031a2c438bbc54d",
+        "path": "pi05_flappy_fix_latency_0_200ep/checkpoints/steps_5000_state/model.safetensors",
+        "sha256": "c7e7d86dda1257b48f0465818e7d6186c781e92bbd4f1365eae2a9eed215c942",
+        "size": 11412998004
+      },
+      "central_checkpoint": {
+        "repo": "latency-sensitive-bench/benchmark-models",
+        "repo_type": "model",
+        "path": "zero-latency/flappy/vla/starvla-qwenpi_v3-h1/pi05_flappy_fix_latency_0_200ep/checkpoints/steps_5000_model.safetensors",
+        "status": "publication_blocked",
+        "reason": "HF duplicate endpoint reports target already has sha256 stored without xetHash; no central checkpoint was committed."
+      },
+      "original_evaluation": {
+        "repo": "talha15032/pi05_bridge_flappy_single_latency_clean_data_exp2",
+        "revision": "3403d705255901dd76b7f5a12ab1d6379730072f",
+        "path": "eval/post_train/step_5000.json",
+        "paper_raw_file": "flappy_pi05_step_5000.json",
+        "paper_json_equal": true
+      },
+      "evaluation": {
+        "repo": "latency-sensitive-bench/benchmark-datasets",
+        "repo_type": "dataset",
+        "revision": "efeac490144ca9c94660e2410163415e0a8809a9",
+        "path": "evaluations/flappy/latency-degradation/qwenpi_v3/step_5000.json"
+      },
+      "plotted_latency_labels": [
+        0,
+        1,
+        2,
+        3,
+        4
+      ],
+      "stored_latency_labels": [
+        0,
+        1,
+        2,
+        3,
+        4,
+        5
+      ],
+      "episodes_per_stored_condition": 20,
+      "episode_seeds": [
+        42,
+        61
+      ],
+      "latency_unit_status": "source frame labels retained; game inputs do not establish raw-frame conversion or FPS",
+      "training_data": {
+        "reported_episodes": 200,
+        "reported_training_latency": 0,
+        "exact_dataset_revision": null,
+        "binding_status": "original_training_dataset_revision_unconfirmed"
+      },
+      "available_metadata": [
+        "config.full.yaml",
+        "config.yaml",
+        "dataset_statistics.json",
+        "dataset_statistics_eval.json"
+      ],
+      "runtime_loading_verified": false,
+      "central_metadata": {
+        "repo": "latency-sensitive-bench/benchmark-models",
+        "repo_type": "model",
+        "path": "reports/latency-degradation-20261007/models/flappy/pi05_flappy_fix_latency_0_200ep",
+        "revision": "66341335f4ead6429bbe8d5df4180a0040f9e0cd"
+      }
+    },
+    {
+      "task": "flappy",
+      "framework": "qwengr00t",
+      "action_horizon": 1,
+      "checkpoint_step": 3000,
+      "checkpoint_binding": "exact_checkpoint_missing",
+      "checkpoint": null,
+      "source_repository": {
+        "repo": "talha15032/gr00t_bridge_flappy_single_latency_clean_data_exp2_action_1e-4_backbone_1e-5",
+        "repo_type": "model",
+        "revision": "65d9bc63a9a80519380c16662245592382051fc3"
+      },
+      "source_evaluation_path": "eval/post_train/step_3000.json",
+      "evaluation": {
+        "repo": "latency-sensitive-bench/benchmark-datasets",
+        "repo_type": "dataset",
+        "revision": "efeac490144ca9c94660e2410163415e0a8809a9",
+        "path": "evaluations/flappy/latency-degradation/qwengr00t/step_3000.json"
+      },
+      "plotted_latency_labels": [
+        0,
+        1,
+        2,
+        3,
+        4
+      ],
+      "stored_latency_labels": [
+        0,
+        1,
+        2,
+        3,
+        4,
+        5
+      ],
+      "episodes_per_stored_condition": 20,
+      "episode_seeds": [
+        42,
+        61
+      ],
+      "available_best_step": 2400,
+      "available_best_used_as_replacement": false,
+      "history_audit": {
+        "commits_checked": 34,
+        "best_steps_found": [
+          300,
+          600,
+          1800,
+          2400
+        ]
+      },
+      "training_data": {
+        "reported_episodes": 200,
+        "reported_training_latency": 0,
+        "exact_dataset_revision": null,
+        "binding_status": "original_training_dataset_revision_unconfirmed"
+      }
+    },
+    {
+      "task": "flappy",
+      "paper_model_label": "OpenVLA",
+      "framework": "qwenoft",
+      "checkpoint_step": null,
+      "checkpoint_binding": "unconfirmed_transcribed_curve",
+      "checkpoint": null,
+      "evaluation": {
+        "repo": "latency-sensitive-bench/benchmark-datasets",
+        "repo_type": "dataset",
+        "revision": "68f1b2d3bcf42de973e7508dae68461801c501e3",
+        "path": "evaluations/flappy/latency-degradation/qwenoft/transcribed_curve.json"
+      },
+      "transcribed_values": {
+        "0": 438.67,
+        "1": 18.36,
+        "2": 5.11,
+        "3": 4.97,
+        "4": 4.18
+      },
+      "plotted_latency_labels": [
+        0,
+        1,
+        2,
+        3,
+        4
+      ],
+      "episodes_per_condition": null,
+      "training_data": {
+        "reported_episodes": 200,
+        "reported_training_latency": 0,
+        "exact_dataset_revision": null,
+        "binding_status": "original_training_dataset_revision_unconfirmed"
+      }
+    }
+  ]
+}

experiments/latency-degradation/manifest.json

@@ -0,0 +1,874 @@
+{
+  "experiment": "latency-degradation",
+  "paper_figure": 2,
+  "csv_rows": 95,
+  "logical_curves": 15,
+  "plotted_unique_points": 72,
+  "curves": [
+    {
+      "task": "flappy",
+      "framework": "qwenpi_v3",
+      "action_horizon": 1,
+      "checkpoint_step": 5000,
+      "checkpoint_binding": "confirmed_source",
+      "checkpoint": {
+        "repo": "latency-sensitive-bench/pi05_fix_latency",
+        "repo_type": "model",
+        "revision": "d7c75396e88ae9be7f6988869031a2c438bbc54d",
+        "path": "pi05_flappy_fix_latency_0_200ep/checkpoints/steps_5000_state/model.safetensors",
+        "sha256": "c7e7d86dda1257b48f0465818e7d6186c781e92bbd4f1365eae2a9eed215c942",
+        "size": 11412998004
+      },
+      "central_checkpoint": {
+        "repo": "latency-sensitive-bench/benchmark-models",
+        "repo_type": "model",
+        "path": "zero-latency/flappy/vla/starvla-qwenpi_v3-h1/pi05_flappy_fix_latency_0_200ep/checkpoints/steps_5000_model.safetensors",
+        "status": "publication_blocked",
+        "reason": "HF duplicate endpoint reports target already has sha256 stored without xetHash; no central checkpoint was committed."
+      },
+      "original_evaluation": {
+        "repo": "talha15032/pi05_bridge_flappy_single_latency_clean_data_exp2",
+        "revision": "3403d705255901dd76b7f5a12ab1d6379730072f",
+        "path": "eval/post_train/step_5000.json",
+        "paper_raw_file": "flappy_pi05_step_5000.json",
+        "paper_json_equal": true
+      },
+      "evaluation": {
+        "repo": "latency-sensitive-bench/benchmark-datasets",
+        "repo_type": "dataset",
+        "revision": "efeac490144ca9c94660e2410163415e0a8809a9",
+        "path": "evaluations/flappy/latency-degradation/qwenpi_v3/step_5000.json"
+      },
+      "plotted_latency_labels": [
+        0,
+        1,
+        2,
+        3,
+        4
+      ],
+      "stored_latency_labels": [
+        0,
+        1,
+        2,
+        3,
+        4,
+        5
+      ],
+      "episodes_per_stored_condition": 20,
+      "episode_seeds": [
+        42,
+        61
+      ],
+      "latency_unit_status": "source frame labels retained; game inputs do not establish raw-frame conversion or FPS",
+      "training_data": {
+        "reported_episodes": 200,
+        "reported_training_latency": 0,
+        "exact_dataset_revision": null,
+        "binding_status": "original_training_dataset_revision_unconfirmed"
+      },
+      "available_metadata": [
+        "config.full.yaml",
+        "config.yaml",
+        "dataset_statistics.json",
+        "dataset_statistics_eval.json"
+      ],
+      "runtime_loading_verified": false,
+      "central_metadata": {
+        "repo": "latency-sensitive-bench/benchmark-models",
+        "repo_type": "model",
+        "path": "reports/latency-degradation-20261007/models/flappy/pi05_flappy_fix_latency_0_200ep",
+        "revision": "66341335f4ead6429bbe8d5df4180a0040f9e0cd"
+      }
+    },
+    {
+      "task": "demon-attack",
+      "framework": "qwenpi_v3",
+      "action_horizon": 1,
+      "checkpoint_step": 5000,
+      "checkpoint_binding": "confirmed_source",
+      "checkpoint": {
+        "repo": "talha15032/pi05_bridge_demon_attack_single_latency_clean_data_exp1",
+        "repo_type": "model",
+        "revision": "66966cf0ba7889f447b031aa33b865d3ada419da",
+        "path": "steps_5000_state/model.safetensors",
+        "sha256": "168cab931db1793c763bdcf454c8b8bf6817221d150e980c7ec15c42adf64c2f",
+        "size": 11412998004
+      },
+      "central_checkpoint": {
+        "repo": "latency-sensitive-bench/benchmark-models",
+        "repo_type": "model",
+        "path": "zero-latency/demon-attack/vla/starvla-qwenpi_v3-h1/pi05_bridge_demon_attack_single_latency_clean_data_exp1/checkpoints/steps_5000_model.safetensors",
+        "status": "publication_blocked",
+        "reason": "HF duplicate endpoint reports target already has sha256 stored without xetHash; no central checkpoint was committed."
+      },
+      "original_evaluation": {
+        "repo": "talha15032/pi05_bridge_demon_attack_single_latency_clean_data_exp1",
+        "revision": "66966cf0ba7889f447b031aa33b865d3ada419da",
+        "path": "eval/post_train/step_5000.json",
+        "paper_raw_file": "demon_pi05_step_5000.json",
+        "paper_json_equal": true
+      },
+      "evaluation": {
+        "repo": "latency-sensitive-bench/benchmark-datasets",
+        "repo_type": "dataset",
+        "revision": "efeac490144ca9c94660e2410163415e0a8809a9",
+        "path": "evaluations/demon-attack/latency-degradation/qwenpi_v3/step_5000.json"
+      },
+      "plotted_latency_labels": [
+        0,
+        2,
+        4
+      ],
+      "stored_latency_labels": [
+        0,
+        1,
+        2,
+        3,
+        4,
+        5
+      ],
+      "episodes_per_stored_condition": 20,
+      "episode_seeds": [
+        42,
+        61
+      ],
+      "latency_unit_status": "source frame labels retained; game inputs do not establish raw-frame conversion or FPS",
+      "training_data": {
+        "reported_episodes": 200,
+        "reported_training_latency": 0,
+        "exact_dataset_revision": null,
+        "binding_status": "original_training_dataset_revision_unconfirmed"
+      },
+      "available_metadata": [],
+      "runtime_loading_verified": false,
+      "central_metadata": {
+        "repo": "latency-sensitive-bench/benchmark-models",
+        "repo_type": "model",
+        "path": "reports/latency-degradation-20261007/models/demon-attack/pi05_bridge_demon_attack_single_latency_clean_data_exp1",
+        "revision": "66341335f4ead6429bbe8d5df4180a0040f9e0cd"
+      }
+    },
+    {
+      "task": "demon-attack",
+      "framework": "qwengr00t",
+      "action_horizon": 1,
+      "checkpoint_step": 5000,
+      "checkpoint_binding": "confirmed_source",
+      "checkpoint": {
+        "repo": "latency-sensitive-bench/gr00t_bridge_demon_attack_single_latency_clean_data_exp3_action_1e-5_backbone_1e-6",
+        "repo_type": "model",
+        "revision": "fb0b8c4ef4d76b8c7ad97edf4990341f2b93cf70",
+        "path": "checkpoints/steps_5000_model.safetensors",
+        "sha256": "adb7a7b841e0495b5f18a081184a8960a6817c2468645c673f8ac1c090c06cd9",
+        "size": 9521659380
+      },
+      "central_checkpoint": {
+        "repo": "latency-sensitive-bench/benchmark-models",
+        "repo_type": "model",
+        "path": "zero-latency/demon-attack/vla/starvla-qwengr00t-h1/gr00t_bridge_demon_attack_single_latency_clean_data_exp3_action_1e-5_backbone_1e-6/checkpoints/steps_5000_model.safetensors",
+        "status": "publication_blocked",
+        "reason": "HF duplicate endpoint reports target already has sha256 stored without xetHash; no central checkpoint was committed."
+      },
+      "original_evaluation": {
+        "repo": "latency-sensitive-bench/gr00t_bridge_demon_attack_single_latency_clean_data_exp3_action_1e-5_backbone_1e-6",
+        "revision": "fb0b8c4ef4d76b8c7ad97edf4990341f2b93cf70",
+        "path": "eval/post_train/step_5000.json",
+        "paper_raw_file": "demon_gr00t_action_1e-5_backbone_1e-6_step_5000.json",
+        "paper_json_equal": true
+      },
+      "evaluation": {
+        "repo": "latency-sensitive-bench/benchmark-datasets",
+        "repo_type": "dataset",
+        "revision": "efeac490144ca9c94660e2410163415e0a8809a9",
+        "path": "evaluations/demon-attack/latency-degradation/qwengr00t/step_5000.json"
+      },
+      "plotted_latency_labels": [
+        0,
+        2,
+        4
+      ],
+      "stored_latency_labels": [
+        0,
+        2,
+        4,
+        6,
+        8
+      ],
+      "episodes_per_stored_condition": 20,
+      "episode_seeds": [
+        42,
+        61
+      ],
+      "latency_unit_status": "source frame labels retained; game inputs do not establish raw-frame conversion or FPS",
+      "training_data": {
+        "reported_episodes": 200,
+        "reported_training_latency": 0,
+        "exact_dataset_revision": null,
+        "binding_status": "original_training_dataset_revision_unconfirmed"
+      },
+      "available_metadata": [
+        "config.full.yaml",
+        "config.yaml",
+        "dataset_statistics.json",
+        "dataset_statistics_eval.json"
+      ],
+      "runtime_loading_verified": false,
+      "central_metadata": {
+        "repo": "latency-sensitive-bench/benchmark-models",
+        "repo_type": "model",
+        "path": "reports/latency-degradation-20261007/models/demon-attack/gr00t_bridge_demon_attack_single_latency_clean_data_exp3_action_1e-5_backbone_1e-6",
+        "revision": "66341335f4ead6429bbe8d5df4180a0040f9e0cd"
+      }
+    },
+    {
+      "task": "deadly-corridor",
+      "framework": "qwenoft",
+      "action_horizon": 1,
+      "checkpoint_step": 500,
+      "checkpoint_binding": "confirmed_source",
+      "checkpoint": {
+        "repo": "talha15032/openvla_bridge_deadly_corridor_single_latency_clean_data_bce_exp2",
+        "repo_type": "model",
+        "revision": "31983a233371ba1e755e28b38de3b0b72ded39f7",
+        "path": "steps_500_state/model.safetensors",
+        "sha256": "c6293db2208230c4caa45484b185d417babe948ac5222a600aa9fc62589056a5",
+        "size": 9138230516
+      },
+      "central_checkpoint": {
+        "repo": "latency-sensitive-bench/benchmark-models",
+        "repo_type": "model",
+        "path": "zero-latency/deadly-corridor/vla/starvla-qwenoft-h1/openvla_bridge_deadly_corridor_single_latency_clean_data_bce_exp2/checkpoints/steps_500_model.safetensors",
+        "status": "publication_blocked",
+        "reason": "HF duplicate endpoint reports target already has sha256 stored without xetHash; no central checkpoint was committed."
+      },
+      "original_evaluation": {
+        "repo": "talha15032/openvla_bridge_deadly_corridor_single_latency_clean_data_bce_exp2",
+        "revision": "31983a233371ba1e755e28b38de3b0b72ded39f7",
+        "path": "eval/post_train/step_500.json",
+        "paper_raw_file": "deadly_openvla_step_500.json",
+        "paper_json_equal": true
+      },
+      "evaluation": {
+        "repo": "latency-sensitive-bench/benchmark-datasets",
+        "repo_type": "dataset",
+        "revision": "efeac490144ca9c94660e2410163415e0a8809a9",
+        "path": "evaluations/deadly-corridor/latency-degradation/qwenoft/step_500.json"
+      },
+      "plotted_latency_labels": [
+        0,
+        1,
+        2,
+        3,
+        4,
+        5
+      ],
+      "stored_latency_labels": [
+        0,
+        1,
+        2,
+        3,
+        4,
+        5,
+        6,
+        7,
+        8,
+        9
+      ],
+      "episodes_per_stored_condition": 20,
+      "episode_seeds": [
+        42,
+        61
+      ],
+      "latency_unit_status": "source frame labels retained; game inputs do not establish raw-frame conversion or FPS",
+      "training_data": {
+        "reported_episodes": 1000,
+        "reported_training_latency": 0,
+        "exact_dataset_revision": null,
+        "binding_status": "original_training_dataset_revision_unconfirmed"
+      },
+      "available_metadata": [],
+      "runtime_loading_verified": false,
+      "central_metadata": {
+        "repo": "latency-sensitive-bench/benchmark-models",
+        "repo_type": "model",
+        "path": "reports/latency-degradation-20261007/models/deadly-corridor/openvla_bridge_deadly_corridor_single_latency_clean_data_bce_exp2",
+        "revision": "66341335f4ead6429bbe8d5df4180a0040f9e0cd"
+      }
+    },
+    {
+      "task": "deadly-corridor",
+      "framework": "qwenpi_v3",
+      "action_horizon": 1,
+      "checkpoint_step": 500,
+      "checkpoint_binding": "confirmed_source",
+      "checkpoint": {
+        "repo": "talha15032/pi05_bridge_deadly_corridor_single_latency_clean_data_exp2",
+        "repo_type": "model",
+        "revision": "4d80b52ab5160e20327c843d7737c25ecfcb5ec7",
+        "path": "steps_500_state/model.safetensors",
+        "sha256": "171da5334fbd3a5c9d2b581de32dfd0a06165af73ffca77413fcf70fe96d0945",
+        "size": 11412998004
+      },
+      "central_checkpoint": {
+        "repo": "latency-sensitive-bench/benchmark-models",
+        "repo_type": "model",
+        "path": "zero-latency/deadly-corridor/vla/starvla-qwenpi_v3-h1/pi05_bridge_deadly_corridor_single_latency_clean_data_exp2/checkpoints/steps_500_model.safetensors",
+        "status": "publication_blocked",
+        "reason": "HF duplicate endpoint reports target already has sha256 stored without xetHash; no central checkpoint was committed."
+      },
+      "original_evaluation": {
+        "repo": "talha15032/pi05_bridge_deadly_corridor_single_latency_clean_data_exp2",
+        "revision": "4d80b52ab5160e20327c843d7737c25ecfcb5ec7",
+        "path": "eval/post_train/step_500.json",
+        "paper_raw_file": "deadly_pi05_step_500.json",
+        "paper_json_equal": true
+      },
+      "evaluation": {
+        "repo": "latency-sensitive-bench/benchmark-datasets",
+        "repo_type": "dataset",
+        "revision": "efeac490144ca9c94660e2410163415e0a8809a9",
+        "path": "evaluations/deadly-corridor/latency-degradation/qwenpi_v3/step_500.json"
+      },
+      "plotted_latency_labels": [
+        0,
+        1,
+        2,
+        3,
+        4,
+        5
+      ],
+      "stored_latency_labels": [
+        0,
+        1,
+        2,
+        3,
+        4,
+        5,
+        6,
+        7,
+        8,
+        9,
+        10
+      ],
+      "episodes_per_stored_condition": 20,
+      "episode_seeds": [
+        42,
+        61
+      ],
+      "latency_unit_status": "source frame labels retained; game inputs do not establish raw-frame conversion or FPS",
+      "training_data": {
+        "reported_episodes": 1000,
+        "reported_training_latency": 0,
+        "exact_dataset_revision": null,
+        "binding_status": "original_training_dataset_revision_unconfirmed"
+      },
+      "available_metadata": [],
+      "runtime_loading_verified": false,
+      "central_metadata": {
+        "repo": "latency-sensitive-bench/benchmark-models",
+        "repo_type": "model",
+        "path": "reports/latency-degradation-20261007/models/deadly-corridor/pi05_bridge_deadly_corridor_single_latency_clean_data_exp2",
+        "revision": "66341335f4ead6429bbe8d5df4180a0040f9e0cd"
+      }
+    },
+    {
+      "task": "deadly-corridor",
+      "framework": "qwengr00t",
+      "action_horizon": 1,
+      "checkpoint_step": 500,
+      "checkpoint_binding": "confirmed_source",
+      "checkpoint": {
+        "repo": "talha15032/gr00t_bridge_deadly_corridor_single_latency_clean_data_bce_exp2",
+        "repo_type": "model",
+        "revision": "e93458b10c98dd1f4c712071a74f46dda2fa905d",
+        "path": "steps_500_state/model.safetensors",
+        "sha256": "6e4b8e4858ee32c19303680674afa8dfd22eaf8fab4e90842acf29358102cff6",
+        "size": 9521659380
+      },
+      "central_checkpoint": {
+        "repo": "latency-sensitive-bench/benchmark-models",
+        "repo_type": "model",
+        "path": "zero-latency/deadly-corridor/vla/starvla-qwengr00t-h1/gr00t_bridge_deadly_corridor_single_latency_clean_data_bce_exp2/checkpoints/steps_500_model.safetensors",
+        "status": "publication_blocked",
+        "reason": "HF duplicate endpoint reports target already has sha256 stored without xetHash; no central checkpoint was committed."
+      },
+      "original_evaluation": {
+        "repo": "talha15032/gr00t_bridge_deadly_corridor_single_latency_clean_data_bce_exp2",
+        "revision": "e93458b10c98dd1f4c712071a74f46dda2fa905d",
+        "path": "eval/post_train/step_500.json",
+        "paper_raw_file": "deadly_gr00t_step_500.json",
+        "paper_json_equal": true
+      },
+      "evaluation": {
+        "repo": "latency-sensitive-bench/benchmark-datasets",
+        "repo_type": "dataset",
+        "revision": "efeac490144ca9c94660e2410163415e0a8809a9",
+        "path": "evaluations/deadly-corridor/latency-degradation/qwengr00t/step_500.json"
+      },
+      "plotted_latency_labels": [
+        0,
+        1,
+        2,
+        3,
+        4,
+        5
+      ],
+      "stored_latency_labels": [
+        0,
+        1,
+        2,
+        3,
+        4,
+        5,
+        6,
+        7,
+        8,
+        9,
+        10
+      ],
+      "episodes_per_stored_condition": 20,
+      "episode_seeds": [
+        42,
+        61
+      ],
+      "latency_unit_status": "source frame labels retained; game inputs do not establish raw-frame conversion or FPS",
+      "training_data": {
+        "reported_episodes": 1000,
+        "reported_training_latency": 0,
+        "exact_dataset_revision": null,
+        "binding_status": "original_training_dataset_revision_unconfirmed"
+      },
+      "available_metadata": [],
+      "runtime_loading_verified": false,
+      "central_metadata": {
+        "repo": "latency-sensitive-bench/benchmark-models",
+        "repo_type": "model",
+        "path": "reports/latency-degradation-20261007/models/deadly-corridor/gr00t_bridge_deadly_corridor_single_latency_clean_data_bce_exp2",
+        "revision": "66341335f4ead6429bbe8d5df4180a0040f9e0cd"
+      }
+    },
+    {
+      "task": "flappy",
+      "framework": "qwengr00t",
+      "action_horizon": 1,
+      "checkpoint_step": 3000,
+      "checkpoint_binding": "exact_checkpoint_missing",
+      "checkpoint": null,
+      "source_repository": {
+        "repo": "talha15032/gr00t_bridge_flappy_single_latency_clean_data_exp2_action_1e-4_backbone_1e-5",
+        "repo_type": "model",
+        "revision": "65d9bc63a9a80519380c16662245592382051fc3"
+      },
+      "source_evaluation_path": "eval/post_train/step_3000.json",
+      "evaluation": {
+        "repo": "latency-sensitive-bench/benchmark-datasets",
+        "repo_type": "dataset",
+        "revision": "efeac490144ca9c94660e2410163415e0a8809a9",
+        "path": "evaluations/flappy/latency-degradation/qwengr00t/step_3000.json"
+      },
+      "plotted_latency_labels": [
+        0,
+        1,
+        2,
+        3,
+        4
+      ],
+      "stored_latency_labels": [
+        0,
+        1,
+        2,
+        3,
+        4,
+        5
+      ],
+      "episodes_per_stored_condition": 20,
+      "episode_seeds": [
+        42,
+        61
+      ],
+      "available_best_step": 2400,
+      "available_best_used_as_replacement": false,
+      "history_audit": {
+        "commits_checked": 34,
+        "best_steps_found": [
+          300,
+          600,
+          1800,
+          2400
+        ]
+      },
+      "training_data": {
+        "reported_episodes": 200,
+        "reported_training_latency": 0,
+        "exact_dataset_revision": null,
+        "binding_status": "original_training_dataset_revision_unconfirmed"
+      }
+    },
+    {
+      "task": "flappy",
+      "paper_model_label": "OpenVLA",
+      "framework": "qwenoft",
+      "checkpoint_step": null,
+      "checkpoint_binding": "unconfirmed_transcribed_curve",
+      "checkpoint": null,
+      "evaluation": {
+        "repo": "latency-sensitive-bench/benchmark-datasets",
+        "repo_type": "dataset",
+        "revision": "68f1b2d3bcf42de973e7508dae68461801c501e3",
+        "path": "evaluations/flappy/latency-degradation/qwenoft/transcribed_curve.json"
+      },
+      "transcribed_values": {
+        "0": 438.67,
+        "1": 18.36,
+        "2": 5.11,
+        "3": 4.97,
+        "4": 4.18
+      },
+      "plotted_latency_labels": [
+        0,
+        1,
+        2,
+        3,
+        4
+      ],
+      "episodes_per_condition": null,
+      "training_data": {
+        "reported_episodes": 200,
+        "reported_training_latency": 0,
+        "exact_dataset_revision": null,
+        "binding_status": "original_training_dataset_revision_unconfirmed"
+      }
+    },
+    {
+      "task": "demon-attack",
+      "paper_model_label": "OpenVLA",
+      "framework": "qwenoft",
+      "checkpoint_step": null,
+      "checkpoint_binding": "unconfirmed_transcribed_curve",
+      "checkpoint": null,
+      "evaluation": {
+        "repo": "latency-sensitive-bench/benchmark-datasets",
+        "repo_type": "dataset",
+        "revision": "68f1b2d3bcf42de973e7508dae68461801c501e3",
+        "path": "evaluations/demon-attack/latency-degradation/qwenoft/transcribed_curve.json"
+      },
+      "transcribed_values": {
+        "0": 6557.75,
+        "2": 209.25,
+        "4": 159.25,
+        "6": 92.25,
+        "8": 60.5
+      },
+      "plotted_latency_labels": [
+        0,
+        2,
+        4
+      ],
+      "episodes_per_condition": null,
+      "training_data": {
+        "reported_episodes": 200,
+        "reported_training_latency": 0,
+        "exact_dataset_revision": null,
+        "binding_status": "original_training_dataset_revision_unconfirmed"
+      }
+    },
+    {
+      "task": "mikasa-intercept-grab-fast",
+      "framework": "qwenoft",
+      "action_horizon": 1,
+      "checkpoint_step": 5000,
+      "checkpoint_binding": "confirmed_central",
+      "checkpoint": {
+        "repo": "latency-sensitive-bench/benchmark-models",
+        "repo_type": "model",
+        "revision": "56b2c095d76ea5b10dd304e6e03c2ca81801630b",
+        "path": "zero-latency/mikasa-intercept-grab-fast/vla/starvla-qwenoft-h1/mikasa_intercept_grab_fast_qwenoft_formal_latest_r8gpu8_20260907T082154Z/checkpoints/model.pt"
+      },
+      "evaluation": {
+        "repo": "latency-sensitive-bench/benchmark-datasets",
+        "repo_type": "dataset",
+        "revision": "efeac490144ca9c94660e2410163415e0a8809a9",
+        "path": "evaluations/mikasa-intercept-grab-fast/latency-degradation/combined_results.json",
+        "curve_key": "qwenoft_h1"
+      },
+      "execution_record_directory": "evaluations/mikasa-intercept-grab-fast/latency-degradation/qwenoft/h1",
+      "plotted_latency_labels": [
+        0,
+        1,
+        2,
+        3,
+        4
+      ],
+      "episodes_per_condition": 200,
+      "episode_seeds": [
+        4242424242,
+        4242424441
+      ],
+      "env_fps": 20,
+      "max_steps": 60,
+      "executor": "simulated",
+      "training_data": {
+        "bootstrap_source": {
+          "repo": "latency-sensitive-bench/benchmark-datasets",
+          "repo_type": "dataset",
+          "revision": "5421c9d272d7a939bb04045b047af84519c1fba8",
+          "path": "zero-latency/mikasa-intercept-grab-fast/teacher-rollouts-h1-state-v1"
+        },
+        "h8_sft_conversion": false,
+        "dagger_data_source": null,
+        "dagger_data_fixed_hub_revision": null
+      },
+      "inference_config_override": null
+    },
+    {
+      "task": "mikasa-intercept-grab-fast",
+      "framework": "qwenoft",
+      "action_horizon": 8,
+      "checkpoint_step": 5000,
+      "checkpoint_binding": "confirmed_central",
+      "checkpoint": {
+        "repo": "latency-sensitive-bench/benchmark-models",
+        "repo_type": "model",
+        "revision": "5b533937a72e9fa923ecf28b1c1be5a16bad488d",
+        "path": "zero-latency/mikasa-intercept-grab-fast/vla/starvla-qwenoft-h8/h8-naive-dagger-20260923/checkpoints/model.pt"
+      },
+      "evaluation": {
+        "repo": "latency-sensitive-bench/benchmark-datasets",
+        "repo_type": "dataset",
+        "revision": "efeac490144ca9c94660e2410163415e0a8809a9",
+        "path": "evaluations/mikasa-intercept-grab-fast/latency-degradation/combined_results.json",
+        "curve_key": "qwenoft_h8"
+      },
+      "execution_record_directory": "evaluations/mikasa-intercept-grab-fast/latency-degradation/qwenoft/h8",
+      "plotted_latency_labels": [
+        0,
+        1,
+        2,
+        3,
+        4
+      ],
+      "episodes_per_condition": 200,
+      "episode_seeds": [
+        4242424242,
+        4242424441
+      ],
+      "env_fps": 20,
+      "max_steps": 60,
+      "executor": "simulated",
+      "training_data": {
+        "bootstrap_source": {
+          "repo": "latency-sensitive-bench/benchmark-datasets",
+          "repo_type": "dataset",
+          "revision": "5421c9d272d7a939bb04045b047af84519c1fba8",
+          "path": "zero-latency/mikasa-intercept-grab-fast/teacher-rollouts-h1-state-v1"
+        },
+        "h8_sft_conversion": true,
+        "dagger_data_source": "student rollouts labeled by privileged PPO teacher; original per-run local data paths retained in published provenance",
+        "dagger_data_fixed_hub_revision": null
+      },
+      "inference_config_override": null
+    },
+    {
+      "task": "mikasa-intercept-grab-fast",
+      "framework": "qwenpi_v3",
+      "action_horizon": 1,
+      "checkpoint_step": 5000,
+      "checkpoint_binding": "confirmed_central",
+      "checkpoint": {
+        "repo": "latency-sensitive-bench/benchmark-models",
+        "repo_type": "model",
+        "revision": "56b2c095d76ea5b10dd304e6e03c2ca81801630b",
+        "path": "zero-latency/mikasa-intercept-grab-fast/vla/starvla-qwenpi_v3-h1/mikasa_intercept_grab_fast_qwenpi_v3_formal_latest_r8gpu8_20260907T082154Z/checkpoints/model.pt"
+      },
+      "evaluation": {
+        "repo": "latency-sensitive-bench/benchmark-datasets",
+        "repo_type": "dataset",
+        "revision": "efeac490144ca9c94660e2410163415e0a8809a9",
+        "path": "evaluations/mikasa-intercept-grab-fast/latency-degradation/combined_results.json",
+        "curve_key": "qwenpi_v3_h1"
+      },
+      "execution_record_directory": "evaluations/mikasa-intercept-grab-fast/latency-degradation/qwenpi_v3/h1",
+      "plotted_latency_labels": [
+        0,
+        1,
+        2,
+        3,
+        4
+      ],
+      "episodes_per_condition": 200,
+      "episode_seeds": [
+        4242424242,
+        4242424441
+      ],
+      "env_fps": 20,
+      "max_steps": 60,
+      "executor": "simulated",
+      "training_data": {
+        "bootstrap_source": {
+          "repo": "latency-sensitive-bench/benchmark-datasets",
+          "repo_type": "dataset",
+          "revision": "5421c9d272d7a939bb04045b047af84519c1fba8",
+          "path": "zero-latency/mikasa-intercept-grab-fast/teacher-rollouts-h1-state-v1"
+        },
+        "h8_sft_conversion": false,
+        "dagger_data_source": null,
+        "dagger_data_fixed_hub_revision": null
+      },
+      "inference_config_override": {
+        "field": "framework.action_model.diffusion_model_cfg.interleave_self_attention",
+        "hub_value": true,
+        "eval_value": false
+      }
+    },
+    {
+      "task": "mikasa-intercept-grab-fast",
+      "framework": "qwenpi_v3",
+      "action_horizon": 8,
+      "checkpoint_step": 4000,
+      "checkpoint_binding": "confirmed_central",
+      "checkpoint": {
+        "repo": "latency-sensitive-bench/benchmark-models",
+        "repo_type": "model",
+        "revision": "5b533937a72e9fa923ecf28b1c1be5a16bad488d",
+        "path": "zero-latency/mikasa-intercept-grab-fast/vla/starvla-qwenpi_v3-h8/h8-naive-dagger-20260923/checkpoints/model.pt"
+      },
+      "evaluation": {
+        "repo": "latency-sensitive-bench/benchmark-datasets",
+        "repo_type": "dataset",
+        "revision": "efeac490144ca9c94660e2410163415e0a8809a9",
+        "path": "evaluations/mikasa-intercept-grab-fast/latency-degradation/combined_results.json",
+        "curve_key": "qwenpi_v3_h8"
+      },
+      "execution_record_directory": "evaluations/mikasa-intercept-grab-fast/latency-degradation/qwenpi_v3/h8",
+      "plotted_latency_labels": [
+        0,
+        1,
+        2,
+        3,
+        4
+      ],
+      "episodes_per_condition": 200,
+      "episode_seeds": [
+        4242424242,
+        4242424441
+      ],
+      "env_fps": 20,
+      "max_steps": 60,
+      "executor": "simulated",
+      "training_data": {
+        "bootstrap_source": {
+          "repo": "latency-sensitive-bench/benchmark-datasets",
+          "repo_type": "dataset",
+          "revision": "5421c9d272d7a939bb04045b047af84519c1fba8",
+          "path": "zero-latency/mikasa-intercept-grab-fast/teacher-rollouts-h1-state-v1"
+        },
+        "h8_sft_conversion": true,
+        "dagger_data_source": "student rollouts labeled by privileged PPO teacher; original per-run local data paths retained in published provenance",
+        "dagger_data_fixed_hub_revision": null
+      },
+      "inference_config_override": null
+    },
+    {
+      "task": "mikasa-intercept-grab-fast",
+      "framework": "qwengr00t",
+      "action_horizon": 1,
+      "checkpoint_step": 5000,
+      "checkpoint_binding": "confirmed_central",
+      "checkpoint": {
+        "repo": "latency-sensitive-bench/benchmark-models",
+        "repo_type": "model",
+        "revision": "56b2c095d76ea5b10dd304e6e03c2ca81801630b",
+        "path": "zero-latency/mikasa-intercept-grab-fast/vla/starvla-qwengr00t-h1/mikasa_intercept_grab_fast_qwengr00t_formal_latest_r8gpu7_20260907T075003Z/checkpoints/model.pt"
+      },
+      "evaluation": {
+        "repo": "latency-sensitive-bench/benchmark-datasets",
+        "repo_type": "dataset",
+        "revision": "efeac490144ca9c94660e2410163415e0a8809a9",
+        "path": "evaluations/mikasa-intercept-grab-fast/latency-degradation/combined_results.json",
+        "curve_key": "qwengr00t_h1"
+      },
+      "execution_record_directory": "evaluations/mikasa-intercept-grab-fast/latency-degradation/qwengr00t/h1",
+      "plotted_latency_labels": [
+        0,
+        1,
+        2,
+        3,
+        4
+      ],
+      "episodes_per_condition": 200,
+      "episode_seeds": [
+        4242424242,
+        4242424441
+      ],
+      "env_fps": 20,
+      "max_steps": 60,
+      "executor": "simulated",
+      "training_data": {
+        "bootstrap_source": {
+          "repo": "latency-sensitive-bench/benchmark-datasets",
+          "repo_type": "dataset",
+          "revision": "5421c9d272d7a939bb04045b047af84519c1fba8",
+          "path": "zero-latency/mikasa-intercept-grab-fast/teacher-rollouts-h1-state-v1"
+        },
+        "h8_sft_conversion": false,
+        "dagger_data_source": null,
+        "dagger_data_fixed_hub_revision": null
+      },
+      "inference_config_override": null
+    },
+    {
+      "task": "mikasa-intercept-grab-fast",
+      "framework": "qwengr00t",
+      "action_horizon": 8,
+      "checkpoint_step": 3000,
+      "checkpoint_binding": "confirmed_central",
+      "checkpoint": {
+        "repo": "latency-sensitive-bench/benchmark-models",
+        "repo_type": "model",
+        "revision": "5b533937a72e9fa923ecf28b1c1be5a16bad488d",
+        "path": "zero-latency/mikasa-intercept-grab-fast/vla/starvla-qwengr00t-h8/h8-naive-dagger-20260923/checkpoints/model.pt"
+      },
+      "evaluation": {
+        "repo": "latency-sensitive-bench/benchmark-datasets",
+        "repo_type": "dataset",
+        "revision": "efeac490144ca9c94660e2410163415e0a8809a9",
+        "path": "evaluations/mikasa-intercept-grab-fast/latency-degradation/combined_results.json",
+        "curve_key": "qwengr00t_h8"
+      },
+      "execution_record_directory": "evaluations/mikasa-intercept-grab-fast/latency-degradation/qwengr00t/h8",
+      "plotted_latency_labels": [
+        0,
+        1,
+        2,
+        3,
+        4
+      ],
+      "episodes_per_condition": 200,
+      "episode_seeds": [
+        4242424242,
+        4242424441
+      ],
+      "env_fps": 20,
+      "max_steps": 60,
+      "executor": "simulated",
+      "training_data": {
+        "bootstrap_source": {
+          "repo": "latency-sensitive-bench/benchmark-datasets",
+          "repo_type": "dataset",
+          "revision": "5421c9d272d7a939bb04045b047af84519c1fba8",
+          "path": "zero-latency/mikasa-intercept-grab-fast/teacher-rollouts-h1-state-v1"
+        },
+        "h8_sft_conversion": true,
+        "dagger_data_source": "student rollouts labeled by privileged PPO teacher; original per-run local data paths retained in published provenance",
+        "dagger_data_fixed_hub_revision": null
+      },
+      "inference_config_override": null
+    }
+  ],
+  "unused_paper_raw_json": [
+    "demon_gr00t_step_5000.json"
+  ],
+  "checkpoint_copy_status": "blocked",
+  "central_game_weights_committed": 0,
+  "source_repositories_deleted": 0
+}

experiments/latency-degradation/mikasa-intercept-grab-fast/README.md

@@ -0,0 +1,3 @@
+# Figure 2: mikasa-intercept-grab-fast
+
+[Experiment contract and source bindings](../README.md). [Task manifest](manifest.json) records 6 curves and their fixed model/evaluation references.

experiments/latency-degradation/mikasa-intercept-grab-fast/manifest.json

@@ -0,0 +1,298 @@
+{
+  "experiment": "latency-degradation",
+  "task": "mikasa-intercept-grab-fast",
+  "curves": [
+    {
+      "task": "mikasa-intercept-grab-fast",
+      "framework": "qwenoft",
+      "action_horizon": 1,
+      "checkpoint_step": 5000,
+      "checkpoint_binding": "confirmed_central",
+      "checkpoint": {
+        "repo": "latency-sensitive-bench/benchmark-models",
+        "repo_type": "model",
+        "revision": "56b2c095d76ea5b10dd304e6e03c2ca81801630b",
+        "path": "zero-latency/mikasa-intercept-grab-fast/vla/starvla-qwenoft-h1/mikasa_intercept_grab_fast_qwenoft_formal_latest_r8gpu8_20260907T082154Z/checkpoints/model.pt"
+      },
+      "evaluation": {
+        "repo": "latency-sensitive-bench/benchmark-datasets",
+        "repo_type": "dataset",
+        "revision": "efeac490144ca9c94660e2410163415e0a8809a9",
+        "path": "evaluations/mikasa-intercept-grab-fast/latency-degradation/combined_results.json",
+        "curve_key": "qwenoft_h1"
+      },
+      "execution_record_directory": "evaluations/mikasa-intercept-grab-fast/latency-degradation/qwenoft/h1",
+      "plotted_latency_labels": [
+        0,
+        1,
+        2,
+        3,
+        4
+      ],
+      "episodes_per_condition": 200,
+      "episode_seeds": [
+        4242424242,
+        4242424441
+      ],
+      "env_fps": 20,
+      "max_steps": 60,
+      "executor": "simulated",
+      "training_data": {
+        "bootstrap_source": {
+          "repo": "latency-sensitive-bench/benchmark-datasets",
+          "repo_type": "dataset",
+          "revision": "5421c9d272d7a939bb04045b047af84519c1fba8",
+          "path": "zero-latency/mikasa-intercept-grab-fast/teacher-rollouts-h1-state-v1"
+        },
+        "h8_sft_conversion": false,
+        "dagger_data_source": null,
+        "dagger_data_fixed_hub_revision": null
+      },
+      "inference_config_override": null
+    },
+    {
+      "task": "mikasa-intercept-grab-fast",
+      "framework": "qwenoft",
+      "action_horizon": 8,
+      "checkpoint_step": 5000,
+      "checkpoint_binding": "confirmed_central",
+      "checkpoint": {
+        "repo": "latency-sensitive-bench/benchmark-models",
+        "repo_type": "model",
+        "revision": "5b533937a72e9fa923ecf28b1c1be5a16bad488d",
+        "path": "zero-latency/mikasa-intercept-grab-fast/vla/starvla-qwenoft-h8/h8-naive-dagger-20260923/checkpoints/model.pt"
+      },
+      "evaluation": {
+        "repo": "latency-sensitive-bench/benchmark-datasets",
+        "repo_type": "dataset",
+        "revision": "efeac490144ca9c94660e2410163415e0a8809a9",
+        "path": "evaluations/mikasa-intercept-grab-fast/latency-degradation/combined_results.json",
+        "curve_key": "qwenoft_h8"
+      },
+      "execution_record_directory": "evaluations/mikasa-intercept-grab-fast/latency-degradation/qwenoft/h8",
+      "plotted_latency_labels": [
+        0,
+        1,
+        2,
+        3,
+        4
+      ],
+      "episodes_per_condition": 200,
+      "episode_seeds": [
+        4242424242,
+        4242424441
+      ],
+      "env_fps": 20,
+      "max_steps": 60,
+      "executor": "simulated",
+      "training_data": {
+        "bootstrap_source": {
+          "repo": "latency-sensitive-bench/benchmark-datasets",
+          "repo_type": "dataset",
+          "revision": "5421c9d272d7a939bb04045b047af84519c1fba8",
+          "path": "zero-latency/mikasa-intercept-grab-fast/teacher-rollouts-h1-state-v1"
+        },
+        "h8_sft_conversion": true,
+        "dagger_data_source": "student rollouts labeled by privileged PPO teacher; original per-run local data paths retained in published provenance",
+        "dagger_data_fixed_hub_revision": null
+      },
+      "inference_config_override": null
+    },
+    {
+      "task": "mikasa-intercept-grab-fast",
+      "framework": "qwenpi_v3",
+      "action_horizon": 1,
+      "checkpoint_step": 5000,
+      "checkpoint_binding": "confirmed_central",
+      "checkpoint": {
+        "repo": "latency-sensitive-bench/benchmark-models",
+        "repo_type": "model",
+        "revision": "56b2c095d76ea5b10dd304e6e03c2ca81801630b",
+        "path": "zero-latency/mikasa-intercept-grab-fast/vla/starvla-qwenpi_v3-h1/mikasa_intercept_grab_fast_qwenpi_v3_formal_latest_r8gpu8_20260907T082154Z/checkpoints/model.pt"
+      },
+      "evaluation": {
+        "repo": "latency-sensitive-bench/benchmark-datasets",
+        "repo_type": "dataset",
+        "revision": "efeac490144ca9c94660e2410163415e0a8809a9",
+        "path": "evaluations/mikasa-intercept-grab-fast/latency-degradation/combined_results.json",
+        "curve_key": "qwenpi_v3_h1"
+      },
+      "execution_record_directory": "evaluations/mikasa-intercept-grab-fast/latency-degradation/qwenpi_v3/h1",
+      "plotted_latency_labels": [
+        0,
+        1,
+        2,
+        3,
+        4
+      ],
+      "episodes_per_condition": 200,
+      "episode_seeds": [
+        4242424242,
+        4242424441
+      ],
+      "env_fps": 20,
+      "max_steps": 60,
+      "executor": "simulated",
+      "training_data": {
+        "bootstrap_source": {
+          "repo": "latency-sensitive-bench/benchmark-datasets",
+          "repo_type": "dataset",
+          "revision": "5421c9d272d7a939bb04045b047af84519c1fba8",
+          "path": "zero-latency/mikasa-intercept-grab-fast/teacher-rollouts-h1-state-v1"
+        },
+        "h8_sft_conversion": false,
+        "dagger_data_source": null,
+        "dagger_data_fixed_hub_revision": null
+      },
+      "inference_config_override": {
+        "field": "framework.action_model.diffusion_model_cfg.interleave_self_attention",
+        "hub_value": true,
+        "eval_value": false
+      }
+    },
+    {
+      "task": "mikasa-intercept-grab-fast",
+      "framework": "qwenpi_v3",
+      "action_horizon": 8,
+      "checkpoint_step": 4000,
+      "checkpoint_binding": "confirmed_central",
+      "checkpoint": {
+        "repo": "latency-sensitive-bench/benchmark-models",
+        "repo_type": "model",
+        "revision": "5b533937a72e9fa923ecf28b1c1be5a16bad488d",
+        "path": "zero-latency/mikasa-intercept-grab-fast/vla/starvla-qwenpi_v3-h8/h8-naive-dagger-20260923/checkpoints/model.pt"
+      },
+      "evaluation": {
+        "repo": "latency-sensitive-bench/benchmark-datasets",
+        "repo_type": "dataset",
+        "revision": "efeac490144ca9c94660e2410163415e0a8809a9",
+        "path": "evaluations/mikasa-intercept-grab-fast/latency-degradation/combined_results.json",
+        "curve_key": "qwenpi_v3_h8"
+      },
+      "execution_record_directory": "evaluations/mikasa-intercept-grab-fast/latency-degradation/qwenpi_v3/h8",
+      "plotted_latency_labels": [
+        0,
+        1,
+        2,
+        3,
+        4
+      ],
+      "episodes_per_condition": 200,
+      "episode_seeds": [
+        4242424242,
+        4242424441
+      ],
+      "env_fps": 20,
+      "max_steps": 60,
+      "executor": "simulated",
+      "training_data": {
+        "bootstrap_source": {
+          "repo": "latency-sensitive-bench/benchmark-datasets",
+          "repo_type": "dataset",
+          "revision": "5421c9d272d7a939bb04045b047af84519c1fba8",
+          "path": "zero-latency/mikasa-intercept-grab-fast/teacher-rollouts-h1-state-v1"
+        },
+        "h8_sft_conversion": true,
+        "dagger_data_source": "student rollouts labeled by privileged PPO teacher; original per-run local data paths retained in published provenance",
+        "dagger_data_fixed_hub_revision": null
+      },
+      "inference_config_override": null
+    },
+    {
+      "task": "mikasa-intercept-grab-fast",
+      "framework": "qwengr00t",
+      "action_horizon": 1,
+      "checkpoint_step": 5000,
+      "checkpoint_binding": "confirmed_central",
+      "checkpoint": {
+        "repo": "latency-sensitive-bench/benchmark-models",
+        "repo_type": "model",
+        "revision": "56b2c095d76ea5b10dd304e6e03c2ca81801630b",
+        "path": "zero-latency/mikasa-intercept-grab-fast/vla/starvla-qwengr00t-h1/mikasa_intercept_grab_fast_qwengr00t_formal_latest_r8gpu7_20260907T075003Z/checkpoints/model.pt"
+      },
+      "evaluation": {
+        "repo": "latency-sensitive-bench/benchmark-datasets",
+        "repo_type": "dataset",
+        "revision": "efeac490144ca9c94660e2410163415e0a8809a9",
+        "path": "evaluations/mikasa-intercept-grab-fast/latency-degradation/combined_results.json",
+        "curve_key": "qwengr00t_h1"
+      },
+      "execution_record_directory": "evaluations/mikasa-intercept-grab-fast/latency-degradation/qwengr00t/h1",
+      "plotted_latency_labels": [
+        0,
+        1,
+        2,
+        3,
+        4
+      ],
+      "episodes_per_condition": 200,
+      "episode_seeds": [
+        4242424242,
+        4242424441
+      ],
+      "env_fps": 20,
+      "max_steps": 60,
+      "executor": "simulated",
+      "training_data": {
+        "bootstrap_source": {
+          "repo": "latency-sensitive-bench/benchmark-datasets",
+          "repo_type": "dataset",
+          "revision": "5421c9d272d7a939bb04045b047af84519c1fba8",
+          "path": "zero-latency/mikasa-intercept-grab-fast/teacher-rollouts-h1-state-v1"
+        },
+        "h8_sft_conversion": false,
+        "dagger_data_source": null,
+        "dagger_data_fixed_hub_revision": null
+      },
+      "inference_config_override": null
+    },
+    {
+      "task": "mikasa-intercept-grab-fast",
+      "framework": "qwengr00t",
+      "action_horizon": 8,
+      "checkpoint_step": 3000,
+      "checkpoint_binding": "confirmed_central",
+      "checkpoint": {
+        "repo": "latency-sensitive-bench/benchmark-models",
+        "repo_type": "model",
+        "revision": "5b533937a72e9fa923ecf28b1c1be5a16bad488d",
+        "path": "zero-latency/mikasa-intercept-grab-fast/vla/starvla-qwengr00t-h8/h8-naive-dagger-20260923/checkpoints/model.pt"
+      },
+      "evaluation": {
+        "repo": "latency-sensitive-bench/benchmark-datasets",
+        "repo_type": "dataset",
+        "revision": "efeac490144ca9c94660e2410163415e0a8809a9",
+        "path": "evaluations/mikasa-intercept-grab-fast/latency-degradation/combined_results.json",
+        "curve_key": "qwengr00t_h8"
+      },
+      "execution_record_directory": "evaluations/mikasa-intercept-grab-fast/latency-degradation/qwengr00t/h8",
+      "plotted_latency_labels": [
+        0,
+        1,
+        2,
+        3,
+        4
+      ],
+      "episodes_per_condition": 200,
+      "episode_seeds": [
+        4242424242,
+        4242424441
+      ],
+      "env_fps": 20,
+      "max_steps": 60,
+      "executor": "simulated",
+      "training_data": {
+        "bootstrap_source": {
+          "repo": "latency-sensitive-bench/benchmark-datasets",
+          "repo_type": "dataset",
+          "revision": "5421c9d272d7a939bb04045b047af84519c1fba8",
+          "path": "zero-latency/mikasa-intercept-grab-fast/teacher-rollouts-h1-state-v1"
+        },
+        "h8_sft_conversion": true,
+        "dagger_data_source": "student rollouts labeled by privileged PPO teacher; original per-run local data paths retained in published provenance",
+        "dagger_data_fixed_hub_revision": null
+      },
+      "inference_config_override": null
+    }
+  ]
+}