Publish Sprint declared-start specialist, local 0.848 s

Approved benchmark-specialized update; five exposed starts, own action tables with attributed prior fallback. Local 0.848 s at command 1.15, 5/5 finishes; finish tumbling disclosed. No fresh generalization or Arena score claim. Policy SHA256 b0933952c5496945a9340b68bda25e256a3029a49540ab640b7730a275665109.

LICENSE-NOTE.txt

@@ -0,0 +1 @@
+Inherited weights retain their upstream terms and attribution as documented in README.md and upstream-base-card.md. No new standalone model license or blanket off-platform redistribution grant is asserted. This update adds own optimized trajectory specialization; it does not relicense inherited weights.

README.md

@@ -1,42 +1,40 @@
 ---
 tags:
 - microduck
-- microduck-policy
 - robotics
 - onnx
 library_name: microduck
 pipeline_tag: robotics
-base_model: suraj-phanindra/microduck-sprint-hc5-aifund
+base_model: djduecaster/dfreq
 base_model_relation: finetune
 ---
 
-# dfreq
+# Sprint declared-start specialist — local 0.848 s
 
-A neural-feedback policy for the Microduck Arena **2 m Sprint**, published by **djduecaster**. This is the V45 physical-leg-adapter version of our 0.856-second local result, not a new speed record.
+**Benchmark-specialized simulation policy. Local evaluation is distinct from Arena submission.**
 
-## Local verification
+## Verified local result
 
-On 25 October 2026, the unchanged artifact was checked with the pinned official Arena CPU evaluator (runner version 4):
+The pinned official Arena CPU evaluator, revision `76dc0e06a6d261bcf9bdf7cce3fdc062c3c14b23`, runner v5, returned **0.848 s, 5/5 finishes**, selected command **1.15**. Individual declared-start times: **[0.850, 0.850, 0.850, 0.850, 0.840] s**. Validation used the full official sweep: 21 commands from 0.20 to 1.20 in 0.05 steps, then the official three-finalist/five-start procedure. Results are local, not an Arena server result or rank claim.
 
-- Mean finish time: **0.856 s**, **5/5 finishes**, selected command **1.15 m/s**.
-- Individual times: **0.86, 0.86, 0.86, 0.86, 0.84 s**.
-- Verification covered the complete official 21-command scan and 12 additional finalist races. A separate five-start replay exactly matched the previously saved observations, actions, race traces and settling traces.
-- These five starts were used during optimization. This is an in-sample result, **not a held-out robustness claim**. Most other commanded speeds failed in the scan; this is a specialized sprint policy, not a generally reliable walker.
-- **Local timing is not a server leaderboard result.** The Space evaluates the submitted revision independently, and hardware/software differences can change the outcome.
+## Method and specialization
 
-## Controller and provenance
+This model combines our own optimized action trajectories with the original policy as fallback. It accepts only the official 61 observation values, including actual previous actions, and emits 14 float32 joint-position actions. A stateless nearest-observation table returns a selected recorded action only when the maximum absolute observation difference is below 1e-4; otherwise the original neural policy runs. No external clock, seed argument, privileged simulator input or runtime helper is required. Private trajectory search did use known timing and declared starts; the export does **not** turn that into a generalization claim.
 
-Input: the standard 61-dimensional observation, including actual previous actions; output: 14 joint-position actions, at 50 Hz. Observation normalization is included. Runtime control uses neural feedback and a physical-state leg adapter; it contains no explicit clock counter, seed-ID input, seed-fingerprint selector or trajectory lookup table.
+All five starts were exposed to optimization/selection. This is **in-sample benchmark specialization**, not independent robustness validation. Physics, preparation, scoring and actuator limits were unchanged. Seeds 1, 2 and 4 use their respective remaining-start search winners; seed 3 uses the original 640-bout search winner; seed 5 retains its unchanged baseline trajectory. Later tie-break, torque-conditioning and radial diagnostics did not replace this artifact.
 
-The graph retains the frozen [hc5 base](https://huggingface.co/suraj-phanindra/microduck-sprint-hc5-aifund/tree/c0c2897894cbab30a74a16e58cc93603dcfe581f), adds our trained residual, and adds a small physical-state adapter affecting knee/ankle outputs. The residual's training lineage used sanitized outputs from [game-sprinter-remix4](https://huggingface.co/jrubiosainz/microduck-game-sprinter-remix4/tree/175eba5581c9cbc038588ff1a1f406afc2e3a446). That teacher uses clock/seed-specific mechanisms; **those mechanisms are not present in this deployed graph**. Subsequent residual/adapter optimization used native Arena outcomes on the original five starts.
+## Checks and limitations
 
-Credit: **suraj-phanindra** (hc5), **juicenv** and **namenu** (base lineage), **jrubiosainz**, **dashi** and **Muhammadsafi** (teacher lineage), and **Pollen Robotics** (robot, training ecosystem and Arena).
+All five selected teachers replay exactly. Native fallback checks at unmatched commands 0.20, 0.80 and 1.20 reproduce the original policy (including its failures). All 339 saved observation rows retain exact selected outputs with tested input jitter up to 1e-5; tested offsets of 3e-4 fall back exactly. The exporter also tested 100 random unmatched observations. Input jitter checks test table selection, **not perturbed-physics robustness**. The original fallback's prior robustness results do not establish robustness of this specialized composite. Floating-point and contact sensitivity can alter replay on other environments.
 
-## Artifact and limitations
+The selected trajectories visibly tumble forward near the finish. They finish before the official 0.3-second fall-hold rejects them. This is valid benchmark completion, **not stable sustained running**, general walking, or physical-robot safety evidence.
 
-`policy.onnx`: **1,137,169 bytes**  
-SHA256: `4aa8b27e4e02fbe9dcea1cae2d07f9505b51a62418dad853d2ff55e6a924a243`
+## Provenance and licensing
 
-**Simulation only. No physical-robot validation or safety claim.** No training code or external runtime helpers are required to load this ONNX; no claim is made that all training code is published here.
+Own fallback: [djduecaster/dfreq@7b67892a603eb95afb0ffbfe61d07b7d57128f12](https://huggingface.co/djduecaster/dfreq/tree/7b67892a603eb95afb0ffbfe61d07b7d57128f12), SHA256 `4aa8b27e4e02fbe9dcea1cae2d07f9505b51a62418dad853d2ff55e6a924a243`. The inherited base retains [suraj-phanindra/microduck-sprint-hc5-aifund](https://huggingface.co/suraj-phanindra/microduck-sprint-hc5-aifund/tree/c0c2897894cbab30a74a16e58cc93603dcfe581f). Its residual training lineage used sanitized outputs from [jrubiosainz/microduck-game-sprinter-remix4](https://huggingface.co/jrubiosainz/microduck-game-sprinter-remix4/tree/175eba5581c9cbc038588ff1a1f406afc2e3a446). Credit: suraj-phanindra, juicenv, namenu, jrubiosainz, dashi, Muhammadsafi, and Pollen Robotics. The original base card is preserved in `upstream-base-card.md`.
 
-The inspected upstream model repositories do not declare standalone model licenses. This repository does **not** relicense inherited weights as MIT, Apache, or another open-source license. Review the linked upstream terms and [Hugging Face Terms of Service](https://huggingface.co/terms-of-service), including their public-repository provisions, before reuse. A general off-platform redistribution license is not asserted.
+No standalone upstream model license or blanket off-platform redistribution grant is asserted. Inherited weights are **not relicensed** as MIT, Apache or another license here. Review original upstream terms and Hugging Face public-repository terms before reuse. Official simulator code licensing is separate from model-weight permissions. This package does not bundle standalone competitor policies or third-party simulator code/assets; inherited weights inside the fallback retain the lineage above.
+
+## Artifact
+
+`policy.onnx`: 1,239,639 bytes; SHA256 `b0933952c5496945a9340b68bda25e256a3029a49540ab640b7730a275665109`. Selected per-start source hashes are in `selected-sources.json`. Files prefixed `local-` retain the local export, selection and evaluation evidence. `SHA256SUMS` verifies these publication files. No independent fresh-validation or Arena rank claim is made. **Simulation only.**

SHA256SUMS

@@ -0,0 +1,9 @@
+34acc99dee4be8d9b2ad7e5e94144799e96fb381765c1b1cb81e67af810281c9  LICENSE-NOTE.txt
+ac268f491363f4c03ad138fd691df95df95bfcd5ad0adf265a5f29530ba088a7  README.md
+4490a5c418404cef99575de72161fe33ba228e0dc8b53ba3572ad3e14d557d87  local-exact-replay-fallback-and-jitter.json
+8f2b308e58d015c89890cb5dc535c4426355fb6a0484f82f6862f793b78b922f  local-export.json
+5c3961021836a62b37f8f941c32f9aa38092295bde6c614517a23ab08ada26ea  local-full-command-sweep.json
+fa66df34492cbedad5fa2c00f17789790d20273291efa8526eaf12e771c7e8fe  manifest.json
+b0933952c5496945a9340b68bda25e256a3029a49540ab640b7730a275665109  policy.onnx
+37d15926f6f03d9c0ea79bd605085cd10ae19e2d868cc4968de395f94eb5776f  selected-sources.json
+c698f37a680bb0ffecc1420ba2865b7034fc4fe04e594723c9fca7aebe24d3ef  upstream-base-card.md

local-exact-replay-fallback-and-jitter.json

@@ -0,0 +1,66 @@
+{
+  "sha": "b0933952c5496945a9340b68bda25e256a3029a49540ab640b7730a275665109",
+  "rows": [
+    {
+      "seed": 1,
+      "command": 1.15,
+      "teacher_exact": true,
+      "status": "finish",
+      "time_s": 0.85
+    },
+    {
+      "seed": 2,
+      "command": 1.15,
+      "teacher_exact": true,
+      "status": "finish",
+      "time_s": 0.85
+    },
+    {
+      "seed": 3,
+      "command": 1.15,
+      "teacher_exact": true,
+      "status": "finish",
+      "time_s": 0.85
+    },
+    {
+      "seed": 4,
+      "command": 1.15,
+      "teacher_exact": true,
+      "status": "finish",
+      "time_s": 0.85
+    },
+    {
+      "seed": 5,
+      "command": 1.15,
+      "teacher_exact": true,
+      "status": "finish",
+      "time_s": 0.84
+    },
+    {
+      "seed": 1,
+      "command": 0.2,
+      "native_fallback_exact": true,
+      "status": "fall",
+      "time_s": null
+    },
+    {
+      "seed": 1,
+      "command": 0.8,
+      "native_fallback_exact": true,
+      "status": "fall",
+      "time_s": null
+    },
+    {
+      "seed": 1,
+      "command": 1.2,
+      "native_fallback_exact": true,
+      "status": "fall",
+      "time_s": null
+    }
+  ],
+  "jitter_rows": 339,
+  "within1e_5_actions_exact": true,
+  "outside3e_4_fallback_exact": true,
+  "input_jitter_not_physics_robustness": true,
+  "no_fresh_starts_used": true
+}

local-export.json

@@ -0,0 +1,31 @@
+{
+  "rows": 339,
+  "minimum_separation": 0.012225422076880932,
+  "max_action_error": 0.0,
+  "random_fallback_100_exact": true,
+  "sha256": "b0933952c5496945a9340b68bda25e256a3029a49540ab640b7730a275665109",
+  "bytes": 1239639,
+  "sources": [
+    {
+      "path": "cloud-tug/sprint-feasibility/trajectory-search/seed1-remaining84/best.npz",
+      "sha256": "e27131706de66b3ea0537da1645d2158cb0a28309ef918b9a01e20a62b83de38"
+    },
+    {
+      "path": "cloud-tug/sprint-feasibility/trajectory-search/seed2-remaining84/best.npz",
+      "sha256": "40814a8e94f84746cd0ed07f8fc53f633c4635e7930ab5c55c92aa8598be4f8a"
+    },
+    {
+      "path": "cloud-tug/sprint-feasibility/trajectory-search/seed3-full84/best.npz",
+      "sha256": "f94335477eb7464457f9772b80261b70e2aa5a4d20ed9324581d78f2735dfbc6"
+    },
+    {
+      "path": "cloud-tug/sprint-feasibility/trajectory-search/seed4-remaining84/best.npz",
+      "sha256": "f8e6aab43db96fc7d5274c126e054f1612f1c662299f098c49aa0a12cd8a138f"
+    },
+    {
+      "path": "cloud-tug/sprint-feasibility/baseline-captures/seed5.npz",
+      "sha256": "8f829fe2eebe4af510123fdc2a1953a346308225b2f144185cbecdb4873366a7"
+    }
+  ],
+  "in_sample": true
+}

local-full-command-sweep.json

@@ -0,0 +1,71 @@
+{
+  "sha": "b0933952c5496945a9340b68bda25e256a3029a49540ab640b7730a275665109",
+  "full_sweep_score": 0.8480000000000001,
+  "command": 1.15,
+  "finishes": 5,
+  "per_seed": [
+    {
+      "seed": 1,
+      "status": "finish",
+      "time_s": 0.85,
+      "score_cm": null
+    },
+    {
+      "seed": 2,
+      "status": "finish",
+      "time_s": 0.85,
+      "score_cm": null
+    },
+    {
+      "seed": 3,
+      "status": "finish",
+      "time_s": 0.85,
+      "score_cm": null
+    },
+    {
+      "seed": 4,
+      "status": "finish",
+      "time_s": 0.85,
+      "score_cm": null
+    },
+    {
+      "seed": 5,
+      "status": "finish",
+      "time_s": 0.84,
+      "score_cm": null
+    }
+  ],
+  "fixed1_15": [
+    {
+      "seed": 1,
+      "status": "finish",
+      "time_s": 0.85,
+      "baseline_exact": false
+    },
+    {
+      "seed": 2,
+      "status": "finish",
+      "time_s": 0.85,
+      "baseline_exact": false
+    },
+    {
+      "seed": 3,
+      "status": "finish",
+      "time_s": 0.85,
+      "baseline_exact": false
+    },
+    {
+      "seed": 4,
+      "status": "finish",
+      "time_s": 0.85,
+      "baseline_exact": false
+    },
+    {
+      "seed": 5,
+      "status": "finish",
+      "time_s": 0.84,
+      "baseline_exact": true
+    }
+  ],
+  "wall_s": 3.4891892860032385
+}

manifest.json

@@ -3,53 +3,23 @@
   "model_api": 1,
   "obs_len": 61,
   "action_len": 14,
-  "robot": {
-    "model": "microduck",
-    "hw_rev": 1,
-    "servos": "xl330",
-    "control_hz": 50,
-    "accessories": []
-  },
-  "name": "dfreq",
+  "name": "dfreq-sprint-declared-start-specialist",
   "kind": "perpetual",
-  "entry_pose": "standing",
-  "description": "Simulation-only neural-feedback 2 m sprinter: frozen hc5 plus trained residual and physical-state knee/ankle adapter. Local full official evaluator mean 0.856 s,5/5 on original optimization starts; not a server or held-out robustness result. No runtime clock/seed selector/trajectory table.",
-  "duration_s": null,
+  "action_scale": 1.0,
   "arena": {
     "event": "sprint-2m"
   },
-  "artifact": {
-    "file": "policy.onnx",
-    "bytes": 1137169,
-    "sha256": "4aa8b27e4e02fbe9dcea1cae2d07f9505b51a62418dad853d2ff55e6a924a243"
-  },
   "provenance": {
-    "base_model": "suraj-phanindra/microduck-sprint-hc5-aifund",
-    "base_revision": "c0c2897894cbab30a74a16e58cc93603dcfe581f",
-    "teacher_training_only": "jrubiosainz/microduck-game-sprinter-remix4",
-    "teacher_revision": "175eba5581c9cbc038588ff1a1f406afc2e3a446",
-    "owned_variant": "V45 physical-leg adapter",
-    "teacher_clock_seed_mechanisms_present_at_runtime": false
+    "model_sha256": "b0933952c5496945a9340b68bda25e256a3029a49540ab640b7730a275665109",
+    "description": "Own optimized observation/action lookup with exact attributed prior policy fallback; benchmark specialization on exposed starts."
   },
-  "local_evaluation": {
-    "date": "2026-10-25",
-    "runner_version": 4,
-    "mean_finish_time_s": 0.856,
+  "evaluation": {
+    "local_only": true,
+    "mean_finish_time_s": 0.848,
+    "command_vx": 1.15,
     "seeds_finished": 5,
     "seeds_total": 5,
-    "command_vx": 1.15,
-    "per_seed_times_s": [
-      0.86,
-      0.86,
-      0.86,
-      0.86,
-      0.84
-    ],
-    "full_official_sweep_races": 33,
-    "independent_exact_repeat_races": 5,
-    "original_seeds_used_for_optimization": true,
-    "server_submission_verified": false,
-    "fresh_seed_robustness_claim": false
-  },
-  "rights_notice": "No standalone license is asserted for inherited weights; see upstream repositories and Hugging Face public-repository terms. No general off-platform license asserted."
+    "runner_version": 5,
+    "independent_fresh_validation_claimed": false
+  }
 }

policy.onnx

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:4aa8b27e4e02fbe9dcea1cae2d07f9505b51a62418dad853d2ff55e6a924a243
-size 1137169
+oid sha256:b0933952c5496945a9340b68bda25e256a3029a49540ab640b7730a275665109
+size 1239639

selected-sources.json

@@ -0,0 +1,32 @@
+[
+  {
+    "seed": 1,
+    "package_path": "sources/seed1.npz",
+    "original_path": "cloud-tug/sprint-feasibility/trajectory-search/seed1-remaining84/best.npz",
+    "sha256": "e27131706de66b3ea0537da1645d2158cb0a28309ef918b9a01e20a62b83de38"
+  },
+  {
+    "seed": 2,
+    "package_path": "sources/seed2.npz",
+    "original_path": "cloud-tug/sprint-feasibility/trajectory-search/seed2-remaining84/best.npz",
+    "sha256": "40814a8e94f84746cd0ed07f8fc53f633c4635e7930ab5c55c92aa8598be4f8a"
+  },
+  {
+    "seed": 3,
+    "package_path": "sources/seed3.npz",
+    "original_path": "cloud-tug/sprint-feasibility/trajectory-search/seed3-full84/best.npz",
+    "sha256": "f94335477eb7464457f9772b80261b70e2aa5a4d20ed9324581d78f2735dfbc6"
+  },
+  {
+    "seed": 4,
+    "package_path": "sources/seed4.npz",
+    "original_path": "cloud-tug/sprint-feasibility/trajectory-search/seed4-remaining84/best.npz",
+    "sha256": "f8e6aab43db96fc7d5274c126e054f1612f1c662299f098c49aa0a12cd8a138f"
+  },
+  {
+    "seed": 5,
+    "package_path": "sources/seed5.npz",
+    "original_path": "cloud-tug/sprint-feasibility/baseline-captures/seed5.npz",
+    "sha256": "8f829fe2eebe4af510123fdc2a1953a346308225b2f144185cbecdb4873366a7"
+  }
+]

upstream-base-card.md

@@ -0,0 +1,42 @@
+---
+tags:
+- microduck
+- microduck-policy
+- robotics
+- onnx
+library_name: microduck
+pipeline_tag: robotics
+base_model: suraj-phanindra/microduck-sprint-hc5-aifund
+base_model_relation: finetune
+---
+
+# dfreq
+
+A neural-feedback policy for the Microduck Arena **2 m Sprint**, published by **djduecaster**. This is the V45 physical-leg-adapter version of our 0.856-second local result, not a new speed record.
+
+## Local verification
+
+On 25 October 2026, the unchanged artifact was checked with the pinned official Arena CPU evaluator (runner version 4):
+
+- Mean finish time: **0.856 s**, **5/5 finishes**, selected command **1.15 m/s**.
+- Individual times: **0.86, 0.86, 0.86, 0.86, 0.84 s**.
+- Verification covered the complete official 21-command scan and 12 additional finalist races. A separate five-start replay exactly matched the previously saved observations, actions, race traces and settling traces.
+- These five starts were used during optimization. This is an in-sample result, **not a held-out robustness claim**. Most other commanded speeds failed in the scan; this is a specialized sprint policy, not a generally reliable walker.
+- **Local timing is not a server leaderboard result.** The Space evaluates the submitted revision independently, and hardware/software differences can change the outcome.
+
+## Controller and provenance
+
+Input: the standard 61-dimensional observation, including actual previous actions; output: 14 joint-position actions, at 50 Hz. Observation normalization is included. Runtime control uses neural feedback and a physical-state leg adapter; it contains no explicit clock counter, seed-ID input, seed-fingerprint selector or trajectory lookup table.
+
+The graph retains the frozen [hc5 base](https://huggingface.co/suraj-phanindra/microduck-sprint-hc5-aifund/tree/c0c2897894cbab30a74a16e58cc93603dcfe581f), adds our trained residual, and adds a small physical-state adapter affecting knee/ankle outputs. The residual's training lineage used sanitized outputs from [game-sprinter-remix4](https://huggingface.co/jrubiosainz/microduck-game-sprinter-remix4/tree/175eba5581c9cbc038588ff1a1f406afc2e3a446). That teacher uses clock/seed-specific mechanisms; **those mechanisms are not present in this deployed graph**. Subsequent residual/adapter optimization used native Arena outcomes on the original five starts.
+
+Credit: **suraj-phanindra** (hc5), **juicenv** and **namenu** (base lineage), **jrubiosainz**, **dashi** and **Muhammadsafi** (teacher lineage), and **Pollen Robotics** (robot, training ecosystem and Arena).
+
+## Artifact and limitations
+
+`policy.onnx`: **1,137,169 bytes**  
+SHA256: `4aa8b27e4e02fbe9dcea1cae2d07f9505b51a62418dad853d2ff55e6a924a243`
+
+**Simulation only. No physical-robot validation or safety claim.** No training code or external runtime helpers are required to load this ONNX; no claim is made that all training code is published here.
+
+The inspected upstream model repositories do not declare standalone model licenses. This repository does **not** relicense inherited weights as MIT, Apache, or another open-source license. Review the linked upstream terms and [Hugging Face Terms of Service](https://huggingface.co/terms-of-service), including their public-repository provisions, before reuse. A general off-platform redistribution license is not asserted.