Publish two unqualified PiPER endpoint models and eight native development outcomes

LICENSE

@@ -0,0 +1,21 @@
+MIT License
+
+Copyright (c) 2026 DVIDIA contributors
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.

README.md

@@ -0,0 +1,75 @@
+---
+license: mit
+language:
+- en
+tags:
+- robotics
+- mujoco
+- imitation-learning
+- numpy
+- research-only
+---
+
+# DVIDIA PiPER reference actions v0.1
+
+Two compact CPU research candidates learned from 96,000 synchronized simulation demonstration rows. Both trained in about 5 seconds on the measured host and passed save/load replay checks. **Neither is a qualified skill: both achieved 0/4 strict placements in the bounded native development smoke.** Both candidates and negative results are retained.
+
+This repository contains model JSON files, cards and evidence for a **custom Python API**. It is not a Transformers model, a qualified-skill CLI installation package or a physical robot driver. No model was selected as a winner.
+
+| Measurement | Absolute reference (Cabs) | Prior-reference delta (Cdelta) |
+|---|---:|---:|
+| Training rows | 96,000 | 96,000 |
+| Fit time on measured shared host | 5.230713 s | 5.097628 s |
+| Model JSON size | 539,345 bytes | 538,021 bytes |
+| Saved-model replay parity | 1,000/1,000 outputs identical | 1,000/1,000 outputs identical |
+| Complete native development episodes | 4/4 | 4/4 |
+| Strict placements | **0/4** | **0/4** |
+| Bilateral grasp contact seen | 3/4 | 4/4 |
+| Qualified lift seen | 1/4 | 0/4 |
+| Carry completed | 0/4 | 0/4 |
+| Episodes with unexpected contacts | 2/4 | 4/4 |
+| Unexpected-contact native ticks | 7,606 | 19,751 |
+| Native callback including disk guard p95 | 0.295460 ms | 0.284297 ms |
+| Full offline boundary p95 | 22.362537 ms | 22.049043 ms |
+
+[Absolute model card](models/absolute/README.md) · [Delta model card](models/delta/README.md) · [Data card](data/README.md) · [Native audit](evidence/native-smoke-v3-audit.json) · [Failure diagnosis](evidence/native-smoke-diagnosis.json).
+
+## Load a candidate
+
+Install the source revision used for measurement and its pinned NumPy dependency in a virtual environment:
+
+```sh
+python -m pip install "git+https://github.com/Dvidia-Inference/dvidia-training.git@78ea4fbb4b2733e7738eb425eeecd7cb67a52195"
+```
+
+Download this repository's files and run Python from its root:
+
+```python
+from dvidia_training import camera_action_reference_model as reference_model
+
+model = reference_model.load_model("models/delta/model.json")
+assert model["task_qualified"] is False
+assert model["physical_robot_ready"] is False
+policy = reference_model.ModelPolicy(model)
+print(model["variant"], model["training"]["samples"])
+```
+
+This only loads and inspects the candidate; it starts no simulator or hardware. `policy(context)` expects the recorder's exact causal camera/encoder context, historical initial visual anchor, 160 ms history, availability masks and an actual retained float64 previous servo reference starting at the registered reset. An arbitrary image or joint vector does not satisfy that input contract. The source's new model format is not handled by the qualified-skill CLI installer.
+
+## What was measured
+
+Exactly two real training-data fits used the same 64,000 authored-teacher simulation rows plus 32,000 earlier mixed-control correction rows. Recipe: seed 17, two 64-unit tanh layers, 50 epochs, batch 512, Adam 0.003. Parent cache and exact prior-reference/feasible-target identities are bound in the [training input receipt](evidence/training-input-identity.json). Raw videos and derived training arrays are not redistributed in this compact model repository.
+
+Both candidates replayed 1,000 steps of the first **training** teacher trajectory. Predictions never fed into later replay observations or references; this is serialization/inference integration, not generalization. A separate native batch used the first two development groups, both goals and both unchanged candidates. All eight 20-second simulated episodes completed with no teacher intervention or fallback. No confirmation data or hardware was used.
+
+Neither candidate completed carry or strict placement. Seven of eight episodes never reached the required sustained bilateral table-free lift. All four delta trajectories crossed the strict scorer's unchanged finite TCP coordinate envelope of ±0.5 m; this was not a NaN or solver-warning diagnosis. Cabs visual estimates were available 59.9% of decisions and Cdelta 53.7%, with longest missing intervals 13.36 s and 6.16 s. These are observations from four cases per candidate, not a causal diagnosis or a statistical ranking.
+
+Sixteen earlier infrastructure refusals are preserved separately: eight at the wrapper's import guard, then eight at CoreGraphics initialization. A tiny static-frame preflight established scoped graphics access before the third batch. The source, models, roster, physics, guards and scorer remained unchanged across the two later batches. [All-batch accounting](evidence/development-smoke-status.json) and [graphics preflight](evidence/graphics-preflight.json) retain those boundaries. Do not pool infrastructure refusals into a 0/24 manipulation score.
+
+Measurements used a shared macOS 26.5.2 arm64 host, Python 3.12.9, NumPy 2.5.3, MuJoCo 3.15.0 and declared BLAS threads 1. Fit timing includes validation and normalization. Whole-process integration peak RSS was 1,345,404,928 bytes, not a model-only allocation or minimum hardware requirement. Callback timing excludes camera/rendering/physics/transport; the native callback includes its disk guard. Full boundary includes file writes and alternate camera work while the simulator pauses. Its 22 ms p95 is not a 50 Hz real-time guarantee. The renderer reported a depth-accuracy limitation due to unavailable `ARB_clip_control`; no backend depth calibration was established.
+
+These remain bounded PiPER block-placement research candidates. No arbitrary-video skill recovery, other-arm transfer, hardware qualification, release gate, negative-control suite or comparator superiority is claimed. The proposed 164-attempt screen remains unrun and frozen v0.1 remains 5/64 strict learned placements, unqualified. The [research report](https://research.dvidia.org/papers/reference-action-v01/) and [all eight recorded attempts](https://research.dvidia.org/downloads/topics/physical-grounding/runs/2026-11-09-reference-action-v01/viewer/index.html) are publicly available. Anonymous readback verified all 414 published report, diagnostic and viewer files against research commit `32fcceeb`.
+
+## License and provenance
+
+DVIDIA-generated weights, derived-data documentation and source use [MIT](LICENSE). Dependencies and upstream robot assets retain their own licenses and are not bundled here. No raw human footage or credentials are included. Candidate files are byte-identical to the retained training outputs; their file and canonical content hashes are listed in the model cards. Runtime source was [78ea4fbb](https://github.com/Dvidia-Inference/dvidia-training/tree/78ea4fbb4b2733e7738eb425eeecd7cb67a52195). The [source, model cards and benchmark evidence](https://github.com/Dvidia-Inference/dvidia-training/tree/8774265fdf2a4516520cdf36ff2eb41dcd04dfda/benchmarks/camera-action-reference-model-integration-v01) were published at `8774265f`; this documentation commit preserves the measured runtime above. The model's `scientific_registration` remains null; external receipts bind these operational experiments.

data/README.md

@@ -0,0 +1,26 @@
+---
+license: mit
+language:
+- en
+tags:
+- robotics
+- simulation
+- demonstrations
+- research
+---
+
+# DVIDIA PiPER reference training data
+
+This card describes the derived training inputs for both retained reference-model candidates. This data directory contains an identity receipt; the benchmark does not redistribute raw videos, caches or reference arrays. Both candidate model files are included separately.
+
+The data comprises 96,000 control rows from 96 existing simulation episodes in 32 training starting-position groups: 64 authored-teacher demonstrations covering both destinations, followed by 32 previously collected mixed-control corrections. These are synchronized simulated camera/action data, not recovered actions from human video. Development and confirmation episodes are excluded from fitting.
+
+Each parent row contains 269 causal features, the issued-teacher or separately recorded teacher-query target, group ownership and provenance. Each reference export adds two C-order float64 `(N, 7)` arrays: exact previous normal servo references and feasible target endpoints. Six dimensions use radians; the seventh is the single-finger jaw coordinate in metres. Reset references come from the recording reset; later references come from the preceding step's actual normal ramp state. The original 20 updates are applied to teacher labels. For mixed control this is counterfactual normal-reference supervision, not a statement that those labels were executed.
+
+The parent cache was built through the original strict exporter. The reference-export construction verified sealed recording metadata and caller-pinned parent data; cache/reference loads validate derived bytes without re-reading raw RGB-D frames. This integration experiment did not revalidate raw recordings. Array, protocol, stage, source, parent dataset and manifest identities are in [training-input-identity.json](../evidence/training-input-identity.json).
+
+Source: [DVIDIA training at the benchmark commit](https://github.com/Dvidia-Inference/dvidia-training/tree/78ea4fbb4b2733e7738eb425eeecd7cb67a52195). Training uses only training-data normalization. Current teacher actions, future observations and native object/contact truth are excluded from features. Missing visual observations remain masked, and historical estimates retain their age semantics.
+
+These bounded synthetic demonstrations do not establish transfer to shoes, arbitrary videos, other arms or physical hardware. Model training losses and a single training-trajectory replay do not establish generalization. The original dataset ownership and failure accounting are preserved.
+
+DVIDIA-generated derived data and accompanying documentation use [MIT](../LICENSE). Upstream simulator and robot asset licenses remain separate. No raw human footage, credentials or upstream model weights are included in this benchmark.

evidence/development-smoke-status.json

@@ -0,0 +1,362 @@
+{
+  "critical_contact_metrics": "See per-variant critical_contact_ticks and episodes_with_critical_contact in native-smoke-v3-audit.json.",
+  "evidence_state": "locally_validated_unqualified; remote hosting not asserted",
+  "funnel_metrics": {
+    "absolute": {
+      "callback_including_guard": {
+        "calls": 4000,
+        "maximum_seconds": 0.008621292014140636,
+        "median_seconds": 0.00023227051133289933,
+        "p95_seconds": 0.00029546011064667254
+      },
+      "carry_complete": 0,
+      "complete_episodes": 4,
+      "critical_contact_ticks": 7606,
+      "episodes_with_critical_contact": 2,
+      "full_boundary": {
+        "calls": 4000,
+        "maximum_seconds": 0.051054958981694654,
+        "median_seconds": 0.013028145520365797,
+        "p95_seconds": 0.022362537491426335
+      },
+      "grasp_seen": 3,
+      "lift_seen": 1,
+      "qualified": false,
+      "retention_lost": 0,
+      "scheduled": 4,
+      "scored_control_trajectories": 4,
+      "strict_successes": 0,
+      "terminal_only_successes": 0
+    },
+    "delta": {
+      "callback_including_guard": {
+        "calls": 4000,
+        "maximum_seconds": 0.021589000010862947,
+        "median_seconds": 0.00023095900542102754,
+        "p95_seconds": 0.0002842972404323518
+      },
+      "carry_complete": 0,
+      "complete_episodes": 4,
+      "critical_contact_ticks": 19751,
+      "episodes_with_critical_contact": 4,
+      "full_boundary": {
+        "calls": 4000,
+        "maximum_seconds": 0.058383833995321766,
+        "median_seconds": 0.013461853988701478,
+        "p95_seconds": 0.02204904312093276
+      },
+      "grasp_seen": 4,
+      "lift_seen": 0,
+      "qualified": false,
+      "retention_lost": 0,
+      "scheduled": 4,
+      "scored_control_trajectories": 4,
+      "strict_successes": 0,
+      "terminal_only_successes": 0
+    }
+  },
+  "independently_scored_control_trajectories": 8,
+  "initial_infrastructure_batches": [
+    {
+      "initialization_refusals": 8,
+      "plan_sha256": "3818be616fe37f976e2b31f86a21f5c429f11b4e4131a43d5174ed1bb771e289",
+      "policy_callbacks": 0,
+      "recorded_control_intervals": 0,
+      "result": {
+        "bytes": 6130,
+        "sha256": "2e9a40eaa5799f3bc6d0d36ee10f5fe8a6cf0bfe387f428f399d66a4e9c5c6fe"
+      },
+      "rows": [
+        {
+          "files_verified": 2,
+          "id": "development-00-goal0-absolute",
+          "recorded_complete_intervals": 0,
+          "rollout_error": {
+            "message": "No network or child process inside native worker.",
+            "type": "RuntimeError"
+          },
+          "startup_permission_closed": null,
+          "startup_probe_calls": 0,
+          "variant": "absolute",
+          "worker_receipt": {
+            "bytes": 2444,
+            "sha256": "059ef7b8732056b28e73b2dd50bd56dd95bbc22d9c633a9736609131f841ba42"
+          }
+        },
+        {
+          "files_verified": 2,
+          "id": "development-00-goal0-delta",
+          "recorded_complete_intervals": 0,
+          "rollout_error": {
+            "message": "No network or child process inside native worker.",
+            "type": "RuntimeError"
+          },
+          "startup_permission_closed": null,
+          "startup_probe_calls": 0,
+          "variant": "delta",
+          "worker_receipt": {
+            "bytes": 2435,
+            "sha256": "fb67f2ba4ed0a69f294541987288b40be51e7886302b976be43de547b5021dcc"
+          }
+        },
+        {
+          "files_verified": 2,
+          "id": "development-00-goal1-absolute",
+          "recorded_complete_intervals": 0,
+          "rollout_error": {
+            "message": "No network or child process inside native worker.",
+            "type": "RuntimeError"
+          },
+          "startup_permission_closed": null,
+          "startup_probe_calls": 0,
+          "variant": "absolute",
+          "worker_receipt": {
+            "bytes": 2443,
+            "sha256": "71285854d1e4f590a3b58fd4f97819ae46eb4d58d50d4932ddf34abed93884fd"
+          }
+        },
+        {
+          "files_verified": 2,
+          "id": "development-00-goal1-delta",
+          "recorded_complete_intervals": 0,
+          "rollout_error": {
+            "message": "No network or child process inside native worker.",
+            "type": "RuntimeError"
+          },
+          "startup_permission_closed": null,
+          "startup_probe_calls": 0,
+          "variant": "delta",
+          "worker_receipt": {
+            "bytes": 2434,
+            "sha256": "651a0ac311d20c08023a84945604f9591cacd12bfe7bc41fda763927ae4927c7"
+          }
+        },
+        {
+          "files_verified": 2,
+          "id": "development-01-goal0-absolute",
+          "recorded_complete_intervals": 0,
+          "rollout_error": {
+            "message": "No network or child process inside native worker.",
+            "type": "RuntimeError"
+          },
+          "startup_permission_closed": null,
+          "startup_probe_calls": 0,
+          "variant": "absolute",
+          "worker_receipt": {
+            "bytes": 2443,
+            "sha256": "fbc5c5419755eef28cd877949f534146c0594b65a6cfc44d784c21552c544516"
+          }
+        },
+        {
+          "files_verified": 2,
+          "id": "development-01-goal0-delta",
+          "recorded_complete_intervals": 0,
+          "rollout_error": {
+            "message": "No network or child process inside native worker.",
+            "type": "RuntimeError"
+          },
+          "startup_permission_closed": null,
+          "startup_probe_calls": 0,
+          "variant": "delta",
+          "worker_receipt": {
+            "bytes": 2434,
+            "sha256": "527b2af2befa2a1461db9d65b2730f790800937a3f97e2ae73f7b49c8f6c3214"
+          }
+        },
+        {
+          "files_verified": 2,
+          "id": "development-01-goal1-absolute",
+          "recorded_complete_intervals": 0,
+          "rollout_error": {
+            "message": "No network or child process inside native worker.",
+            "type": "RuntimeError"
+          },
+          "startup_permission_closed": null,
+          "startup_probe_calls": 0,
+          "variant": "absolute",
+          "worker_receipt": {
+            "bytes": 2443,
+            "sha256": "47c9609434b19a8c8da621cf0bb4a6c21fc64b1249bb26104a9f02669be40ff6"
+          }
+        },
+        {
+          "files_verified": 2,
+          "id": "development-01-goal1-delta",
+          "recorded_complete_intervals": 0,
+          "rollout_error": {
+            "message": "No network or child process inside native worker.",
+            "type": "RuntimeError"
+          },
+          "startup_permission_closed": null,
+          "startup_probe_calls": 0,
+          "variant": "delta",
+          "worker_receipt": {
+            "bytes": 2435,
+            "sha256": "9ae30452ca490e2adab099bc62206abdae3d029fbaa7c8c7e281ae20780a3ae8"
+          }
+        }
+      ],
+      "run_id": "reference-model-development-smoke-20261109",
+      "scheduled_attempts": 8
+    },
+    {
+      "initialization_refusals": 8,
+      "plan_sha256": "600b2129fc44c21e092390269e50ff43766262808b88add6d4718ea301c0d8e8",
+      "policy_callbacks": 0,
+      "recorded_control_intervals": 0,
+      "result": {
+        "bytes": 6664,
+        "sha256": "017d7ae05f64b8539d950ec00063345f30571e4a89086e06411b1b038c9a335c"
+      },
+      "rows": [
+        {
+          "files_verified": 2,
+          "id": "development-00-goal0-absolute",
+          "recorded_complete_intervals": 0,
+          "rollout_error": {
+            "message": "invalid CoreGraphics connection",
+            "type": "CGLError"
+          },
+          "startup_permission_closed": true,
+          "startup_probe_calls": 1,
+          "variant": "absolute",
+          "worker_receipt": {
+            "bytes": 2918,
+            "sha256": "d3f14709e112bec83b112bf5391f61aebe4277b39adb5956030ebb4a11d6add6"
+          }
+        },
+        {
+          "files_verified": 2,
+          "id": "development-00-goal0-delta",
+          "recorded_complete_intervals": 0,
+          "rollout_error": {
+            "message": "invalid CoreGraphics connection",
+            "type": "CGLError"
+          },
+          "startup_permission_closed": true,
+          "startup_probe_calls": 1,
+          "variant": "delta",
+          "worker_receipt": {
+            "bytes": 2910,
+            "sha256": "965fba1d8c1e06eef6694ef65ec4d87dbf58966013712cd9aa8e6a2e8a426b04"
+          }
+        },
+        {
+          "files_verified": 2,
+          "id": "development-00-goal1-absolute",
+          "recorded_complete_intervals": 0,
+          "rollout_error": {
+            "message": "invalid CoreGraphics connection",
+            "type": "CGLError"
+          },
+          "startup_permission_closed": true,
+          "startup_probe_calls": 1,
+          "variant": "absolute",
+          "worker_receipt": {
+            "bytes": 2917,
+            "sha256": "de91ed38603f1ca847de2fbf0416687e52d5c4084a9bdf2242d5726c600c5fa8"
+          }
+        },
+        {
+          "files_verified": 2,
+          "id": "development-00-goal1-delta",
+          "recorded_complete_intervals": 0,
+          "rollout_error": {
+            "message": "invalid CoreGraphics connection",
+            "type": "CGLError"
+          },
+          "startup_permission_closed": true,
+          "startup_probe_calls": 1,
+          "variant": "delta",
+          "worker_receipt": {
+            "bytes": 2909,
+            "sha256": "aa0c157fbc73e6b501d9e1085fa526e153a322144709c175de18d212ee9fd4a7"
+          }
+        },
+        {
+          "files_verified": 2,
+          "id": "development-01-goal0-absolute",
+          "recorded_complete_intervals": 0,
+          "rollout_error": {
+            "message": "invalid CoreGraphics connection",
+            "type": "CGLError"
+          },
+          "startup_permission_closed": true,
+          "startup_probe_calls": 1,
+          "variant": "absolute",
+          "worker_receipt": {
+            "bytes": 2919,
+            "sha256": "c0cc1bacebe5f93cd1392df3f59dd617a6bcefb01caa508ae04116eedd377306"
+          }
+        },
+        {
+          "files_verified": 2,
+          "id": "development-01-goal0-delta",
+          "recorded_complete_intervals": 0,
+          "rollout_error": {
+            "message": "invalid CoreGraphics connection",
+            "type": "CGLError"
+          },
+          "startup_permission_closed": true,
+          "startup_probe_calls": 1,
+          "variant": "delta",
+          "worker_receipt": {
+            "bytes": 2909,
+            "sha256": "39b0786aaaeff0455416928bb05bdb8e69738573d3250bd88ba7a1a065dff260"
+          }
+        },
+        {
+          "files_verified": 2,
+          "id": "development-01-goal1-absolute",
+          "recorded_complete_intervals": 0,
+          "rollout_error": {
+            "message": "invalid CoreGraphics connection",
+            "type": "CGLError"
+          },
+          "startup_permission_closed": true,
+          "startup_probe_calls": 1,
+          "variant": "absolute",
+          "worker_receipt": {
+            "bytes": 2918,
+            "sha256": "ed64a04c503121a0649365fd24aa63a6aa357ab83633743820f1b4352a69ab3c"
+          }
+        },
+        {
+          "files_verified": 2,
+          "id": "development-01-goal1-delta",
+          "recorded_complete_intervals": 0,
+          "rollout_error": {
+            "message": "invalid CoreGraphics connection",
+            "type": "CGLError"
+          },
+          "startup_permission_closed": true,
+          "startup_probe_calls": 1,
+          "variant": "delta",
+          "worker_receipt": {
+            "bytes": 2909,
+            "sha256": "885c5b7ec717d1b8790cec4299c1a29e52024abac6cbd70b8bb3c5a3bf059cc7"
+          }
+        }
+      ],
+      "run_id": "reference-model-development-smoke-v2-20261109",
+      "scheduled_attempts": 8
+    }
+  ],
+  "interpretation": "Report0/4 strict placements per model on four reused development scenarios. The prior16 infrastructure refusals are retained separately, not pooled into a0/24 manipulation score. This is not a generalization or comparator study.",
+  "learned_callback_latency": "See per-variant callback_including_guard distributions in native-smoke-v3-audit.json.",
+  "native_audit_sha256": "098c3c7b8c678790cc089dbec50d5c7c874e0a54da72844e48c49222683f9d7f",
+  "native_initialization_scope": "First eight import refusals and second eight graphics refusals remain separate. Third batch used separately registered scoped graphics access and completed all8 native learned-control attempts; neither model succeeded.",
+  "native_plan_sha256": "9912b3b86abe889d81073b1fee1c9cad9655264fdb325abbbf60dbebd89a490b",
+  "native_result_file": {
+    "bytes": 7153,
+    "sha256": "597277fd6c4671a7d2791aa861beb5da896573fcbf51f8db08e30d4bc501284e"
+  },
+  "next_prerequisite": "A separately registered failure-correction experiment; no automatic retries, tuning or new qualification.",
+  "policy_callbacks": 8000,
+  "qualified": false,
+  "recorded_control_intervals": 8000,
+  "review_receipt_sha256": "2270d1043e8dd68286627d15f2c4f039a7fdda9913d843ebe5e7551e1523367e",
+  "scheduled_attempts_across_batches": 24,
+  "schema": "dvidia.reference-model-development-smoke-status/v2",
+  "selection_performed": false
+}

evidence/graphics-preflight.json

@@ -0,0 +1,50 @@
+{
+  "results": {
+    "default-sandbox.json": {
+      "original_file": {
+        "bytes": 498,
+        "sha256": "74c96d8856d24c3f89680202b63ab67efe5a45b42a97e17ff44429d347827d7d"
+      },
+      "result": {
+        "error": "invalid CoreGraphics connection",
+        "error_type": "CGLError",
+        "execution_context": "default-sandbox",
+        "machine": "arm64",
+        "mujoco": "3.15.0",
+        "numpy": "2.5.3",
+        "python": "3.12.9",
+        "scope": "Tiny authored static XML, forward kinematics only, one48x64RGB frame; no robot/assets/dataset/task/model weights or dynamics steps.",
+        "status": "failed",
+        "total_wall_seconds": 0.12905491699348204
+      }
+    },
+    "scoped-graphics-access.json": {
+      "original_file": {
+        "bytes": 674,
+        "sha256": "dbfaef196312bcd905ad97288a63732a9b57bca5eb3a4119e03a885f9af3af55"
+      },
+      "result": {
+        "execution_context": "scoped-graphics-access",
+        "frame_dtype": "uint8",
+        "frame_max": 185,
+        "frame_min": 0,
+        "frame_sha256": "a1b17db25079e446d7030d4f72820efaf9c5163c0cdaccab6a7449cdee82916e",
+        "frame_shape": [
+          48,
+          64,
+          3
+        ],
+        "machine": "arm64",
+        "mujoco": "3.15.0",
+        "numpy": "2.5.3",
+        "python": "3.12.9",
+        "render_wall_seconds": 0.020181250001769513,
+        "scope": "Tiny authored static XML, forward kinematics only, one48x64RGB frame; no robot/assets/dataset/task/model weights or dynamics steps.",
+        "status": "passed",
+        "total_wall_seconds": 0.16178770799888298
+      }
+    }
+  },
+  "schema": "dvidia.reference-model-graphics-preflight-summary/v1",
+  "scope": "One static48x64frame, forward kinematics only; no robot/assets/data/model weights or dynamics. Interpreter path omitted from this public summary."
+}

evidence/initial-smoke-accounting.json

@@ -0,0 +1,14 @@
+{
+  "cause": "The experiment wrapper prohibited MuJoCo macOS import-time sysctl subprocess; each attempt stopped before policy callbacks.",
+  "interpretation": "Infrastructure refusal, not measured manipulation failure. All eight scheduled attempts remain accounted for.",
+  "native_evaluation": "No valid closed-loop result yet. Any further attempt needs separate registration and must preserve these results.",
+  "native_initialization_refusals": 8,
+  "original_result_sha256": "2e9a40eaa5799f3bc6d0d36ee10f5fe8a6cf0bfe387f428f399d66a4e9c5c6fe",
+  "plan_sha256": "3818be616fe37f976e2b31f86a21f5c429f11b4e4131a43d5174ed1bb771e289",
+  "policy_callbacks": 0,
+  "qualified": false,
+  "scheduled_attempts": 8,
+  "schema": "dvidia.reference-model-initial-smoke-accounting/v1",
+  "selection_performed": false,
+  "strict_scored_rollouts": 0
+}

evidence/integration-result.json

@@ -0,0 +1,217 @@
+{
+  "causal_replay_features_equal_parent": true,
+  "common_initialization_and_normalization": true,
+  "contract": {
+    "absolute_decoder": "r0 + D*g(x)",
+    "action_units": [
+      "rad",
+      "rad",
+      "rad",
+      "rad",
+      "rad",
+      "rad",
+      "single_finger_m"
+    ],
+    "adam_moments_dtype": "float32",
+    "blind": false,
+    "control_period_s": 0.02,
+    "decoder_loss_and_gradients_dtype": "float64",
+    "delta_decoder": "r + D*g(x)",
+    "feature_contract": "camera-encoder-absolute-causal160/v1",
+    "feature_contract_sha256": "e2d45f2cb03b39d424559c76f11252978facdf911af541e8e294b7bb8cb03fe8",
+    "hidden_units": [
+      64,
+      64
+    ],
+    "history_seconds": 0.16,
+    "increment_scale": [
+      0.03,
+      0.03,
+      0.03,
+      0.03,
+      0.03,
+      0.03,
+      0.001
+    ],
+    "joint_order": [
+      "joint1",
+      "joint2",
+      "joint3",
+      "joint4",
+      "joint5",
+      "joint6",
+      "single_finger_jaw"
+    ],
+    "loss": "mean(((unprojected_endpoint-feasible_target)/S)**2)",
+    "loss_scale": [
+      3.0,
+      3.0,
+      3.0,
+      3.0,
+      3.0,
+      3.0,
+      0.035
+    ],
+    "network_dtype": "float32",
+    "projection": "native bounds then exact original 20 normal reference updates",
+    "reference_kind": "previous_applied_normal_servo_reference",
+    "reset_reference": [
+      0.0,
+      1.57,
+      -1.3485,
+      0.0,
+      0.0,
+      0.0,
+      0.035
+    ]
+  },
+  "evidence_state": "locally_validated_unqualified; remote hosting not asserted",
+  "feasible_endpoint_array_sha256": "d5754ace901e4d0fda4793be8e34aa9676009e8d700451937c439bd42c1d0163",
+  "feature_array_sha256": "8e2399f0753e20e86a50555dcf17a85c12b7c0b2583842f13da14a0bad393bee",
+  "fits": 2,
+  "forbidden_extra_input_invariance": true,
+  "hardware": {
+    "blas_threads_declared": 1,
+    "cpu_count": 10,
+    "machine": "arm64",
+    "numpy": "2.5.3",
+    "platform": "macOS-26.5.2-arm64-arm-64bit",
+    "processor": "arm",
+    "python": "3.12.9",
+    "shared_host": true
+  },
+  "held_out_rows_read": 0,
+  "limitations": [
+    "Training errors and one teacher-trajectory open-loop replay are not generalization or task success.",
+    "No controller was actuated, selected, qualified, installed or evaluated in a new environment.",
+    "Both candidates retained; frozenv01 results and unrun164-attemptv02 screen unchanged.",
+    "Registration binds this benchmark externally; model scientific_registration intentionally remains null."
+  ],
+  "native_rollouts": 0,
+  "original_metadata_files_read_count": 1003,
+  "original_result_sha256": "6d44d98488ef26f949412aadc94c32174bd5c4ce879442191b9784deff5b325f",
+  "peak_whole_process_rss_bytes": 1345404928,
+  "physical_robot_ready": false,
+  "previous_reference_array_sha256": "e2f84e643ad7e9d04fd2d993051f6873283a376c327838e2ee84733880c8af54",
+  "qualified": false,
+  "raw_recordings_revalidated": false,
+  "recipe": {
+    "batch_size": 512,
+    "beta1": 0.9,
+    "beta2": 0.999,
+    "epochs": 50,
+    "epsilon": 1e-08,
+    "gradient_clip": 1.0,
+    "learning_rate": 0.003,
+    "parameter_clip": 64.0,
+    "seed": 17
+  },
+  "registered_plan_sha256": "ffb52ed4226d48482010be7dc2a1acf2fa385202a0d7dc142ca3e8904cb7d4f8",
+  "replay_frames": 1000,
+  "replay_next_state_origin": "recorded teacher trajectory only",
+  "replay_predictions_actuated": false,
+  "resource_scope": "Single compute worker plus250ms resource observer. RSS/deadline/free-space stop thresholds are sampled controls, not kernel memory quotas or protection against unrelated disk writers.",
+  "schema": "dvidia.reference-model-integration-publication-summary/v1",
+  "script_sha256": "1cc8f9a82d7381ba56db401c9ea89246b67df4d281e930fc664001b888754a69",
+  "selection_performed": false,
+  "source_commit": "78ea4fbb4b2733e7738eb425eeecd7cb67a52195",
+  "source_identity_sha256": "12ea9757cb0d76573c2e67f80b56ab08271f95e43f18b344b6481cc9153d70bc",
+  "status": "complete",
+  "timing_scope": "Fit includes source/derived-input validation and normalization. Callback timing includes causal encoding, NumPy inference and reference projection; excludes camera/render/physics/transport. RSS is whole-process high-water mark, not an isolated model allocation.",
+  "total_wall_seconds": 12.93291958401096,
+  "training_rows_per_fit": 96000,
+  "variants": {
+    "absolute": {
+      "final_training_loss": 7.926108883701983e-06,
+      "fit_wall_seconds": 5.230713041994022,
+      "initial_training_loss": 0.04009072663323015,
+      "loaded_open_loop_callback_wall": {
+        "calls": 1000,
+        "maximum_seconds": 0.0003341670089866966,
+        "median_seconds": 0.00018152050324715674,
+        "p95_seconds": 0.0002062101018964313,
+        "sum_seconds": 0.18355957081075758
+      },
+      "model_content_sha256": "36af3fe1670494b04c48b5114d9be1a5f91632aa7992ba529d6587969ea37460",
+      "model_file": {
+        "bytes": 539345,
+        "sha256": "3dbf6911d92a9072297c9c3bb7c09f45809808272455ffd5562f4553a173f362"
+      },
+      "model_save_load_identity": true,
+      "open_loop_callback_wall": {
+        "calls": 1000,
+        "maximum_seconds": 0.00027541699819266796,
+        "median_seconds": 0.0001818129967432469,
+        "p95_seconds": 0.00020484020933508873,
+        "sum_seconds": 0.18381304285139777
+      },
+      "open_loop_prediction_file": {
+        "bytes": 56128,
+        "sha256": "5dfbcebbb33a0a628b9e82ff5508097b7ec7b190566b2979a5f20255943dc914"
+      },
+      "open_loop_saved_policy_outputs_bitwise_equal": true,
+      "peak_whole_process_rss_bytes_so_far": 1334378496,
+      "qualified": false,
+      "training": {
+        "algorithm": "numpy-tanh-2x64-adam-native-endpoint-mse/v1",
+        "final_loss": 7.926108883701983e-06,
+        "group_count": 32,
+        "initial_loss": 0.04009072663323015,
+        "initial_parameters_sha256": "9c50504160213ad5241d6d96ab9be938d6f49b1c4bf49ff39d0770dac83dce7e",
+        "normalization": "training features only; no label whitening",
+        "optimizer_updates": 9400,
+        "parent_dataset_sha256": "b500b3ba348bba44cc79e377266f92910bd7e0165c36cd8b2fe260909b30a2f9",
+        "parent_provenance_sha256": "52cbeda62e3aa50e8cd6a34688fba7583f9fcec927cb0eb434a6907c99695f4f",
+        "reference_dataset_sha256": "a9dcd66a81f177bb1910e08171ac7607f115ed47b51d93412107eb0e204aed73",
+        "samples": 96000
+      },
+      "training_trajectory_projected_endpoint_scaled_mse": 1.8801026187180245e-06
+    },
+    "delta": {
+      "final_training_loss": 3.531327787737655e-06,
+      "fit_wall_seconds": 5.097628457995597,
+      "initial_training_loss": 8.72223173973333e-05,
+      "loaded_open_loop_callback_wall": {
+        "calls": 1000,
+        "maximum_seconds": 0.00024516700068488717,
+        "median_seconds": 0.0001821249898057431,
+        "p95_seconds": 0.00020668574870796871,
+        "sum_seconds": 0.18321760100661777
+      },
+      "model_content_sha256": "4545be31f50267c9da29b211510c7a32df500678ba107c7a100604bb0c4c18ab",
+      "model_file": {
+        "bytes": 538021,
+        "sha256": "fece6d056e50a1bde41bdb9960bf7cbc81d983c4bfdb03a07632e30b712b1489"
+      },
+      "model_save_load_identity": true,
+      "open_loop_callback_wall": {
+        "calls": 1000,
+        "maximum_seconds": 0.0002842499816324562,
+        "median_seconds": 0.00018272899615112692,
+        "p95_seconds": 0.00020738332968903706,
+        "sum_seconds": 0.18398747508763336
+      },
+      "open_loop_prediction_file": {
+        "bytes": 56128,
+        "sha256": "a741d006667a559c846c4b2cf19c88ea81ea4eaa55d75c53583e415e1f0dd665"
+      },
+      "open_loop_saved_policy_outputs_bitwise_equal": true,
+      "peak_whole_process_rss_bytes_so_far": 1345404928,
+      "qualified": false,
+      "training": {
+        "algorithm": "numpy-tanh-2x64-adam-native-endpoint-mse/v1",
+        "final_loss": 3.531327787737655e-06,
+        "group_count": 32,
+        "initial_loss": 8.72223173973333e-05,
+        "initial_parameters_sha256": "9c50504160213ad5241d6d96ab9be938d6f49b1c4bf49ff39d0770dac83dce7e",
+        "normalization": "training features only; no label whitening",
+        "optimizer_updates": 9400,
+        "parent_dataset_sha256": "b500b3ba348bba44cc79e377266f92910bd7e0165c36cd8b2fe260909b30a2f9",
+        "parent_provenance_sha256": "52cbeda62e3aa50e8cd6a34688fba7583f9fcec927cb0eb434a6907c99695f4f",
+        "reference_dataset_sha256": "a9dcd66a81f177bb1910e08171ac7607f115ed47b51d93412107eb0e204aed73",
+        "samples": 96000
+      },
+      "training_trajectory_projected_endpoint_scaled_mse": 6.222337738975971e-07
+    }
+  }
+}

evidence/native-smoke-diagnosis.json

@@ -0,0 +1,191 @@
+{
+  "delta_strict_violations": [
+    {
+      "all_strict_vector_fields_finite": true,
+      "first_strict_state_envelope_violation": {
+        "object_position_m": [
+          0.4871081710122422,
+          -0.17988504093974045,
+          0.09353123557334052
+        ],
+        "state_index": 739,
+        "t_seconds": 14.779999999997248,
+        "table_object_contact_count": 1,
+        "violations": [
+          {
+            "allowed_inclusive": [
+              -0.5,
+              0.5
+            ],
+            "axis": "x",
+            "component": 0,
+            "field": "tcp_position_m",
+            "finite": true,
+            "value": 0.5018659985967776
+          }
+        ]
+      },
+      "id": "development-00-goal0-delta",
+      "integrity": {
+        "bytes": 567674,
+        "sha256": "d4a89f2555c2e9d86c745fbc0b20be34b39ce3fb2400d037a91c63a6d7d0cd1a"
+      },
+      "native_episode": {
+        "bytes": 3274472,
+        "sha256": "3f91710eb25b61a0fb6c64fa050cc8406f8d11b2eeff3a65dee1502eb2c5fd8f"
+      },
+      "physics_warnings": []
+    },
+    {
+      "all_strict_vector_fields_finite": true,
+      "first_strict_state_envelope_violation": {
+        "object_position_m": [
+          0.32931879716563506,
+          -0.12798862211401657,
+          0.0798922445797837
+        ],
+        "state_index": 713,
+        "t_seconds": 14.259999999997536,
+        "table_object_contact_count": 4,
+        "violations": [
+          {
+            "allowed_inclusive": [
+              -0.5,
+              0.5
+            ],
+            "axis": "x",
+            "component": 0,
+            "field": "tcp_position_m",
+            "finite": true,
+            "value": 0.5033371506888265
+          }
+        ]
+      },
+      "id": "development-00-goal1-delta",
+      "integrity": {
+        "bytes": 567674,
+        "sha256": "d9a2f3b46a3906153b5ee703d915c809b9d5903996abbc948a31ab42fae1e1dd"
+      },
+      "native_episode": {
+        "bytes": 3199780,
+        "sha256": "c38becb927b94c6c7478c76253f86ee4a656f8d8ab1c482258fdfb5ef4dfa8b2"
+      },
+      "physics_warnings": []
+    },
+    {
+      "all_strict_vector_fields_finite": true,
+      "first_strict_state_envelope_violation": {
+        "object_position_m": [
+          0.36982822089212286,
+          -0.17378339194526918,
+          0.0798922445797837
+        ],
+        "state_index": 687,
+        "t_seconds": 13.739999999997824,
+        "table_object_contact_count": 4,
+        "violations": [
+          {
+            "allowed_inclusive": [
+              -0.5,
+              0.5
+            ],
+            "axis": "z",
+            "component": 2,
+            "field": "tcp_position_m",
+            "finite": true,
+            "value": 0.5049848841298252
+          }
+        ]
+      },
+      "id": "development-01-goal0-delta",
+      "integrity": {
+        "bytes": 567674,
+        "sha256": "1f4b2db19b5c0ded4438d9cc549a65e6a25a9f1691118a4363f93bf0646e2a89"
+      },
+      "native_episode": {
+        "bytes": 3081538,
+        "sha256": "51b909802d8da08157c87cc1d1dffac6eea4b28a8f9448669b5cc02797e961ae"
+      },
+      "physics_warnings": []
+    },
+    {
+      "all_strict_vector_fields_finite": true,
+      "first_strict_state_envelope_violation": {
+        "object_position_m": [
+          0.4070060436188808,
+          -0.18591757382039348,
+          0.0798922445794117
+        ],
+        "state_index": 735,
+        "t_seconds": 14.699999999997292,
+        "table_object_contact_count": 4,
+        "violations": [
+          {
+            "allowed_inclusive": [
+              -0.5,
+              0.5
+            ],
+            "axis": "x",
+            "component": 0,
+            "field": "tcp_position_m",
+            "finite": true,
+            "value": 0.5001573804075903
+          }
+        ]
+      },
+      "id": "development-01-goal1-delta",
+      "integrity": {
+        "bytes": 567674,
+        "sha256": "99a748732c1cbb7761ec00b786a680da20f01b0abb038214451be42d069cd8e2"
+      },
+      "native_episode": {
+        "bytes": 3104558,
+        "sha256": "42e5e034ee6c8a6d4cf58cdf7efe2d7545f4554d2e0253b6eca42d477e623da0"
+      },
+      "physics_warnings": []
+    }
+  ],
+  "interpretation": {
+    "causality": "Own-state reference drift is a candidate mechanism; this read-only smoke does not establish its cause or isolate a treatment. No tuning or model ranking conclusion from four cases per variant.",
+    "contact_scope": "Critical ticks count 1ms simulated timesteps with at least one critical contact, not independent collision events. No real hardware force inference.",
+    "delta_state_failure": "All four delta trajectories contain finite TCP coordinates outside the unchanged strict scorer component envelope [-0.5,0.5] metres. Generic invalid_state_value is not a NaN or solver-warning diagnosis.",
+    "dominant_failure": "Seven cases never achieved the required sustained bilateral table-free lift; one absolute case lifted but did not complete supported placement, release or retreat by the 20-second budget. All eight failed retained carry completion.",
+    "floor_case": "development-00-goal0-delta first fails on TCP x while block still has one table contact. It later ends with four explicitly mapped block-floor contacts, no table support, and block centre z=0.019892m.",
+    "prior_batches": "Two distinct eight-attempt initialization-refusal batches are preserved and linked from the v3 result. Those had zero controller calls and are not control performance evaluations.",
+    "timing_scope": "Model diagnostic includes causal feature encoding, NumPy inference and feasible-reference projection. Pooled callback includes separate cooperative disk guard. Full boundary includes captured simulation and measurement work. Rendering estimates are offline, with depth-accuracy warning; no physical latency claim.",
+    "valid_semantics": "Outer valid checks native arrays, recomputed metrics and warning-free native physics; success additionally requires unchanged strict_v01 and native task success. Valid is not a task, workspace or safety qualification."
+  },
+  "model_selection": false,
+  "new_native_executions": 0,
+  "new_training_fits": 0,
+  "original_analysis_file": {
+    "bytes": 51280,
+    "sha256": "7ea36145088fda3e6e39a280c8e537d8ed0a1acab9503f1604ff812f1e971373"
+  },
+  "qualified": false,
+  "result": {
+    "bytes": 7153,
+    "sha256": "597277fd6c4671a7d2791aa861beb5da896573fcbf51f8db08e30d4bc501284e"
+  },
+  "schema": "dvidia.reference-model-smoke-diagnosis-summary/v1",
+  "scope": "Compact read-only diagnosis derived from the separately retained full analysis, not new simulation or a causal treatment experiment.",
+  "source_bindings": {
+    "camera_action_reference_model.py": {
+      "bytes": 19771,
+      "sha256": "952750fd6b99f97eb96054b8fe50710b83f7e8eb4e069601be442292dd0023bc"
+    },
+    "piper_feedback_score.py": {
+      "bytes": 18866,
+      "sha256": "863a272828208c631de4cf57647e973eb3e4f5e99eb7a1eacb4eaa68dc42eb0e"
+    },
+    "piper_imitation_score.py": {
+      "bytes": 11462,
+      "sha256": "0cd4663abfdd7f14f2f33d93e9a2bd45122b998d6160d45b8afcc880512cba2b"
+    },
+    "piper_task.py": {
+      "bytes": 34355,
+      "sha256": "e66a44a8df4e7a8ee42e4f3866f889f8797cdfb8e2688a146e09b75d2c76664f"
+    }
+  },
+  "source_commit": "78ea4fbb4b2733e7738eb425eeecd7cb67a52195"
+}

evidence/native-smoke-v3-audit.json

@@ -0,0 +1,2238 @@
+{
+  "attempts": [
+    {
+      "camera_pairs": 501,
+      "case": {
+        "goal_id": 0,
+        "goal_position_m": [
+          0.35,
+          0.08,
+          0.08
+        ],
+        "group_id": "development-00",
+        "id": "development-00-goal0",
+        "object_position_m": [
+          0.303692,
+          -0.033335,
+          0.082
+        ],
+        "seed": 311465,
+        "split": "development"
+      },
+      "files_verified": 5010,
+      "frame_bytes": 168843964,
+      "guard": {
+        "checks": 1000,
+        "forwarded_calls": 1000,
+        "guard_wall_seconds": 0.00523591969977133,
+        "minimum_free_bytes": 40582459392,
+        "scope": "Once per simulated20ms callback, not a wall-clock50Hz monitor; no task features changed.",
+        "stop_reason": null
+      },
+      "id": "development-00-goal0-absolute",
+      "inspection": {
+        "files_verified": 5010,
+        "physical_robot_ready": false,
+        "physics_executed": false,
+        "recorded_complete_intervals": 1000,
+        "schema": "dvidia.camera-action-rollout/v2/inspection",
+        "score": {
+          "evidence_scope": "Recomputed simulator-authored numeric evidence; no physics rerun, source attestation or hardware sensing.",
+          "false_alarms": {
+            "reason": "Continuous residual policy has no declared discrete alarm events.",
+            "status": "not_applicable"
+          },
+          "grip_events": [
+            {
+              "kind": "supported_grip_interruption",
+              "loss_onset_s": 12.229999999998661,
+              "loss_qualified_s": 12.24899999999865,
+              "onset_to_qualified_restoration_s": null,
+              "onset_to_restoration_s": null,
+              "relative_motion_since_loss_m": 0.037587862924821816,
+              "restoration_onset_s": null,
+              "restoration_qualified_s": null,
+              "restored": false,
+              "supported_at_onset": true
+            }
+          ],
+          "grip_summary": {
+            "airborne_events": 0,
+            "loss_events": 1,
+            "qualification_latency_s": [],
+            "restored_events": 0,
+            "supported_events": 1,
+            "unresolved_events": 1
+          },
+          "native_diagnostics": {
+            "premature_open_jaw": {
+              "count": 0,
+              "intervals": [],
+              "maximum_s": 0.0,
+              "total_s": 0.0
+            },
+            "support_touchdown_before_destination": {
+              "count": 0,
+              "intervals": [],
+              "maximum_s": 0.0,
+              "total_s": 0.0
+            }
+          },
+          "native_metrics_mismatches": [],
+          "native_recomputed": {
+            "carry_complete": false,
+            "carry_contact_loss_events": 0,
+            "dwell_seconds": 0.0,
+            "first_success_time_seconds": null,
+            "grasp_seen": true,
+            "lift_dwell_seconds": 0.0,
+            "lift_seen": false,
+            "max_carry_contact_loss_seconds": 0.0,
+            "max_object_lift_m": 0.003952023743962793,
+            "max_pad_normal_force_n": 2.795753702369255,
+            "retention_lost": false,
+            "success": false,
+            "unexpected_contact_native_ticks": 7454
+          },
+          "reason": "native_outcome_failed",
+          "restoration_scope": "Loaded contact restored for 100 ms after sustained loss; does not establish task recovery or airborne regrasp.",
+          "schema": "dvidia.piper-feedback-score/v1",
+          "strict_v01": {
+            "native_metrics_source": "simulator-authored episode record; not authenticated sensor input",
+            "native_outcome_success": false,
+            "native_outcome_valid": true,
+            "reason": "native_outcome_failed",
+            "reference_xy_m": [
+              0.35,
+              0.08
+            ],
+            "required_dwell_s": 0.25,
+            "required_lift_dwell_s": 0.1,
+            "sample_interval_s": 0.02,
+            "success": false,
+            "terminal_dwell_s": 0.0,
+            "terminal_reference_success": false,
+            "terminal_xy_error_m": 0.13859914919704855,
+            "trace_lift_dwell_s": 0.0,
+            "trajectory_complete": true
+          },
+          "success": false,
+          "terminal_only": {
+            "dwell_s": 0.0,
+            "scope": "Terminal-only geometry/release/settling/dwell; excludes lift, carry and retention history.",
+            "success": false,
+            "terminal_xy_error_m": 0.13859914919704855
+          },
+          "valid": true
+        },
+        "source_authenticated": false,
+        "status": "complete",
+        "strict_success": false
+      },
+      "integrity_file": {
+        "bytes": 567674,
+        "sha256": "714cc6d64114af14d9d75967c57f7f977ed0c40df9ea87e601f3836eee99990a"
+      },
+      "perception": {
+        "available": 384,
+        "decisions": 1000,
+        "longest_missing_seconds": 12.120000000000001,
+        "missing": 616
+      },
+      "policy": {
+        "model_sha256": "36af3fe1670494b04c48b5114d9be1a5f91632aa7992ba529d6587969ea37460",
+        "physical_robot_ready": false,
+        "policy_origin": "camera_encoder_reference_endpoint_numpy_mlp",
+        "runtime_teacher_available": false,
+        "student_inference_calls": 1000,
+        "student_inference_p95_seconds": 0.00026722680195234716,
+        "student_inference_seconds": 0.23363784674438648,
+        "task_qualified": false,
+        "teacher_fallback_calls": 0,
+        "timing_scope": "Causal feature encoding, NumPy inference and reference projection; excludes camera/render/actuation.",
+        "variant": "absolute"
+      },
+      "qualified": false,
+      "recorded_complete_intervals": 1000,
+      "rollout_error": null,
+      "rollout_file": {
+        "bytes": 326965,
+        "sha256": "1104da3991eb6c3872c85c46eec6d4a5dcb89c0a10752d57c5ebc8071b14dee9"
+      },
+      "rollout_status": "complete",
+      "rollout_wall_seconds": 12.169444583007134,
+      "score": {
+        "evidence_scope": "Recomputed simulator-authored numeric evidence; no physics rerun, source attestation or hardware sensing.",
+        "false_alarms": {
+          "reason": "Continuous residual policy has no declared discrete alarm events.",
+          "status": "not_applicable"
+        },
+        "grip_events": [
+          {
+            "kind": "supported_grip_interruption",
+            "loss_onset_s": 12.229999999998661,
+            "loss_qualified_s": 12.24899999999865,
+            "onset_to_qualified_restoration_s": null,
+            "onset_to_restoration_s": null,
+            "relative_motion_since_loss_m": 0.037587862924821816,
+            "restoration_onset_s": null,
+            "restoration_qualified_s": null,
+            "restored": false,
+            "supported_at_onset": true
+          }
+        ],
+        "grip_summary": {
+          "airborne_events": 0,
+          "loss_events": 1,
+          "qualification_latency_s": [],
+          "restored_events": 0,
+          "supported_events": 1,
+          "unresolved_events": 1
+        },
+        "native_diagnostics": {
+          "premature_open_jaw": {
+            "count": 0,
+            "intervals": [],
+            "maximum_s": 0.0,
+            "total_s": 0.0
+          },
+          "support_touchdown_before_destination": {
+            "count": 0,
+            "intervals": [],
+            "maximum_s": 0.0,
+            "total_s": 0.0
+          }
+        },
+        "native_metrics_mismatches": [],
+        "native_recomputed": {
+          "carry_complete": false,
+          "carry_contact_loss_events": 0,
+          "dwell_seconds": 0.0,
+          "first_success_time_seconds": null,
+          "grasp_seen": true,
+          "lift_dwell_seconds": 0.0,
+          "lift_seen": false,
+          "max_carry_contact_loss_seconds": 0.0,
+          "max_object_lift_m": 0.003952023743962793,
+          "max_pad_normal_force_n": 2.795753702369255,
+          "retention_lost": false,
+          "success": false,
+          "unexpected_contact_native_ticks": 7454
+        },
+        "reason": "native_outcome_failed",
+        "restoration_scope": "Loaded contact restored for 100 ms after sustained loss; does not establish task recovery or airborne regrasp.",
+        "schema": "dvidia.piper-feedback-score/v1",
+        "strict_v01": {
+          "native_metrics_source": "simulator-authored episode record; not authenticated sensor input",
+          "native_outcome_success": false,
+          "native_outcome_valid": true,
+          "reason": "native_outcome_failed",
+          "reference_xy_m": [
+            0.35,
+            0.08
+          ],
+          "required_dwell_s": 0.25,
+          "required_lift_dwell_s": 0.1,
+          "sample_interval_s": 0.02,
+          "success": false,
+          "terminal_dwell_s": 0.0,
+          "terminal_reference_success": false,
+          "terminal_xy_error_m": 0.13859914919704855,
+          "trace_lift_dwell_s": 0.0,
+          "trajectory_complete": true
+        },
+        "success": false,
+        "terminal_only": {
+          "dwell_s": 0.0,
+          "scope": "Terminal-only geometry/release/settling/dwell; excludes lift, carry and retention history.",
+          "success": false,
+          "terminal_xy_error_m": 0.13859914919704855
+        },
+        "valid": true
+      },
+      "scored_control_trajectory": true,
+      "strict_success": false,
+      "supervisor_cause": null,
+      "supervisor_status": "finished",
+      "variant": "absolute",
+      "worker_receipt": {
+        "bytes": 5163,
+        "sha256": "f3907fe87706a0f8f4593fe2e790c81ebc8018c185fa6ed17cd46f63c441019b"
+      },
+      "worker_receipt_available": true,
+      "worker_wall_seconds": 13.846535541000776
+    },
+    {
+      "camera_pairs": 501,
+      "case": {
+        "goal_id": 0,
+        "goal_position_m": [
+          0.35,
+          0.08,
+          0.08
+        ],
+        "group_id": "development-00",
+        "id": "development-00-goal0",
+        "object_position_m": [
+          0.303692,
+          -0.033335,
+          0.082
+        ],
+        "seed": 311465,
+        "split": "development"
+      },
+      "files_verified": 5010,
+      "frame_bytes": 168625930,
+      "guard": {
+        "checks": 1000,
+        "forwarded_calls": 1000,
+        "guard_wall_seconds": 0.005588873696979135,
+        "minimum_free_bytes": 40378089472,
+        "scope": "Once per simulated20ms callback, not a wall-clock50Hz monitor; no task features changed.",
+        "stop_reason": null
+      },
+      "id": "development-00-goal0-delta",
+      "inspection": {
+        "files_verified": 5010,
+        "physical_robot_ready": false,
+        "physics_executed": false,
+        "recorded_complete_intervals": 1000,
+        "schema": "dvidia.camera-action-rollout/v2/inspection",
+        "score": {
+          "evidence_scope": "Recomputed simulator-authored numeric evidence; no physics rerun, source attestation or hardware sensing.",
+          "false_alarms": {
+            "reason": "Continuous residual policy has no declared discrete alarm events.",
+            "status": "not_applicable"
+          },
+          "grip_events": [
+            {
+              "kind": "supported_grip_interruption",
+              "loss_onset_s": 12.32299999999861,
+              "loss_qualified_s": 12.341999999998599,
+              "onset_to_qualified_restoration_s": 0.4259999999997639,
+              "onset_to_restoration_s": 0.32699999999981877,
+              "relative_motion_since_loss_m": 0.033405944668013256,
+              "restoration_onset_s": 12.649999999998428,
+              "restoration_qualified_s": 12.748999999998373,
+              "restoration_supported": false,
+              "restored": true,
+              "supported_at_onset": true
+            },
+            {
+              "kind": "supported_grip_interruption",
+              "loss_onset_s": 13.181999999998133,
+              "loss_qualified_s": 13.200999999998123,
+              "onset_to_qualified_restoration_s": 0.21699999999987973,
+              "onset_to_restoration_s": 0.1179999999999346,
+              "relative_motion_since_loss_m": 0.0201096558077051,
+              "restoration_onset_s": 13.299999999998068,
+              "restoration_qualified_s": 13.398999999998013,
+              "restoration_supported": true,
+              "restored": true,
+              "supported_at_onset": true
+            },
+            {
+              "kind": "supported_grip_interruption",
+              "loss_onset_s": 17.108999999997927,
+              "loss_qualified_s": 17.12799999999795,
+              "onset_to_qualified_restoration_s": null,
+              "onset_to_restoration_s": null,
+              "relative_motion_since_loss_m": 0.2010864554938082,
+              "restoration_onset_s": null,
+              "restoration_qualified_s": null,
+              "restored": false,
+              "supported_at_onset": true
+            }
+          ],
+          "grip_summary": {
+            "airborne_events": 0,
+            "loss_events": 3,
+            "qualification_latency_s": [
+              0.4259999999997639,
+              0.21699999999987973
+            ],
+            "restored_events": 2,
+            "supported_events": 3,
+            "unresolved_events": 1
+          },
+          "native_diagnostics": {
+            "premature_open_jaw": {
+              "count": 0,
+              "intervals": [],
+              "maximum_s": 0.0,
+              "total_s": 0.0
+            },
+            "support_touchdown_before_destination": {
+              "count": 0,
+              "intervals": [],
+              "maximum_s": 0.0,
+              "total_s": 0.0
+            }
+          },
+          "native_metrics_mismatches": [],
+          "native_recomputed": {
+            "carry_complete": false,
+            "carry_contact_loss_events": 0,
+            "dwell_seconds": 0.0,
+            "first_success_time_seconds": null,
+            "grasp_seen": true,
+            "lift_dwell_seconds": 0.0,
+            "lift_seen": false,
+            "max_carry_contact_loss_seconds": 0.0,
+            "max_object_lift_m": 0.032622070113360255,
+            "max_pad_normal_force_n": 16.822725404528523,
+            "retention_lost": false,
+            "success": false,
+            "unexpected_contact_native_ticks": 7280
+          },
+          "reason": "invalid_state_value",
+          "restoration_scope": "Loaded contact restored for 100 ms after sustained loss; does not establish task recovery or airborne regrasp.",
+          "schema": "dvidia.piper-feedback-score/v1",
+          "strict_v01": {
+            "reason": "invalid_state_value",
+            "reference_xy_m": [
+              0.35,
+              0.08
+            ],
+            "state_index": 739,
+            "success": false
+          },
+          "success": false,
+          "terminal_only": {
+            "dwell_s": 0.0,
+            "scope": "Terminal-only geometry/release/settling/dwell; excludes lift, carry and retention history.",
+            "success": false,
+            "terminal_xy_error_m": 0.35894977561012875
+          },
+          "valid": true
+        },
+        "source_authenticated": false,
+        "status": "complete",
+        "strict_success": false
+      },
+      "integrity_file": {
+        "bytes": 567674,
+        "sha256": "d4a89f2555c2e9d86c745fbc0b20be34b39ce3fb2400d037a91c63a6d7d0cd1a"
+      },
+      "perception": {
+        "available": 472,
+        "decisions": 1000,
+        "longest_missing_seconds": 6.16,
+        "missing": 528
+      },
+      "policy": {
+        "model_sha256": "4545be31f50267c9da29b211510c7a32df500678ba107c7a100604bb0c4c18ab",
+        "physical_robot_ready": false,
+        "policy_origin": "camera_encoder_reference_endpoint_numpy_mlp",
+        "runtime_teacher_available": false,
+        "student_inference_calls": 1000,
+        "student_inference_p95_seconds": 0.00028345415339572346,
+        "student_inference_seconds": 0.2566282909247093,
+        "task_qualified": false,
+        "teacher_fallback_calls": 0,
+        "timing_scope": "Causal feature encoding, NumPy inference and reference projection; excludes camera/render/actuation.",
+        "variant": "delta"
+      },
+      "qualified": false,
+      "recorded_complete_intervals": 1000,
+      "rollout_error": null,
+      "rollout_file": {
+        "bytes": 327328,
+        "sha256": "328cf546022fd2f79da3ec1a6161c0d3ebd777bd254edd89903a2629966a5c8e"
+      },
+      "rollout_status": "complete",
+      "rollout_wall_seconds": 13.231359249999514,
+      "score": {
+        "evidence_scope": "Recomputed simulator-authored numeric evidence; no physics rerun, source attestation or hardware sensing.",
+        "false_alarms": {
+          "reason": "Continuous residual policy has no declared discrete alarm events.",
+          "status": "not_applicable"
+        },
+        "grip_events": [
+          {
+            "kind": "supported_grip_interruption",
+            "loss_onset_s": 12.32299999999861,
+            "loss_qualified_s": 12.341999999998599,
+            "onset_to_qualified_restoration_s": 0.4259999999997639,
+            "onset_to_restoration_s": 0.32699999999981877,
+            "relative_motion_since_loss_m": 0.033405944668013256,
+            "restoration_onset_s": 12.649999999998428,
+            "restoration_qualified_s": 12.748999999998373,
+            "restoration_supported": false,
+            "restored": true,
+            "supported_at_onset": true
+          },
+          {
+            "kind": "supported_grip_interruption",
+            "loss_onset_s": 13.181999999998133,
+            "loss_qualified_s": 13.200999999998123,
+            "onset_to_qualified_restoration_s": 0.21699999999987973,
+            "onset_to_restoration_s": 0.1179999999999346,
+            "relative_motion_since_loss_m": 0.0201096558077051,
+            "restoration_onset_s": 13.299999999998068,
+            "restoration_qualified_s": 13.398999999998013,
+            "restoration_supported": true,
+            "restored": true,
+            "supported_at_onset": true
+          },
+          {
+            "kind": "supported_grip_interruption",
+            "loss_onset_s": 17.108999999997927,
+            "loss_qualified_s": 17.12799999999795,
+            "onset_to_qualified_restoration_s": null,
+            "onset_to_restoration_s": null,
+            "relative_motion_since_loss_m": 0.2010864554938082,
+            "restoration_onset_s": null,
+            "restoration_qualified_s": null,
+            "restored": false,
+            "supported_at_onset": true
+          }
+        ],
+        "grip_summary": {
+          "airborne_events": 0,
+          "loss_events": 3,
+          "qualification_latency_s": [
+            0.4259999999997639,
+            0.21699999999987973
+          ],
+          "restored_events": 2,
+          "supported_events": 3,
+          "unresolved_events": 1
+        },
+        "native_diagnostics": {
+          "premature_open_jaw": {
+            "count": 0,
+            "intervals": [],
+            "maximum_s": 0.0,
+            "total_s": 0.0
+          },
+          "support_touchdown_before_destination": {
+            "count": 0,
+            "intervals": [],
+            "maximum_s": 0.0,
+            "total_s": 0.0
+          }
+        },
+        "native_metrics_mismatches": [],
+        "native_recomputed": {
+          "carry_complete": false,
+          "carry_contact_loss_events": 0,
+          "dwell_seconds": 0.0,
+          "first_success_time_seconds": null,
+          "grasp_seen": true,
+          "lift_dwell_seconds": 0.0,
+          "lift_seen": false,
+          "max_carry_contact_loss_seconds": 0.0,
+          "max_object_lift_m": 0.032622070113360255,
+          "max_pad_normal_force_n": 16.822725404528523,
+          "retention_lost": false,
+          "success": false,
+          "unexpected_contact_native_ticks": 7280
+        },
+        "reason": "invalid_state_value",
+        "restoration_scope": "Loaded contact restored for 100 ms after sustained loss; does not establish task recovery or airborne regrasp.",
+        "schema": "dvidia.piper-feedback-score/v1",
+        "strict_v01": {
+          "reason": "invalid_state_value",
+          "reference_xy_m": [
+            0.35,
+            0.08
+          ],
+          "state_index": 739,
+          "success": false
+        },
+        "success": false,
+        "terminal_only": {
+          "dwell_s": 0.0,
+          "scope": "Terminal-only geometry/release/settling/dwell; excludes lift, carry and retention history.",
+          "success": false,
+          "terminal_xy_error_m": 0.35894977561012875
+        },
+        "valid": true
+      },
+      "scored_control_trajectory": true,
+      "strict_success": false,
+      "supervisor_cause": null,
+      "supervisor_status": "finished",
+      "variant": "delta",
+      "worker_receipt": {
+        "bytes": 5647,
+        "sha256": "96102ceb6a288286b88dfac78b6e48fd2ad5f7fec02e5278b641802e0c650e7b"
+      },
+      "worker_receipt_available": true,
+      "worker_wall_seconds": 14.912654209008906
+    },
+    {
+      "camera_pairs": 501,
+      "case": {
+        "goal_id": 1,
+        "goal_position_m": [
+          0.4,
+          -0.08,
+          0.08
+        ],
+        "group_id": "development-00",
+        "id": "development-00-goal1",
+        "object_position_m": [
+          0.303692,
+          -0.033335,
+          0.082
+        ],
+        "seed": 311466,
+        "split": "development"
+      },
+      "files_verified": 5010,
+      "frame_bytes": 168809295,
+      "guard": {
+        "checks": 1000,
+        "forwarded_calls": 1000,
+        "guard_wall_seconds": 0.005257312848698348,
+        "minimum_free_bytes": 40164065280,
+        "scope": "Once per simulated20ms callback, not a wall-clock50Hz monitor; no task features changed.",
+        "stop_reason": null
+      },
+      "id": "development-00-goal1-absolute",
+      "inspection": {
+        "files_verified": 5010,
+        "physical_robot_ready": false,
+        "physics_executed": false,
+        "recorded_complete_intervals": 1000,
+        "schema": "dvidia.camera-action-rollout/v2/inspection",
+        "score": {
+          "evidence_scope": "Recomputed simulator-authored numeric evidence; no physics rerun, source attestation or hardware sensing.",
+          "false_alarms": {
+            "reason": "Continuous residual policy has no declared discrete alarm events.",
+            "status": "not_applicable"
+          },
+          "grip_events": [
+            {
+              "kind": "supported_grip_interruption",
+              "loss_onset_s": 11.205999999999229,
+              "loss_qualified_s": 11.224999999999218,
+              "onset_to_qualified_restoration_s": null,
+              "onset_to_restoration_s": null,
+              "relative_motion_since_loss_m": 0.032740883409154146,
+              "restoration_onset_s": null,
+              "restoration_qualified_s": null,
+              "restored": false,
+              "supported_at_onset": true
+            }
+          ],
+          "grip_summary": {
+            "airborne_events": 0,
+            "loss_events": 1,
+            "qualification_latency_s": [],
+            "restored_events": 0,
+            "supported_events": 1,
+            "unresolved_events": 1
+          },
+          "native_diagnostics": {
+            "premature_open_jaw": {
+              "count": 0,
+              "intervals": [],
+              "maximum_s": 0.0,
+              "total_s": 0.0
+            },
+            "support_touchdown_before_destination": {
+              "count": 0,
+              "intervals": [],
+              "maximum_s": 0.0,
+              "total_s": 0.0
+            }
+          },
+          "native_metrics_mismatches": [],
+          "native_recomputed": {
+            "carry_complete": false,
+            "carry_contact_loss_events": 0,
+            "dwell_seconds": 0.0,
+            "first_success_time_seconds": null,
+            "grasp_seen": true,
+            "lift_dwell_seconds": 0.0,
+            "lift_seen": false,
+            "max_carry_contact_loss_seconds": 0.0,
+            "max_object_lift_m": 0.005454372207929728,
+            "max_pad_normal_force_n": 1.8253660550428599,
+            "retention_lost": false,
+            "success": false,
+            "unexpected_contact_native_ticks": 152
+          },
+          "reason": "native_outcome_failed",
+          "restoration_scope": "Loaded contact restored for 100 ms after sustained loss; does not establish task recovery or airborne regrasp.",
+          "schema": "dvidia.piper-feedback-score/v1",
+          "strict_v01": {
+            "native_metrics_source": "simulator-authored episode record; not authenticated sensor input",
+            "native_outcome_success": false,
+            "native_outcome_valid": true,
+            "reason": "native_outcome_failed",
+            "reference_xy_m": [
+              0.4,
+              -0.08
+            ],
+            "required_dwell_s": 0.25,
+            "required_lift_dwell_s": 0.1,
+            "sample_interval_s": 0.02,
+            "success": false,
+            "terminal_dwell_s": 0.0,
+            "terminal_reference_success": false,
+            "terminal_xy_error_m": 0.004566382464358538,
+            "trace_lift_dwell_s": 0.0,
+            "trajectory_complete": true
+          },
+          "success": false,
+          "terminal_only": {
+            "dwell_s": 0.0,
+            "scope": "Terminal-only geometry/release/settling/dwell; excludes lift, carry and retention history.",
+            "success": false,
+            "terminal_xy_error_m": 0.004566382464358537
+          },
+          "valid": true
+        },
+        "source_authenticated": false,
+        "status": "complete",
+        "strict_success": false
+      },
+      "integrity_file": {
+        "bytes": 567674,
+        "sha256": "08f7fbfcfa7a595575b5bf9e35f64a22a265862f3a33044282199847598bf095"
+      },
+      "perception": {
+        "available": 320,
+        "decisions": 1000,
+        "longest_missing_seconds": 13.36,
+        "missing": 680
+      },
+      "policy": {
+        "model_sha256": "36af3fe1670494b04c48b5114d9be1a5f91632aa7992ba529d6587969ea37460",
+        "physical_robot_ready": false,
+        "policy_origin": "camera_encoder_reference_endpoint_numpy_mlp",
+        "runtime_teacher_available": false,
+        "student_inference_calls": 1000,
+        "student_inference_p95_seconds": 0.0002635062483022921,
+        "student_inference_seconds": 0.22442970832344145,
+        "task_qualified": false,
+        "teacher_fallback_calls": 0,
+        "timing_scope": "Causal feature encoding, NumPy inference and reference projection; excludes camera/render/actuation.",
+        "variant": "absolute"
+      },
+      "qualified": false,
+      "recorded_complete_intervals": 1000,
+      "rollout_error": null,
+      "rollout_file": {
+        "bytes": 326955,
+        "sha256": "8b851f39d22f39b02709c7857d1c2a6afa2420e83d200e7ee9564baeeba82b07"
+      },
+      "rollout_status": "complete",
+      "rollout_wall_seconds": 12.444784249993972,
+      "score": {
+        "evidence_scope": "Recomputed simulator-authored numeric evidence; no physics rerun, source attestation or hardware sensing.",
+        "false_alarms": {
+          "reason": "Continuous residual policy has no declared discrete alarm events.",
+          "status": "not_applicable"
+        },
+        "grip_events": [
+          {
+            "kind": "supported_grip_interruption",
+            "loss_onset_s": 11.205999999999229,
+            "loss_qualified_s": 11.224999999999218,
+            "onset_to_qualified_restoration_s": null,
+            "onset_to_restoration_s": null,
+            "relative_motion_since_loss_m": 0.032740883409154146,
+            "restoration_onset_s": null,
+            "restoration_qualified_s": null,
+            "restored": false,
+            "supported_at_onset": true
+          }
+        ],
+        "grip_summary": {
+          "airborne_events": 0,
+          "loss_events": 1,
+          "qualification_latency_s": [],
+          "restored_events": 0,
+          "supported_events": 1,
+          "unresolved_events": 1
+        },
+        "native_diagnostics": {
+          "premature_open_jaw": {
+            "count": 0,
+            "intervals": [],
+            "maximum_s": 0.0,
+            "total_s": 0.0
+          },
+          "support_touchdown_before_destination": {
+            "count": 0,
+            "intervals": [],
+            "maximum_s": 0.0,
+            "total_s": 0.0
+          }
+        },
+        "native_metrics_mismatches": [],
+        "native_recomputed": {
+          "carry_complete": false,
+          "carry_contact_loss_events": 0,
+          "dwell_seconds": 0.0,
+          "first_success_time_seconds": null,
+          "grasp_seen": true,
+          "lift_dwell_seconds": 0.0,
+          "lift_seen": false,
+          "max_carry_contact_loss_seconds": 0.0,
+          "max_object_lift_m": 0.005454372207929728,
+          "max_pad_normal_force_n": 1.8253660550428599,
+          "retention_lost": false,
+          "success": false,
+          "unexpected_contact_native_ticks": 152
+        },
+        "reason": "native_outcome_failed",
+        "restoration_scope": "Loaded contact restored for 100 ms after sustained loss; does not establish task recovery or airborne regrasp.",
+        "schema": "dvidia.piper-feedback-score/v1",
+        "strict_v01": {
+          "native_metrics_source": "simulator-authored episode record; not authenticated sensor input",
+          "native_outcome_success": false,
+          "native_outcome_valid": true,
+          "reason": "native_outcome_failed",
+          "reference_xy_m": [
+            0.4,
+            -0.08
+          ],
+          "required_dwell_s": 0.25,
+          "required_lift_dwell_s": 0.1,
+          "sample_interval_s": 0.02,
+          "success": false,
+          "terminal_dwell_s": 0.0,
+          "terminal_reference_success": false,
+          "terminal_xy_error_m": 0.004566382464358538,
+          "trace_lift_dwell_s": 0.0,
+          "trajectory_complete": true
+        },
+        "success": false,
+        "terminal_only": {
+          "dwell_s": 0.0,
+          "scope": "Terminal-only geometry/release/settling/dwell; excludes lift, carry and retention history.",
+          "success": false,
+          "terminal_xy_error_m": 0.004566382464358537
+        },
+        "valid": true
+      },
+      "scored_control_trajectory": true,
+      "strict_success": false,
+      "supervisor_cause": null,
+      "supervisor_status": "finished",
+      "variant": "absolute",
+      "worker_receipt": {
+        "bytes": 5151,
+        "sha256": "d1456147a5784b4e80621edf48f807401a953ad8d5f168ea559c202e39eff446"
+      },
+      "worker_receipt_available": true,
+      "worker_wall_seconds": 14.14837254100712
+    },
+    {
+      "camera_pairs": 501,
+      "case": {
+        "goal_id": 1,
+        "goal_position_m": [
+          0.4,
+          -0.08,
+          0.08
+        ],
+        "group_id": "development-00",
+        "id": "development-00-goal1",
+        "object_position_m": [
+          0.303692,
+          -0.033335,
+          0.082
+        ],
+        "seed": 311466,
+        "split": "development"
+      },
+      "files_verified": 5010,
+      "frame_bytes": 168741922,
+      "guard": {
+        "checks": 1000,
+        "forwarded_calls": 1000,
+        "guard_wall_seconds": 0.005355292814783752,
+        "minimum_free_bytes": 39958847488,
+        "scope": "Once per simulated20ms callback, not a wall-clock50Hz monitor; no task features changed.",
+        "stop_reason": null
+      },
+      "id": "development-00-goal1-delta",
+      "inspection": {
+        "files_verified": 5010,
+        "physical_robot_ready": false,
+        "physics_executed": false,
+        "recorded_complete_intervals": 1000,
+        "schema": "dvidia.camera-action-rollout/v2/inspection",
+        "score": {
+          "evidence_scope": "Recomputed simulator-authored numeric evidence; no physics rerun, source attestation or hardware sensing.",
+          "false_alarms": {
+            "reason": "Continuous residual policy has no declared discrete alarm events.",
+            "status": "not_applicable"
+          },
+          "grip_events": [],
+          "grip_summary": {
+            "airborne_events": 0,
+            "loss_events": 0,
+            "qualification_latency_s": [],
+            "restored_events": 0,
+            "supported_events": 0,
+            "unresolved_events": 0
+          },
+          "native_diagnostics": {
+            "premature_open_jaw": {
+              "count": 0,
+              "intervals": [],
+              "maximum_s": 0.0,
+              "total_s": 0.0
+            },
+            "support_touchdown_before_destination": {
+              "count": 0,
+              "intervals": [],
+              "maximum_s": 0.0,
+              "total_s": 0.0
+            }
+          },
+          "native_metrics_mismatches": [],
+          "native_recomputed": {
+            "carry_complete": false,
+            "carry_contact_loss_events": 0,
+            "dwell_seconds": 0.0,
+            "first_success_time_seconds": null,
+            "grasp_seen": true,
+            "lift_dwell_seconds": 0.0,
+            "lift_seen": false,
+            "max_carry_contact_loss_seconds": 0.0,
+            "max_object_lift_m": 0.0011216542473738017,
+            "max_pad_normal_force_n": 1.7540589205645145,
+            "retention_lost": false,
+            "success": false,
+            "unexpected_contact_native_ticks": 8087
+          },
+          "reason": "invalid_state_value",
+          "restoration_scope": "Loaded contact restored for 100 ms after sustained loss; does not establish task recovery or airborne regrasp.",
+          "schema": "dvidia.piper-feedback-score/v1",
+          "strict_v01": {
+            "reason": "invalid_state_value",
+            "reference_xy_m": [
+              0.4,
+              -0.08
+            ],
+            "state_index": 713,
+            "success": false
+          },
+          "success": false,
+          "terminal_only": {
+            "dwell_s": 0.0,
+            "scope": "Terminal-only geometry/release/settling/dwell; excludes lift, carry and retention history.",
+            "success": false,
+            "terminal_xy_error_m": 0.08543266510853878
+          },
+          "valid": true
+        },
+        "source_authenticated": false,
+        "status": "complete",
+        "strict_success": false
+      },
+      "integrity_file": {
+        "bytes": 567674,
+        "sha256": "d9a2f3b46a3906153b5ee703d915c809b9d5903996abbc948a31ab42fae1e1dd"
+      },
+      "perception": {
+        "available": 436,
+        "decisions": 1000,
+        "longest_missing_seconds": 4.92,
+        "missing": 564
+      },
+      "policy": {
+        "model_sha256": "4545be31f50267c9da29b211510c7a32df500678ba107c7a100604bb0c4c18ab",
+        "physical_robot_ready": false,
+        "policy_origin": "camera_encoder_reference_endpoint_numpy_mlp",
+        "runtime_teacher_available": false,
+        "student_inference_calls": 1000,
+        "student_inference_p95_seconds": 0.0002676416508620605,
+        "student_inference_seconds": 0.2281716027064249,
+        "task_qualified": false,
+        "teacher_fallback_calls": 0,
+        "timing_scope": "Causal feature encoding, NumPy inference and reference projection; excludes camera/render/actuation.",
+        "variant": "delta"
+      },
+      "qualified": false,
+      "recorded_complete_intervals": 1000,
+      "rollout_error": null,
+      "rollout_file": {
+        "bytes": 326204,
+        "sha256": "83dee39e2cb1cbfd3ef72c741d854627bf54c021852ffb1430c66fe4b8fe8452"
+      },
+      "rollout_status": "complete",
+      "rollout_wall_seconds": 12.688531916995998,
+      "score": {
+        "evidence_scope": "Recomputed simulator-authored numeric evidence; no physics rerun, source attestation or hardware sensing.",
+        "false_alarms": {
+          "reason": "Continuous residual policy has no declared discrete alarm events.",
+          "status": "not_applicable"
+        },
+        "grip_events": [],
+        "grip_summary": {
+          "airborne_events": 0,
+          "loss_events": 0,
+          "qualification_latency_s": [],
+          "restored_events": 0,
+          "supported_events": 0,
+          "unresolved_events": 0
+        },
+        "native_diagnostics": {
+          "premature_open_jaw": {
+            "count": 0,
+            "intervals": [],
+            "maximum_s": 0.0,
+            "total_s": 0.0
+          },
+          "support_touchdown_before_destination": {
+            "count": 0,
+            "intervals": [],
+            "maximum_s": 0.0,
+            "total_s": 0.0
+          }
+        },
+        "native_metrics_mismatches": [],
+        "native_recomputed": {
+          "carry_complete": false,
+          "carry_contact_loss_events": 0,
+          "dwell_seconds": 0.0,
+          "first_success_time_seconds": null,
+          "grasp_seen": true,
+          "lift_dwell_seconds": 0.0,
+          "lift_seen": false,
+          "max_carry_contact_loss_seconds": 0.0,
+          "max_object_lift_m": 0.0011216542473738017,
+          "max_pad_normal_force_n": 1.7540589205645145,
+          "retention_lost": false,
+          "success": false,
+          "unexpected_contact_native_ticks": 8087
+        },
+        "reason": "invalid_state_value",
+        "restoration_scope": "Loaded contact restored for 100 ms after sustained loss; does not establish task recovery or airborne regrasp.",
+        "schema": "dvidia.piper-feedback-score/v1",
+        "strict_v01": {
+          "reason": "invalid_state_value",
+          "reference_xy_m": [
+            0.4,
+            -0.08
+          ],
+          "state_index": 713,
+          "success": false
+        },
+        "success": false,
+        "terminal_only": {
+          "dwell_s": 0.0,
+          "scope": "Terminal-only geometry/release/settling/dwell; excludes lift, carry and retention history.",
+          "success": false,
+          "terminal_xy_error_m": 0.08543266510853878
+        },
+        "valid": true
+      },
+      "scored_control_trajectory": true,
+      "strict_success": false,
+      "supervisor_cause": null,
+      "supervisor_status": "finished",
+      "variant": "delta",
+      "worker_receipt": {
+        "bytes": 4443,
+        "sha256": "2d7d66b18c09275dd56980c72ee73cf6ab3b7b3f9a25d438e7175c8dbbbbaa03"
+      },
+      "worker_receipt_available": true,
+      "worker_wall_seconds": 14.580125999986194
+    },
+    {
+      "camera_pairs": 501,
+      "case": {
+        "goal_id": 0,
+        "goal_position_m": [
+          0.35,
+          0.08,
+          0.08
+        ],
+        "group_id": "development-01",
+        "id": "development-01-goal0",
+        "object_position_m": [
+          0.303637,
+          -0.039966,
+          0.082
+        ],
+        "seed": 311467,
+        "split": "development"
+      },
+      "files_verified": 5010,
+      "frame_bytes": 169510528,
+      "guard": {
+        "checks": 1000,
+        "forwarded_calls": 1000,
+        "guard_wall_seconds": 0.005813365976791829,
+        "minimum_free_bytes": 39743258624,
+        "scope": "Once per simulated20ms callback, not a wall-clock50Hz monitor; no task features changed.",
+        "stop_reason": null
+      },
+      "id": "development-01-goal0-absolute",
+      "inspection": {
+        "files_verified": 5010,
+        "physical_robot_ready": false,
+        "physics_executed": false,
+        "recorded_complete_intervals": 1000,
+        "schema": "dvidia.camera-action-rollout/v2/inspection",
+        "score": {
+          "evidence_scope": "Recomputed simulator-authored numeric evidence; no physics rerun, source attestation or hardware sensing.",
+          "false_alarms": {
+            "reason": "Continuous residual policy has no declared discrete alarm events.",
+            "status": "not_applicable"
+          },
+          "grip_events": [],
+          "grip_summary": {
+            "airborne_events": 0,
+            "loss_events": 0,
+            "qualification_latency_s": [],
+            "restored_events": 0,
+            "supported_events": 0,
+            "unresolved_events": 0
+          },
+          "native_diagnostics": {
+            "premature_open_jaw": {
+              "count": 0,
+              "intervals": [],
+              "maximum_s": 0.0,
+              "total_s": 0.0
+            },
+            "support_touchdown_before_destination": {
+              "count": 0,
+              "intervals": [],
+              "maximum_s": 0.0,
+              "total_s": 0.0
+            }
+          },
+          "native_metrics_mismatches": [],
+          "native_recomputed": {
+            "carry_complete": false,
+            "carry_contact_loss_events": 0,
+            "dwell_seconds": 0.0,
+            "first_success_time_seconds": null,
+            "grasp_seen": false,
+            "lift_dwell_seconds": 0.0,
+            "lift_seen": false,
+            "max_carry_contact_loss_seconds": 0.0,
+            "max_object_lift_m": 6.22998874710845e-11,
+            "max_pad_normal_force_n": 0.0,
+            "retention_lost": false,
+            "success": false,
+            "unexpected_contact_native_ticks": 0
+          },
+          "reason": "native_outcome_failed",
+          "restoration_scope": "Loaded contact restored for 100 ms after sustained loss; does not establish task recovery or airborne regrasp.",
+          "schema": "dvidia.piper-feedback-score/v1",
+          "strict_v01": {
+            "native_metrics_source": "simulator-authored episode record; not authenticated sensor input",
+            "native_outcome_success": false,
+            "native_outcome_valid": true,
+            "reason": "native_outcome_failed",
+            "reference_xy_m": [
+              0.35,
+              0.08
+            ],
+            "required_dwell_s": 0.25,
+            "required_lift_dwell_s": 0.1,
+            "sample_interval_s": 0.02,
+            "success": false,
+            "terminal_dwell_s": 0.0,
+            "terminal_reference_success": false,
+            "terminal_xy_error_m": 0.12861307125253946,
+            "trace_lift_dwell_s": 0.0,
+            "trajectory_complete": true
+          },
+          "success": false,
+          "terminal_only": {
+            "dwell_s": 0.0,
+            "scope": "Terminal-only geometry/release/settling/dwell; excludes lift, carry and retention history.",
+            "success": false,
+            "terminal_xy_error_m": 0.12861307125253946
+          },
+          "valid": true
+        },
+        "source_authenticated": false,
+        "status": "complete",
+        "strict_success": false
+      },
+      "integrity_file": {
+        "bytes": 567674,
+        "sha256": "22443164911963ff89e31663dc3a938cfeffd151a8e569f2d98f2563f18a8065"
+      },
+      "perception": {
+        "available": 1000,
+        "decisions": 1000,
+        "longest_missing_seconds": 0.0,
+        "missing": 0
+      },
+      "policy": {
+        "model_sha256": "36af3fe1670494b04c48b5114d9be1a5f91632aa7992ba529d6587969ea37460",
+        "physical_robot_ready": false,
+        "policy_origin": "camera_encoder_reference_endpoint_numpy_mlp",
+        "runtime_teacher_available": false,
+        "student_inference_calls": 1000,
+        "student_inference_p95_seconds": 0.00030651460547233,
+        "student_inference_seconds": 0.24899018558789976,
+        "task_qualified": false,
+        "teacher_fallback_calls": 0,
+        "timing_scope": "Causal feature encoding, NumPy inference and reference projection; excludes camera/render/actuation.",
+        "variant": "absolute"
+      },
+      "qualified": false,
+      "recorded_complete_intervals": 1000,
+      "rollout_error": null,
+      "rollout_file": {
+        "bytes": 326367,
+        "sha256": "47f735604508f95ad78f9d863ca88f13f77b0bb4f68a4fcc4b067b967704fb97"
+      },
+      "rollout_status": "complete",
+      "rollout_wall_seconds": 13.759285707987146,
+      "score": {
+        "evidence_scope": "Recomputed simulator-authored numeric evidence; no physics rerun, source attestation or hardware sensing.",
+        "false_alarms": {
+          "reason": "Continuous residual policy has no declared discrete alarm events.",
+          "status": "not_applicable"
+        },
+        "grip_events": [],
+        "grip_summary": {
+          "airborne_events": 0,
+          "loss_events": 0,
+          "qualification_latency_s": [],
+          "restored_events": 0,
+          "supported_events": 0,
+          "unresolved_events": 0
+        },
+        "native_diagnostics": {
+          "premature_open_jaw": {
+            "count": 0,
+            "intervals": [],
+            "maximum_s": 0.0,
+            "total_s": 0.0
+          },
+          "support_touchdown_before_destination": {
+            "count": 0,
+            "intervals": [],
+            "maximum_s": 0.0,
+            "total_s": 0.0
+          }
+        },
+        "native_metrics_mismatches": [],
+        "native_recomputed": {
+          "carry_complete": false,
+          "carry_contact_loss_events": 0,
+          "dwell_seconds": 0.0,
+          "first_success_time_seconds": null,
+          "grasp_seen": false,
+          "lift_dwell_seconds": 0.0,
+          "lift_seen": false,
+          "max_carry_contact_loss_seconds": 0.0,
+          "max_object_lift_m": 6.22998874710845e-11,
+          "max_pad_normal_force_n": 0.0,
+          "retention_lost": false,
+          "success": false,
+          "unexpected_contact_native_ticks": 0
+        },
+        "reason": "native_outcome_failed",
+        "restoration_scope": "Loaded contact restored for 100 ms after sustained loss; does not establish task recovery or airborne regrasp.",
+        "schema": "dvidia.piper-feedback-score/v1",
+        "strict_v01": {
+          "native_metrics_source": "simulator-authored episode record; not authenticated sensor input",
+          "native_outcome_success": false,
+          "native_outcome_valid": true,
+          "reason": "native_outcome_failed",
+          "reference_xy_m": [
+            0.35,
+            0.08
+          ],
+          "required_dwell_s": 0.25,
+          "required_lift_dwell_s": 0.1,
+          "sample_interval_s": 0.02,
+          "success": false,
+          "terminal_dwell_s": 0.0,
+          "terminal_reference_success": false,
+          "terminal_xy_error_m": 0.12861307125253946,
+          "trace_lift_dwell_s": 0.0,
+          "trajectory_complete": true
+        },
+        "success": false,
+        "terminal_only": {
+          "dwell_s": 0.0,
+          "scope": "Terminal-only geometry/release/settling/dwell; excludes lift, carry and retention history.",
+          "success": false,
+          "terminal_xy_error_m": 0.12861307125253946
+        },
+        "valid": true
+      },
+      "scored_control_trajectory": true,
+      "strict_success": false,
+      "supervisor_cause": null,
+      "supervisor_status": "finished",
+      "variant": "absolute",
+      "worker_receipt": {
+        "bytes": 4794,
+        "sha256": "170cecc5eb6733f7b1f3398af31f4f795b0115cc0860d9f8d13075d97e1297ed"
+      },
+      "worker_receipt_available": true,
+      "worker_wall_seconds": 15.491917999985162
+    },
+    {
+      "camera_pairs": 501,
+      "case": {
+        "goal_id": 0,
+        "goal_position_m": [
+          0.35,
+          0.08,
+          0.08
+        ],
+        "group_id": "development-01",
+        "id": "development-01-goal0",
+        "object_position_m": [
+          0.303637,
+          -0.039966,
+          0.082
+        ],
+        "seed": 311467,
+        "split": "development"
+      },
+      "files_verified": 5010,
+      "frame_bytes": 166698225,
+      "guard": {
+        "checks": 1000,
+        "forwarded_calls": 1000,
+        "guard_wall_seconds": 0.005421159323304892,
+        "minimum_free_bytes": 39533481984,
+        "scope": "Once per simulated20ms callback, not a wall-clock50Hz monitor; no task features changed.",
+        "stop_reason": null
+      },
+      "id": "development-01-goal0-delta",
+      "inspection": {
+        "files_verified": 5010,
+        "physical_robot_ready": false,
+        "physics_executed": false,
+        "recorded_complete_intervals": 1000,
+        "schema": "dvidia.camera-action-rollout/v2/inspection",
+        "score": {
+          "evidence_scope": "Recomputed simulator-authored numeric evidence; no physics rerun, source attestation or hardware sensing.",
+          "false_alarms": {
+            "reason": "Continuous residual policy has no declared discrete alarm events.",
+            "status": "not_applicable"
+          },
+          "grip_events": [],
+          "grip_summary": {
+            "airborne_events": 0,
+            "loss_events": 0,
+            "qualification_latency_s": [],
+            "restored_events": 0,
+            "supported_events": 0,
+            "unresolved_events": 0
+          },
+          "native_diagnostics": {
+            "premature_open_jaw": {
+              "count": 0,
+              "intervals": [],
+              "maximum_s": 0.0,
+              "total_s": 0.0
+            },
+            "support_touchdown_before_destination": {
+              "count": 0,
+              "intervals": [],
+              "maximum_s": 0.0,
+              "total_s": 0.0
+            }
+          },
+          "native_metrics_mismatches": [],
+          "native_recomputed": {
+            "carry_complete": false,
+            "carry_contact_loss_events": 0,
+            "dwell_seconds": 0.0,
+            "first_success_time_seconds": null,
+            "grasp_seen": true,
+            "lift_dwell_seconds": 0.0,
+            "lift_seen": false,
+            "max_carry_contact_loss_seconds": 0.0,
+            "max_object_lift_m": 0.01765747992294485,
+            "max_pad_normal_force_n": 2.5154725382030216,
+            "retention_lost": false,
+            "success": false,
+            "unexpected_contact_native_ticks": 1240
+          },
+          "reason": "invalid_state_value",
+          "restoration_scope": "Loaded contact restored for 100 ms after sustained loss; does not establish task recovery or airborne regrasp.",
+          "schema": "dvidia.piper-feedback-score/v1",
+          "strict_v01": {
+            "reason": "invalid_state_value",
+            "reference_xy_m": [
+              0.35,
+              0.08
+            ],
+            "state_index": 687,
+            "success": false
+          },
+          "success": false,
+          "terminal_only": {
+            "dwell_s": 0.0,
+            "scope": "Terminal-only geometry/release/settling/dwell; excludes lift, carry and retention history.",
+            "success": false,
+            "terminal_xy_error_m": 0.2545568078047586
+          },
+          "valid": true
+        },
+        "source_authenticated": false,
+        "status": "complete",
+        "strict_success": false
+      },
+      "integrity_file": {
+        "bytes": 567674,
+        "sha256": "1f4b2db19b5c0ded4438d9cc549a65e6a25a9f1691118a4363f93bf0646e2a89"
+      },
+      "perception": {
+        "available": 732,
+        "decisions": 1000,
+        "longest_missing_seconds": 4.04,
+        "missing": 268
+      },
+      "policy": {
+        "model_sha256": "4545be31f50267c9da29b211510c7a32df500678ba107c7a100604bb0c4c18ab",
+        "physical_robot_ready": false,
+        "policy_origin": "camera_encoder_reference_endpoint_numpy_mlp",
+        "runtime_teacher_available": false,
+        "student_inference_calls": 1000,
+        "student_inference_p95_seconds": 0.0002795872074784711,
+        "student_inference_seconds": 0.2348349518433679,
+        "task_qualified": false,
+        "teacher_fallback_calls": 0,
+        "timing_scope": "Causal feature encoding, NumPy inference and reference projection; excludes camera/render/actuation.",
+        "variant": "delta"
+      },
+      "qualified": false,
+      "recorded_complete_intervals": 1000,
+      "rollout_error": null,
+      "rollout_file": {
+        "bytes": 326043,
+        "sha256": "1302c51513d638910857c79bc78539bc9c59b546aeaf212d08b922ec06372a21"
+      },
+      "rollout_status": "complete",
+      "rollout_wall_seconds": 12.975723750016186,
+      "score": {
+        "evidence_scope": "Recomputed simulator-authored numeric evidence; no physics rerun, source attestation or hardware sensing.",
+        "false_alarms": {
+          "reason": "Continuous residual policy has no declared discrete alarm events.",
+          "status": "not_applicable"
+        },
+        "grip_events": [],
+        "grip_summary": {
+          "airborne_events": 0,
+          "loss_events": 0,
+          "qualification_latency_s": [],
+          "restored_events": 0,
+          "supported_events": 0,
+          "unresolved_events": 0
+        },
+        "native_diagnostics": {
+          "premature_open_jaw": {
+            "count": 0,
+            "intervals": [],
+            "maximum_s": 0.0,
+            "total_s": 0.0
+          },
+          "support_touchdown_before_destination": {
+            "count": 0,
+            "intervals": [],
+            "maximum_s": 0.0,
+            "total_s": 0.0
+          }
+        },
+        "native_metrics_mismatches": [],
+        "native_recomputed": {
+          "carry_complete": false,
+          "carry_contact_loss_events": 0,
+          "dwell_seconds": 0.0,
+          "first_success_time_seconds": null,
+          "grasp_seen": true,
+          "lift_dwell_seconds": 0.0,
+          "lift_seen": false,
+          "max_carry_contact_loss_seconds": 0.0,
+          "max_object_lift_m": 0.01765747992294485,
+          "max_pad_normal_force_n": 2.5154725382030216,
+          "retention_lost": false,
+          "success": false,
+          "unexpected_contact_native_ticks": 1240
+        },
+        "reason": "invalid_state_value",
+        "restoration_scope": "Loaded contact restored for 100 ms after sustained loss; does not establish task recovery or airborne regrasp.",
+        "schema": "dvidia.piper-feedback-score/v1",
+        "strict_v01": {
+          "reason": "invalid_state_value",
+          "reference_xy_m": [
+            0.35,
+            0.08
+          ],
+          "state_index": 687,
+          "success": false
+        },
+        "success": false,
+        "terminal_only": {
+          "dwell_s": 0.0,
+          "scope": "Terminal-only geometry/release/settling/dwell; excludes lift, carry and retention history.",
+          "success": false,
+          "terminal_xy_error_m": 0.2545568078047586
+        },
+        "valid": true
+      },
+      "scored_control_trajectory": true,
+      "strict_success": false,
+      "supervisor_cause": null,
+      "supervisor_status": "finished",
+      "variant": "delta",
+      "worker_receipt": {
+        "bytes": 4438,
+        "sha256": "992235bea602ec6bdcd12f82296d5a2f8b89d6ca7dfc598808e9f35f1256268d"
+      },
+      "worker_receipt_available": true,
+      "worker_wall_seconds": 14.5429857079871
+    },
+    {
+      "camera_pairs": 501,
+      "case": {
+        "goal_id": 1,
+        "goal_position_m": [
+          0.4,
+          -0.08,
+          0.08
+        ],
+        "group_id": "development-01",
+        "id": "development-01-goal1",
+        "object_position_m": [
+          0.303637,
+          -0.039966,
+          0.082
+        ],
+        "seed": 311468,
+        "split": "development"
+      },
+      "files_verified": 5010,
+      "frame_bytes": 169366324,
+      "guard": {
+        "checks": 1000,
+        "forwarded_calls": 1000,
+        "guard_wall_seconds": 0.005538281751796603,
+        "minimum_free_bytes": 39318462464,
+        "scope": "Once per simulated20ms callback, not a wall-clock50Hz monitor; no task features changed.",
+        "stop_reason": null
+      },
+      "id": "development-01-goal1-absolute",
+      "inspection": {
+        "files_verified": 5010,
+        "physical_robot_ready": false,
+        "physics_executed": false,
+        "recorded_complete_intervals": 1000,
+        "schema": "dvidia.camera-action-rollout/v2/inspection",
+        "score": {
+          "evidence_scope": "Recomputed simulator-authored numeric evidence; no physics rerun, source attestation or hardware sensing.",
+          "false_alarms": {
+            "reason": "Continuous residual policy has no declared discrete alarm events.",
+            "status": "not_applicable"
+          },
+          "grip_events": [
+            {
+              "kind": "supported_grip_interruption",
+              "loss_onset_s": 18.28199999999936,
+              "loss_qualified_s": 18.300999999999384,
+              "onset_to_qualified_restoration_s": 0.1490000000001821,
+              "onset_to_restoration_s": 0.05000000000006111,
+              "relative_motion_since_loss_m": 0.0009026515472255445,
+              "restoration_onset_s": 18.33199999999942,
+              "restoration_qualified_s": 18.430999999999543,
+              "restoration_supported": true,
+              "restored": true,
+              "supported_at_onset": true
+            }
+          ],
+          "grip_summary": {
+            "airborne_events": 0,
+            "loss_events": 1,
+            "qualification_latency_s": [
+              0.1490000000001821
+            ],
+            "restored_events": 1,
+            "supported_events": 1,
+            "unresolved_events": 0
+          },
+          "native_diagnostics": {
+            "premature_open_jaw": {
+              "count": 0,
+              "intervals": [],
+              "maximum_s": 0.0,
+              "total_s": 0.0
+            },
+            "support_touchdown_before_destination": {
+              "count": 0,
+              "intervals": [],
+              "maximum_s": 0.0,
+              "total_s": 0.0
+            }
+          },
+          "native_metrics_mismatches": [],
+          "native_recomputed": {
+            "carry_complete": false,
+            "carry_contact_loss_events": 0,
+            "dwell_seconds": 0.0,
+            "first_success_time_seconds": null,
+            "grasp_seen": true,
+            "lift_dwell_seconds": 0.0,
+            "lift_seen": true,
+            "max_carry_contact_loss_seconds": 0.0,
+            "max_object_lift_m": 0.12241943079952637,
+            "max_pad_normal_force_n": 0.6881092160543247,
+            "retention_lost": false,
+            "success": false,
+            "unexpected_contact_native_ticks": 0
+          },
+          "reason": "native_outcome_failed",
+          "restoration_scope": "Loaded contact restored for 100 ms after sustained loss; does not establish task recovery or airborne regrasp.",
+          "schema": "dvidia.piper-feedback-score/v1",
+          "strict_v01": {
+            "native_metrics_source": "simulator-authored episode record; not authenticated sensor input",
+            "native_outcome_success": false,
+            "native_outcome_valid": true,
+            "reason": "native_outcome_failed",
+            "reference_xy_m": [
+              0.4,
+              -0.08
+            ],
+            "required_dwell_s": 0.25,
+            "required_lift_dwell_s": 0.1,
+            "sample_interval_s": 0.02,
+            "success": false,
+            "terminal_dwell_s": 0.0,
+            "terminal_reference_success": false,
+            "terminal_xy_error_m": 0.004469139679330379,
+            "trace_lift_dwell_s": 0.9799999999999969,
+            "trajectory_complete": true
+          },
+          "success": false,
+          "terminal_only": {
+            "dwell_s": 0.0,
+            "scope": "Terminal-only geometry/release/settling/dwell; excludes lift, carry and retention history.",
+            "success": false,
+            "terminal_xy_error_m": 0.004469139679330379
+          },
+          "valid": true
+        },
+        "source_authenticated": false,
+        "status": "complete",
+        "strict_success": false
+      },
+      "integrity_file": {
+        "bytes": 567674,
+        "sha256": "00eea091817448e593ee0d9b6149af6b492291194f9caa914c942a6bc75fd9e7"
+      },
+      "perception": {
+        "available": 692,
+        "decisions": 1000,
+        "longest_missing_seconds": 5.96,
+        "missing": 308
+      },
+      "policy": {
+        "model_sha256": "36af3fe1670494b04c48b5114d9be1a5f91632aa7992ba529d6587969ea37460",
+        "physical_robot_ready": false,
+        "policy_origin": "camera_encoder_reference_endpoint_numpy_mlp",
+        "runtime_teacher_available": false,
+        "student_inference_calls": 1000,
+        "student_inference_p95_seconds": 0.0002899830491514876,
+        "student_inference_seconds": 0.23790540907066315,
+        "task_qualified": false,
+        "teacher_fallback_calls": 0,
+        "timing_scope": "Causal feature encoding, NumPy inference and reference projection; excludes camera/render/actuation.",
+        "variant": "absolute"
+      },
+      "qualified": false,
+      "recorded_complete_intervals": 1000,
+      "rollout_error": null,
+      "rollout_file": {
+        "bytes": 326868,
+        "sha256": "bb86dd380867955a60df4588140dc5c07e4f49129197c09e6b7fae9c64ce0d77"
+      },
+      "rollout_status": "complete",
+      "rollout_wall_seconds": 13.351964041998144,
+      "score": {
+        "evidence_scope": "Recomputed simulator-authored numeric evidence; no physics rerun, source attestation or hardware sensing.",
+        "false_alarms": {
+          "reason": "Continuous residual policy has no declared discrete alarm events.",
+          "status": "not_applicable"
+        },
+        "grip_events": [
+          {
+            "kind": "supported_grip_interruption",
+            "loss_onset_s": 18.28199999999936,
+            "loss_qualified_s": 18.300999999999384,
+            "onset_to_qualified_restoration_s": 0.1490000000001821,
+            "onset_to_restoration_s": 0.05000000000006111,
+            "relative_motion_since_loss_m": 0.0009026515472255445,
+            "restoration_onset_s": 18.33199999999942,
+            "restoration_qualified_s": 18.430999999999543,
+            "restoration_supported": true,
+            "restored": true,
+            "supported_at_onset": true
+          }
+        ],
+        "grip_summary": {
+          "airborne_events": 0,
+          "loss_events": 1,
+          "qualification_latency_s": [
+            0.1490000000001821
+          ],
+          "restored_events": 1,
+          "supported_events": 1,
+          "unresolved_events": 0
+        },
+        "native_diagnostics": {
+          "premature_open_jaw": {
+            "count": 0,
+            "intervals": [],
+            "maximum_s": 0.0,
+            "total_s": 0.0
+          },
+          "support_touchdown_before_destination": {
+            "count": 0,
+            "intervals": [],
+            "maximum_s": 0.0,
+            "total_s": 0.0
+          }
+        },
+        "native_metrics_mismatches": [],
+        "native_recomputed": {
+          "carry_complete": false,
+          "carry_contact_loss_events": 0,
+          "dwell_seconds": 0.0,
+          "first_success_time_seconds": null,
+          "grasp_seen": true,
+          "lift_dwell_seconds": 0.0,
+          "lift_seen": true,
+          "max_carry_contact_loss_seconds": 0.0,
+          "max_object_lift_m": 0.12241943079952637,
+          "max_pad_normal_force_n": 0.6881092160543247,
+          "retention_lost": false,
+          "success": false,
+          "unexpected_contact_native_ticks": 0
+        },
+        "reason": "native_outcome_failed",
+        "restoration_scope": "Loaded contact restored for 100 ms after sustained loss; does not establish task recovery or airborne regrasp.",
+        "schema": "dvidia.piper-feedback-score/v1",
+        "strict_v01": {
+          "native_metrics_source": "simulator-authored episode record; not authenticated sensor input",
+          "native_outcome_success": false,
+          "native_outcome_valid": true,
+          "reason": "native_outcome_failed",
+          "reference_xy_m": [
+            0.4,
+            -0.08
+          ],
+          "required_dwell_s": 0.25,
+          "required_lift_dwell_s": 0.1,
+          "sample_interval_s": 0.02,
+          "success": false,
+          "terminal_dwell_s": 0.0,
+          "terminal_reference_success": false,
+          "terminal_xy_error_m": 0.004469139679330379,
+          "trace_lift_dwell_s": 0.9799999999999969,
+          "trajectory_complete": true
+        },
+        "success": false,
+        "terminal_only": {
+          "dwell_s": 0.0,
+          "scope": "Terminal-only geometry/release/settling/dwell; excludes lift, carry and retention history.",
+          "success": false,
+          "terminal_xy_error_m": 0.004469139679330379
+        },
+        "valid": true
+      },
+      "scored_control_trajectory": true,
+      "strict_success": false,
+      "supervisor_cause": null,
+      "supervisor_status": "finished",
+      "variant": "absolute",
+      "worker_receipt": {
+        "bytes": 5262,
+        "sha256": "9e24e931e0063d24b32beb1642abe26d59756af2c6a6d29e31d975f4a828d185"
+      },
+      "worker_receipt_available": true,
+      "worker_wall_seconds": 15.055389708984876
+    },
+    {
+      "camera_pairs": 501,
+      "case": {
+        "goal_id": 1,
+        "goal_position_m": [
+          0.4,
+          -0.08,
+          0.08
+        ],
+        "group_id": "development-01",
+        "id": "development-01-goal1",
+        "object_position_m": [
+          0.303637,
+          -0.039966,
+          0.082
+        ],
+        "seed": 311468,
+        "split": "development"
+      },
+      "files_verified": 5010,
+      "frame_bytes": 168677614,
+      "guard": {
+        "checks": 1000,
+        "forwarded_calls": 1000,
+        "guard_wall_seconds": 0.005143560847500339,
+        "minimum_free_bytes": 39110926336,
+        "scope": "Once per simulated20ms callback, not a wall-clock50Hz monitor; no task features changed.",
+        "stop_reason": null
+      },
+      "id": "development-01-goal1-delta",
+      "inspection": {
+        "files_verified": 5010,
+        "physical_robot_ready": false,
+        "physics_executed": false,
+        "recorded_complete_intervals": 1000,
+        "schema": "dvidia.camera-action-rollout/v2/inspection",
+        "score": {
+          "evidence_scope": "Recomputed simulator-authored numeric evidence; no physics rerun, source attestation or hardware sensing.",
+          "false_alarms": {
+            "reason": "Continuous residual policy has no declared discrete alarm events.",
+            "status": "not_applicable"
+          },
+          "grip_events": [
+            {
+              "kind": "supported_grip_interruption",
+              "loss_onset_s": 13.27699999999808,
+              "loss_qualified_s": 13.29599999999807,
+              "onset_to_qualified_restoration_s": 0.5129999999997157,
+              "onset_to_restoration_s": 0.41399999999977055,
+              "relative_motion_since_loss_m": 0.04689580619513455,
+              "restoration_onset_s": 13.690999999997851,
+              "restoration_qualified_s": 13.789999999997796,
+              "restoration_supported": false,
+              "restored": true,
+              "supported_at_onset": true
+            },
+            {
+              "kind": "supported_grip_interruption",
+              "loss_onset_s": 13.949999999997708,
+              "loss_qualified_s": 13.968999999997697,
+              "onset_to_qualified_restoration_s": null,
+              "onset_to_restoration_s": null,
+              "relative_motion_since_loss_m": 0.306676529975985,
+              "restoration_onset_s": null,
+              "restoration_qualified_s": null,
+              "restored": false,
+              "supported_at_onset": true
+            }
+          ],
+          "grip_summary": {
+            "airborne_events": 0,
+            "loss_events": 2,
+            "qualification_latency_s": [
+              0.5129999999997157
+            ],
+            "restored_events": 1,
+            "supported_events": 2,
+            "unresolved_events": 1
+          },
+          "native_diagnostics": {
+            "premature_open_jaw": {
+              "count": 0,
+              "intervals": [],
+              "maximum_s": 0.0,
+              "total_s": 0.0
+            },
+            "support_touchdown_before_destination": {
+              "count": 0,
+              "intervals": [],
+              "maximum_s": 0.0,
+              "total_s": 0.0
+            }
+          },
+          "native_metrics_mismatches": [],
+          "native_recomputed": {
+            "carry_complete": false,
+            "carry_contact_loss_events": 0,
+            "dwell_seconds": 0.0,
+            "first_success_time_seconds": null,
+            "grasp_seen": true,
+            "lift_dwell_seconds": 0.0,
+            "lift_seen": false,
+            "max_carry_contact_loss_seconds": 0.0,
+            "max_object_lift_m": 0.030235768269076563,
+            "max_pad_normal_force_n": 2.8633099470994976,
+            "retention_lost": false,
+            "success": false,
+            "unexpected_contact_native_ticks": 3144
+          },
+          "reason": "invalid_state_value",
+          "restoration_scope": "Loaded contact restored for 100 ms after sustained loss; does not establish task recovery or airborne regrasp.",
+          "schema": "dvidia.piper-feedback-score/v1",
+          "strict_v01": {
+            "reason": "invalid_state_value",
+            "reference_xy_m": [
+              0.4,
+              -0.08
+            ],
+            "state_index": 735,
+            "success": false
+          },
+          "success": false,
+          "terminal_only": {
+            "dwell_s": 0.0,
+            "scope": "Terminal-only geometry/release/settling/dwell; excludes lift, carry and retention history.",
+            "success": false,
+            "terminal_xy_error_m": 0.10614903247737481
+          },
+          "valid": true
+        },
+        "source_authenticated": false,
+        "status": "complete",
+        "strict_success": false
+      },
+      "integrity_file": {
+        "bytes": 567674,
+        "sha256": "99a748732c1cbb7761ec00b786a680da20f01b0abb038214451be42d069cd8e2"
+      },
+      "perception": {
+        "available": 508,
+        "decisions": 1000,
+        "longest_missing_seconds": 4.8,
+        "missing": 492
+      },
+      "policy": {
+        "model_sha256": "4545be31f50267c9da29b211510c7a32df500678ba107c7a100604bb0c4c18ab",
+        "physical_robot_ready": false,
+        "policy_origin": "camera_encoder_reference_endpoint_numpy_mlp",
+        "runtime_teacher_available": false,
+        "student_inference_calls": 1000,
+        "student_inference_p95_seconds": 0.0002611646079458296,
+        "student_inference_seconds": 0.23262492148205638,
+        "task_qualified": false,
+        "teacher_fallback_calls": 0,
+        "timing_scope": "Causal feature encoding, NumPy inference and reference projection; excludes camera/render/actuation.",
+        "variant": "delta"
+      },
+      "qualified": false,
+      "recorded_complete_intervals": 1000,
+      "rollout_error": null,
+      "rollout_file": {
+        "bytes": 326951,
+        "sha256": "f0431561b7049a1058db91cc36bcf13cc7626ab9bbc50e57e0244fac52a9f563"
+      },
+      "rollout_status": "complete",
+      "rollout_wall_seconds": 12.816186875017593,
+      "score": {
+        "evidence_scope": "Recomputed simulator-authored numeric evidence; no physics rerun, source attestation or hardware sensing.",
+        "false_alarms": {
+          "reason": "Continuous residual policy has no declared discrete alarm events.",
+          "status": "not_applicable"
+        },
+        "grip_events": [
+          {
+            "kind": "supported_grip_interruption",
+            "loss_onset_s": 13.27699999999808,
+            "loss_qualified_s": 13.29599999999807,
+            "onset_to_qualified_restoration_s": 0.5129999999997157,
+            "onset_to_restoration_s": 0.41399999999977055,
+            "relative_motion_since_loss_m": 0.04689580619513455,
+            "restoration_onset_s": 13.690999999997851,
+            "restoration_qualified_s": 13.789999999997796,
+            "restoration_supported": false,
+            "restored": true,
+            "supported_at_onset": true
+          },
+          {
+            "kind": "supported_grip_interruption",
+            "loss_onset_s": 13.949999999997708,
+            "loss_qualified_s": 13.968999999997697,
+            "onset_to_qualified_restoration_s": null,
+            "onset_to_restoration_s": null,
+            "relative_motion_since_loss_m": 0.306676529975985,
+            "restoration_onset_s": null,
+            "restoration_qualified_s": null,
+            "restored": false,
+            "supported_at_onset": true
+          }
+        ],
+        "grip_summary": {
+          "airborne_events": 0,
+          "loss_events": 2,
+          "qualification_latency_s": [
+            0.5129999999997157
+          ],
+          "restored_events": 1,
+          "supported_events": 2,
+          "unresolved_events": 1
+        },
+        "native_diagnostics": {
+          "premature_open_jaw": {
+            "count": 0,
+            "intervals": [],
+            "maximum_s": 0.0,
+            "total_s": 0.0
+          },
+          "support_touchdown_before_destination": {
+            "count": 0,
+            "intervals": [],
+            "maximum_s": 0.0,
+            "total_s": 0.0
+          }
+        },
+        "native_metrics_mismatches": [],
+        "native_recomputed": {
+          "carry_complete": false,
+          "carry_contact_loss_events": 0,
+          "dwell_seconds": 0.0,
+          "first_success_time_seconds": null,
+          "grasp_seen": true,
+          "lift_dwell_seconds": 0.0,
+          "lift_seen": false,
+          "max_carry_contact_loss_seconds": 0.0,
+          "max_object_lift_m": 0.030235768269076563,
+          "max_pad_normal_force_n": 2.8633099470994976,
+          "retention_lost": false,
+          "success": false,
+          "unexpected_contact_native_ticks": 3144
+        },
+        "reason": "invalid_state_value",
+        "restoration_scope": "Loaded contact restored for 100 ms after sustained loss; does not establish task recovery or airborne regrasp.",
+        "schema": "dvidia.piper-feedback-score/v1",
+        "strict_v01": {
+          "reason": "invalid_state_value",
+          "reference_xy_m": [
+            0.4,
+            -0.08
+          ],
+          "state_index": 735,
+          "success": false
+        },
+        "success": false,
+        "terminal_only": {
+          "dwell_s": 0.0,
+          "scope": "Terminal-only geometry/release/settling/dwell; excludes lift, carry and retention history.",
+          "success": false,
+          "terminal_xy_error_m": 0.10614903247737481
+        },
+        "valid": true
+      },
+      "scored_control_trajectory": true,
+      "strict_success": false,
+      "supervisor_cause": null,
+      "supervisor_status": "finished",
+      "variant": "delta",
+      "worker_receipt": {
+        "bytes": 5207,
+        "sha256": "8f5058b68c1541c1942702bdb95d1a697199b90efad02eab7afa86f0452bc8c5"
+      },
+      "worker_receipt_available": true,
+      "worker_wall_seconds": 14.523490916006267
+    }
+  ],
+  "by_variant": {
+    "absolute": {
+      "callback_including_guard": {
+        "calls": 4000,
+        "maximum_seconds": 0.008621292014140636,
+        "median_seconds": 0.00023227051133289933,
+        "p95_seconds": 0.00029546011064667254
+      },
+      "carry_complete": 0,
+      "complete_episodes": 4,
+      "critical_contact_ticks": 7606,
+      "episodes_with_critical_contact": 2,
+      "full_boundary": {
+        "calls": 4000,
+        "maximum_seconds": 0.051054958981694654,
+        "median_seconds": 0.013028145520365797,
+        "p95_seconds": 0.022362537491426335
+      },
+      "grasp_seen": 3,
+      "lift_seen": 1,
+      "qualified": false,
+      "retention_lost": 0,
+      "scheduled": 4,
+      "scored_control_trajectories": 4,
+      "strict_successes": 0,
+      "terminal_only_successes": 0
+    },
+    "delta": {
+      "callback_including_guard": {
+        "calls": 4000,
+        "maximum_seconds": 0.021589000010862947,
+        "median_seconds": 0.00023095900542102754,
+        "p95_seconds": 0.0002842972404323518
+      },
+      "carry_complete": 0,
+      "complete_episodes": 4,
+      "critical_contact_ticks": 19751,
+      "episodes_with_critical_contact": 4,
+      "full_boundary": {
+        "calls": 4000,
+        "maximum_seconds": 0.058383833995321766,
+        "median_seconds": 0.013461853988701478,
+        "p95_seconds": 0.02204904312093276
+      },
+      "grasp_seen": 4,
+      "lift_seen": 0,
+      "qualified": false,
+      "retention_lost": 0,
+      "scheduled": 4,
+      "scored_control_trajectories": 4,
+      "strict_successes": 0,
+      "terminal_only_successes": 0
+    }
+  },
+  "decision": "pass_integrity_and_accounting",
+  "evidence_scope": "Independent full artifact inventory hash/contract/authority checks and strict scorer reconstruction from saved native arrays, without physics replay or source authentication.",
+  "models": {
+    "absolute": {
+      "content_sha256": "36af3fe1670494b04c48b5114d9be1a5f91632aa7992ba529d6587969ea37460",
+      "file": {
+        "bytes": 539345,
+        "sha256": "3dbf6911d92a9072297c9c3bb7c09f45809808272455ffd5562f4553a173f362"
+      }
+    },
+    "delta": {
+      "content_sha256": "4545be31f50267c9da29b211510c7a32df500678ba107c7a100604bb0c4c18ab",
+      "file": {
+        "bytes": 538021,
+        "sha256": "fece6d056e50a1bde41bdb9960bf7cbc81d983c4bfdb03a07632e30b712b1489"
+      }
+    }
+  },
+  "original_result": {
+    "bytes": 7153,
+    "sha256": "597277fd6c4671a7d2791aa861beb5da896573fcbf51f8db08e30d4bc501284e"
+  },
+  "passed": true,
+  "plan_sha256": "9912b3b86abe889d81073b1fee1c9cad9655264fdb325abbbf60dbebd89a490b",
+  "prior_batches": [
+    {
+      "result_file": {
+        "bytes": 6130,
+        "sha256": "2e9a40eaa5799f3bc6d0d36ee10f5fe8a6cf0bfe387f428f399d66a4e9c5c6fe"
+      },
+      "run_id": "reference-model-development-smoke-20261109",
+      "scheduled_attempts": 8
+    },
+    {
+      "result_file": {
+        "bytes": 6664,
+        "sha256": "017d7ae05f64b8539d950ec00063345f30571e4a89086e06411b1b038c9a335c"
+      },
+      "run_id": "reference-model-development-smoke-v2-20261109",
+      "scheduled_attempts": 8
+    }
+  ],
+  "qualification": false,
+  "result_file": {
+    "bytes": 7153,
+    "sha256": "597277fd6c4671a7d2791aa861beb5da896573fcbf51f8db08e30d4bc501284e"
+  },
+  "review_fits": 0,
+  "review_physics_or_native_import": false,
+  "review_policy_inference": false,
+  "review_wall_seconds": 9.043949083017651,
+  "reviewed_at_utc": "2026-11-09T16:42:28.047717+00:00",
+  "reviewer": "export_cache",
+  "schema": "dvidia.reference-model-development-smoke-v3-result-review/v1",
+  "script_sha256": "8186572b17949cbe3add1ac4d1bc4dcca92cfa76f973892b64e1b4275b78c48a",
+  "selection_performed": false,
+  "source_commit": "78ea4fbb4b2733e7738eb425eeecd7cb67a52195",
+  "source_files": {
+    "src/dvidia_training/action_demonstration.py": {
+      "bytes": 36303,
+      "sha256": "439c7cce558571221660cd51e7dba1675203237706ea8ea0f97f2fcf9a6a24c5"
+    },
+    "src/dvidia_training/camera_action_cache.py": {
+      "bytes": 23490,
+      "sha256": "9191f2eb9f96bd09d4f711c31966a583bd938853a874e6018f8ddebfb0bc00d8"
+    },
+    "src/dvidia_training/camera_action_capture.py": {
+      "bytes": 34257,
+      "sha256": "399f512a08087c72bfa37e382157a2250f413e423974b77b1bd3c0a911ff0a1e"
+    },
+    "src/dvidia_training/camera_action_export.py": {
+      "bytes": 7355,
+      "sha256": "2c4e30587605edc0342150a93430a70ed0559c3ee3eda17d93c5670831370b02"
+    },
+    "src/dvidia_training/camera_action_model.py": {
+      "bytes": 27066,
+      "sha256": "e9a9c4d1d0f16fc7c91cac3ca2e6560d0532370ab78a2af5a4ec84425c23932f"
+    },
+    "src/dvidia_training/camera_action_reference_export.py": {
+      "bytes": 19195,
+      "sha256": "ccf6f74f6100721868ccece67e5b27dae066c1b644c8297efe111125d9517753"
+    },
+    "src/dvidia_training/camera_action_reference_model.py": {
+      "bytes": 19771,
+      "sha256": "952750fd6b99f97eb96054b8fe50710b83f7e8eb4e069601be442292dd0023bc"
+    },
+    "src/dvidia_training/camera_action_study.py": {
+      "bytes": 46551,
+      "sha256": "196f669b03baff3df561fa9236c64dab401c954c929994dedc36721611c8948a"
+    },
+    "src/dvidia_training/embodiment_calibration.py": {
+      "bytes": 14701,
+      "sha256": "0531be511b3bc549576a33930394d33b673c146d5e52c29198096632ea7478c5"
+    },
+    "src/dvidia_training/episode_tape.py": {
+      "bytes": 31209,
+      "sha256": "abae9a4bd0080c7f638c5c82c050fb1364a20769998c054354b918233669f5d2"
+    },
+    "src/dvidia_training/piper_action_capture.py": {
+      "bytes": 29069,
+      "sha256": "da037b2198cf7d2f3a9aa40fc19a4c0f5cc375a22c5dd75606e2c1856f6efc7f"
+    },
+    "src/dvidia_training/piper_feedback_env.py": {
+      "bytes": 11247,
+      "sha256": "0dbae423231968a81bdeed52226d8dffaf8df622013ed3a360d7217148f1bcac"
+    },
+    "src/dvidia_training/piper_feedback_score.py": {
+      "bytes": 18866,
+      "sha256": "863a272828208c631de4cf57647e973eb3e4f5e99eb7a1eacb4eaa68dc42eb0e"
+    },
+    "src/dvidia_training/piper_imitation.py": {
+      "bytes": 25561,
+      "sha256": "849f66512c58441a240cd20230fcc58691234ddd254c5bbcbfd76abc21b14868"
+    },
+    "src/dvidia_training/piper_imitation_score.py": {
+      "bytes": 11462,
+      "sha256": "0cd4663abfdd7f14f2f33d93e9a2bd45122b998d6160d45b8afcc880512cba2b"
+    },
+    "src/dvidia_training/piper_motor_primitive.py": {
+      "bytes": 17476,
+      "sha256": "02a45b88913db6f223e4d9975eac64d4942a92739c5817c6ede4cc9fa332ed1c"
+    },
+    "src/dvidia_training/piper_rgbd_tracker.py": {
+      "bytes": 17461,
+      "sha256": "43202cfcea404061a3a65124d6ad24aeeec5fae8ffaf863caa245f2f54c46c36"
+    },
+    "src/dvidia_training/piper_task.py": {
+      "bytes": 34355,
+      "sha256": "e66a44a8df4e7a8ee42e4f3866f889f8797cdfb8e2688a146e09b75d2c76664f"
+    },
+    "src/dvidia_training/reference_sim.py": {
+      "bytes": 24336,
+      "sha256": "c1201d22eb13dd07a236ee63cc4ba68e129f6067e8db61ed41af2d820eb5e3b8"
+    },
+    "src/dvidia_training/reference_task.py": {
+      "bytes": 34166,
+      "sha256": "bbc768ccc202b8dd41cbd235b206e6478806ce47c1f30ae15dc662c0c1edcc3e"
+    },
+    "src/dvidia_training/robot_profiles.lock.json": {
+      "bytes": 35002,
+      "sha256": "4b2ee2ca675aceb2f826983cb3082a3e08e9aa0ff419b23078abcff08ebcc6fe"
+    },
+    "src/dvidia_training/robot_profiles.py": {
+      "bytes": 26493,
+      "sha256": "fb90782a6a9979e282182118ae0e3e29d00f101255cc57378dd3f692672bb13d"
+    }
+  },
+  "timing_scope": "Stored rollout timings are shared-host offline paused-simulator wall measurements. Callback includes guard+causal encoding+inference+projection; ModelPolicy diagnostic excludes guard. Full boundary includes file writes and alternate camera work; initial acquisition timing row is separate and excluded from callback/boundary percentiles. Neither is physical delivery latency. Review verification time is separate."
+}

evidence/renderer-warnings.json

@@ -0,0 +1,71 @@
+{
+  "attempts": [
+    {
+      "id": "development-00-goal0-absolute",
+      "renderer_warnings": [
+        "ARB_clip_control unavailable while mjDEPTH_ZEROFAR requested, depth accuracy will be limited"
+      ],
+      "rollout_sha256": "1104da3991eb6c3872c85c46eec6d4a5dcb89c0a10752d57c5ebc8071b14dee9",
+      "timing_entries_with_renderer_warnings": 501
+    },
+    {
+      "id": "development-00-goal0-delta",
+      "renderer_warnings": [
+        "ARB_clip_control unavailable while mjDEPTH_ZEROFAR requested, depth accuracy will be limited"
+      ],
+      "rollout_sha256": "328cf546022fd2f79da3ec1a6161c0d3ebd777bd254edd89903a2629966a5c8e",
+      "timing_entries_with_renderer_warnings": 501
+    },
+    {
+      "id": "development-00-goal1-absolute",
+      "renderer_warnings": [
+        "ARB_clip_control unavailable while mjDEPTH_ZEROFAR requested, depth accuracy will be limited"
+      ],
+      "rollout_sha256": "8b851f39d22f39b02709c7857d1c2a6afa2420e83d200e7ee9564baeeba82b07",
+      "timing_entries_with_renderer_warnings": 501
+    },
+    {
+      "id": "development-00-goal1-delta",
+      "renderer_warnings": [
+        "ARB_clip_control unavailable while mjDEPTH_ZEROFAR requested, depth accuracy will be limited"
+      ],
+      "rollout_sha256": "83dee39e2cb1cbfd3ef72c741d854627bf54c021852ffb1430c66fe4b8fe8452",
+      "timing_entries_with_renderer_warnings": 501
+    },
+    {
+      "id": "development-01-goal0-absolute",
+      "renderer_warnings": [
+        "ARB_clip_control unavailable while mjDEPTH_ZEROFAR requested, depth accuracy will be limited"
+      ],
+      "rollout_sha256": "47f735604508f95ad78f9d863ca88f13f77b0bb4f68a4fcc4b067b967704fb97",
+      "timing_entries_with_renderer_warnings": 501
+    },
+    {
+      "id": "development-01-goal0-delta",
+      "renderer_warnings": [
+        "ARB_clip_control unavailable while mjDEPTH_ZEROFAR requested, depth accuracy will be limited"
+      ],
+      "rollout_sha256": "1302c51513d638910857c79bc78539bc9c59b546aeaf212d08b922ec06372a21",
+      "timing_entries_with_renderer_warnings": 501
+    },
+    {
+      "id": "development-01-goal1-absolute",
+      "renderer_warnings": [
+        "ARB_clip_control unavailable while mjDEPTH_ZEROFAR requested, depth accuracy will be limited"
+      ],
+      "rollout_sha256": "bb86dd380867955a60df4588140dc5c07e4f49129197c09e6b7fae9c64ce0d77",
+      "timing_entries_with_renderer_warnings": 501
+    },
+    {
+      "id": "development-01-goal1-delta",
+      "renderer_warnings": [
+        "ARB_clip_control unavailable while mjDEPTH_ZEROFAR requested, depth accuracy will be limited"
+      ],
+      "rollout_sha256": "f0431561b7049a1058db91cc36bcf13cc7626ab9bbc50e57e0244fac52a9f563",
+      "timing_entries_with_renderer_warnings": 501
+    }
+  ],
+  "schema": "dvidia.reference-model-renderer-warning-summary/v1",
+  "scope": "Warnings retained from sealed rollout timing records. No claim of calibrated depth accuracy. Physics/scorer results remain unchanged.",
+  "source_result_sha256": "597277fd6c4671a7d2791aa861beb5da896573fcbf51f8db08e30d4bc501284e"
+}

evidence/training-input-identity.json

@@ -0,0 +1,214 @@
+{
+  "episodes": 96,
+  "groups": 32,
+  "initial_parameters_sha256": "9c50504160213ad5241d6d96ab9be938d6f49b1c4bf49ff39d0770dac83dce7e",
+  "model_source_identity_sha256": "12ea9757cb0d76573c2e67f80b56ab08271f95e43f18b344b6481cc9153d70bc",
+  "parent_dataset_sha256": "b500b3ba348bba44cc79e377266f92910bd7e0165c36cd8b2fe260909b30a2f9",
+  "parent_provenance_sha256": "52cbeda62e3aa50e8cd6a34688fba7583f9fcec927cb0eb434a6907c99695f4f",
+  "raw_recordings_revalidated_during_integration": false,
+  "reference_dataset_sha256": "a9dcd66a81f177bb1910e08171ac7607f115ed47b51d93412107eb0e204aed73",
+  "samples": 96000,
+  "schema": "dvidia.reference-model-training-input-identity/v1",
+  "source_commit": "78ea4fbb4b2733e7738eb425eeecd7cb67a52195",
+  "split": "train",
+  "stages": [
+    {
+      "arrays": {
+        "feasible-targets.npy": {
+          "bytes": 3584128,
+          "dtype": "<f8",
+          "sha256": "7d344cb0c025b97136eeb7aede9bb4b1d550ec9bc4dc9e5cbe2fc44bcf428eb4",
+          "shape": [
+            64000,
+            7
+          ]
+        },
+        "previous-references.npy": {
+          "bytes": 3584128,
+          "dtype": "<f8",
+          "sha256": "ba60ac17901651f5d9e141d39e7cc2a0af5f7bc5fd992b2e77cf93ca90c0a1c0",
+          "shape": [
+            64000,
+            7
+          ]
+        }
+      },
+      "contract": {
+        "action_units": [
+          "rad",
+          "rad",
+          "rad",
+          "rad",
+          "rad",
+          "rad",
+          "single_finger_m"
+        ],
+        "actual_pulse_overridden_controls_are_not_references": true,
+        "clock_tolerance_s": 1e-09,
+        "control_dt_s": 0.02,
+        "endpoint_algorithm": "20 original float64 reference updates, not closed-form projection",
+        "jaw_rate_m_s": 0.05,
+        "joint_order": [
+          "joint1",
+          "joint2",
+          "joint3",
+          "joint4",
+          "joint5",
+          "joint6"
+        ],
+        "joint_rate_rad_s": 1.5,
+        "later_source": "steps[i-1].applied.ramped_targets",
+        "limiter_ast_sha256": "a5356fb9e1bd8b1dc6e6b2b14d657fbe625e018ba8d1e5e668ba4d11191f7921",
+        "limits_native": [
+          [
+            -2.618,
+            2.618
+          ],
+          [
+            0.0,
+            3.14
+          ],
+          [
+            -2.697,
+            0.0
+          ],
+          [
+            -1.832,
+            1.832
+          ],
+          [
+            -1.22,
+            1.22
+          ],
+          [
+            -3.14,
+            3.14
+          ],
+          [
+            0.0,
+            0.035
+          ]
+        ],
+        "microsteps": 20,
+        "numerical_equivalence_tolerance_native": 1e-12,
+        "reference_kind": "previous_applied_normal_servo_reference",
+        "reset_source": "initial_applied_targets",
+        "timestep_s": 0.001
+      },
+      "episodes": 64,
+      "parent_dataset_sha256": "07b175f78e4e77cbdde9a461a4bcae4dcb0db7d389ee4f2e702ff28db8bb3dfc",
+      "parent_provenance_sha256": "7103ad53111cb114cdcd164cfa15bbbf359ebc4ae66e7456ab7c1f38820b2864",
+      "parent_target_sha256": "de98cecd692af8de739bef8bdf0044387d6e92ee1c24945b4de4a5920b4bca6e",
+      "pins": {
+        "expected_manifest_sha256": "48ff59ec719c70b0ae8e42ab2f14e5664e76f135347281570d94b3ceaf36b86c",
+        "expected_parent_manifest_sha256": "8acd728fcc8ed637b58138c3bd7466432a588fa439acefde4b3c90fa1adfdbaf",
+        "expected_parent_source_sha256": "0ab01145c66d651eaca96d763b4bec1706327f2cdd4bd8665d05056cbdd6f677",
+        "expected_protocol_sha256": "576851e321034a0f6d3cb4d0818b3a5ca97975fcc6f2c7fa30b07ffd203a2d76",
+        "expected_source_sha256": "e5b26c86013732007c483a23ad7218271980f0f03563c4ffd95c01ad87fd3041",
+        "expected_split": "train",
+        "expected_stage": "demonstrations"
+      },
+      "raw_source_reverified": false,
+      "samples": 64000
+    },
+    {
+      "arrays": {
+        "feasible-targets.npy": {
+          "bytes": 1792128,
+          "dtype": "<f8",
+          "sha256": "c845a7e50949b10b4b5a2ac60725443b864563017be53e924a0a608012d5b96c",
+          "shape": [
+            32000,
+            7
+          ]
+        },
+        "previous-references.npy": {
+          "bytes": 1792128,
+          "dtype": "<f8",
+          "sha256": "6a150b791c01b79210319b4e9bf45b6c541c3b5ea0955e1318be56f7b260199d",
+          "shape": [
+            32000,
+            7
+          ]
+        }
+      },
+      "contract": {
+        "action_units": [
+          "rad",
+          "rad",
+          "rad",
+          "rad",
+          "rad",
+          "rad",
+          "single_finger_m"
+        ],
+        "actual_pulse_overridden_controls_are_not_references": true,
+        "clock_tolerance_s": 1e-09,
+        "control_dt_s": 0.02,
+        "endpoint_algorithm": "20 original float64 reference updates, not closed-form projection",
+        "jaw_rate_m_s": 0.05,
+        "joint_order": [
+          "joint1",
+          "joint2",
+          "joint3",
+          "joint4",
+          "joint5",
+          "joint6"
+        ],
+        "joint_rate_rad_s": 1.5,
+        "later_source": "steps[i-1].applied.ramped_targets",
+        "limiter_ast_sha256": "a5356fb9e1bd8b1dc6e6b2b14d657fbe625e018ba8d1e5e668ba4d11191f7921",
+        "limits_native": [
+          [
+            -2.618,
+            2.618
+          ],
+          [
+            0.0,
+            3.14
+          ],
+          [
+            -2.697,
+            0.0
+          ],
+          [
+            -1.832,
+            1.832
+          ],
+          [
+            -1.22,
+            1.22
+          ],
+          [
+            -3.14,
+            3.14
+          ],
+          [
+            0.0,
+            0.035
+          ]
+        ],
+        "microsteps": 20,
+        "numerical_equivalence_tolerance_native": 1e-12,
+        "reference_kind": "previous_applied_normal_servo_reference",
+        "reset_source": "initial_applied_targets",
+        "timestep_s": 0.001
+      },
+      "episodes": 32,
+      "parent_dataset_sha256": "0561f8793ac6f68318d28c99e15ee8755c85412f3e34c990e1c4f0da4a5c1789",
+      "parent_provenance_sha256": "0267af1dcc2035453f35a8fcb862bb190a6f2bdb50efb516fe7213f21d392b6a",
+      "parent_target_sha256": "49b2f30ed513b2f575186648b6aac169b24e07021adefaa8e3ace76e50733c1c",
+      "pins": {
+        "expected_manifest_sha256": "dfbb096632e30527cf53bcf9e2627f31e94ad7a669f1820e295cda1b49546334",
+        "expected_parent_manifest_sha256": "0997805831dfaacfe38d87468587f58c4e0b3f748165816e2167c4376676b96b",
+        "expected_parent_source_sha256": "0ab01145c66d651eaca96d763b4bec1706327f2cdd4bd8665d05056cbdd6f677",
+        "expected_protocol_sha256": "576851e321034a0f6d3cb4d0818b3a5ca97975fcc6f2c7fa30b07ffd203a2d76",
+        "expected_source_sha256": "e5b26c86013732007c483a23ad7218271980f0f03563c4ffd95c01ad87fd3041",
+        "expected_split": "train",
+        "expected_stage": "corrections"
+      },
+      "raw_source_reverified": false,
+      "samples": 32000
+    }
+  ]
+}

models/absolute/README.md

@@ -0,0 +1,81 @@
+---
+license: mit
+language:
+- en
+tags:
+- robotics
+- simulation
+- numpy
+- research-candidate
+---
+
+# DVIDIA PiPER reference model — Cabs
+
+**Research candidate; not a qualified skill or hardware controller.** This card documents the retained `absolute` candidate from the separately registered training/inference integration benchmark. A separately registered four-case development smoke completed for this candidate and achieved 0/4 strict placements. It remains unqualified.
+
+The two candidates share the same 269-input, two-layer 64-unit tanh network, feature normalization, seeded initialization and optimizer recipe. Training uses 96,000 rows from 64 authored-teacher simulation demonstrations and 32 earlier mixed-control correction episodes, all within the original training split. Both candidates are retained; no model selection was performed.
+
+Inputs include an explicitly historical initial visual estimate, destination, elapsed time, joint/gripper measurements, encoder-derived tool pose, previous applied commands and 160 ms of causal observation history with availability masks and age. Current teacher commands, teacher phase, native object/contact truth and future observations are excluded. Exact float64 prior normal servo references are separate decoder state; they are not reconstructed from float32 features.
+
+The labels are reachable normal-reference endpoints computed by the original 20 limiter updates. Mixed-control teacher queries are counterfactual supervision; they are not relabeled as commands that were actually executed. This experiment does not recover actions from ordinary human videos.
+
+The decoder is `r0 + D*g(x)`. `r0` is the fixed reset reference; `r` is the actual retained previous normal servo reference. `D` is 0.03 rad per joint and 0.001 m for the single-finger jaw coordinate. Runtime projection uses native bounds and the original 20 normal reference updates. Low training error does not establish stable feedback behavior.
+
+## Recorded integration measurements
+
+| Measurement | Value |
+|---|---:|
+| Training rows | 96,000 |
+| Fit wall time | 5.230713 s |
+| Model JSON bytes | 539,345 |
+| Initial training loss | 0.0400907266332 |
+| Final training loss | 7.9261088837e-06 |
+| Open-loop callback median | 0.181813 ms |
+| Open-loop callback p95 | 0.204840 ms |
+| Serialized policy outputs | Bitwise identical at all 1,000 replay steps |
+
+One fit used seed 17, 50 epochs, batch size 512 and Adam learning rate 0.003. Loss is the mean squared unprojected endpoint error normalized by scales of 3 rad per joint and 0.035 m for the jaw; it is a training-set descriptive metric. The loss includes no task-success scorer.
+
+The 1,000-step replay follows the first protocol-ordered training teacher trajectory. Every next observation and reference remains the recorded teacher value. Predictions never enter simulation or feed the next state. Callback timing includes causal encoding, NumPy inference and reference projection; it excludes camera acquisition, rendering, physics and transport. Fit time includes source/derived-input validation and normalization. The shared host ran macOS 26.5.2, arm64, Python 3.12.9 and NumPy 2.5.3 with declared BLAS threads 1. Whole-process peak RSS across both fits and replays was 1,345,404,928 bytes; this is not a hardware minimum or model-only memory requirement.
+
+## Artifact and source identity
+
+- Artifact: [model.json](model.json), copied byte-for-byte from the retained integration result.
+- File SHA-256: `3dbf6911d92a9072297c9c3bb7c09f45809808272455ffd5562f4553a173f362`.
+- Canonical model content SHA-256: `36af3fe1670494b04c48b5114d9be1a5f91632aa7992ba529d6587969ea37460`.
+- [Pinned source](https://github.com/Dvidia-Inference/dvidia-training/tree/78ea4fbb4b2733e7738eb425eeecd7cb67a52195); model source identity `12ea9757cb0d76573c2e67f80b56ab08271f95e43f18b344b6481cc9153d70bc`.
+- [Integration receipt](../../evidence/integration-result.json) and [training input identity](../../evidence/training-input-identity.json).
+
+The Python API is `dvidia_training.camera_action_reference_model.load_model` / `ModelPolicy` at the pinned source. Existing qualified-skill installers and arbitrary-arm deployment are not validated for this new model type. The model's `scientific_registration` is intentionally null; the external benchmark registration identifies this operational integration test.
+
+## Native evaluation and limitations
+
+Two earlier eight-attempt batches remain retained separately: import-guard refusals, then CoreGraphics renderer refusals. A tiny static-frame preflight isolated the graphics access requirement. A third, separately registered batch used the same models, source, cases, cadence and guards with scoped graphics access. All eight 20-second simulated attempts completed, four per candidate, with zero teacher queries or fallbacks. No training or parameter selection followed these outcomes.
+
+| Native development measurement | This candidate |
+|---|---:|
+| Complete independently checked attempts | 4/4 |
+| Strict placements | 0/4 |
+| Bilateral grasp contact seen | 3/4 |
+| Qualified lift seen | 1/4 |
+| Carry completed | 0/4 |
+| Episodes with unexpected contacts | 2/4 |
+| Unexpected-contact native ticks | 7,606 |
+| Available visual estimates | 2,396/4,000 (59.9%) |
+| Longest missing visual interval | 13.36 s |
+| Callback including storage guard p95 | 0.295460 ms |
+| Full offline boundary p95 | 22.362537 ms |
+
+The absolute candidate reached a qualified lift once, but no carry or placement. Two episodes had unexpected contacts.
+
+These are four development cases per model, with no fixed/teacher/blind comparator, negative controls or uncertainty intervals in this engineering batch. They cannot establish generalization or relative superiority. Zero recorded retention-loss events cannot establish grip reliability when so few episodes reach the scorer's post-lift window.
+
+The shared-host callback distribution includes the storage guard; the model's diagnostic timer excludes it. Full-boundary wall timing includes file writes and alternate camera/tracker work. The simulator pauses for that work; this is not a real-time robot delivery measurement. The renderer warned that `ARB_clip_control` was unavailable for the requested depth mode, limiting depth accuracy. This backend was not depth-calibrated by the smoke test.
+
+[Exact eight-attempt audit](../../evidence/native-smoke-v3-audit.json), [all three batches](../../evidence/development-smoke-status.json), [renderer warnings](../../evidence/renderer-warnings.json) and [graphics preflight](../../evidence/graphics-preflight.json) retain the evidence and limitations.
+
+No confirmation-set evaluation, generalization claim, candidate selection, physical robot action or qualification has occurred. The frozen v0.1 result and the proposed 164-attempt screen remain unchanged. This model applies only to the bounded simulated PiPER block-placement setup and its exact sensing/reference contracts.
+
+## License
+
+DVIDIA candidate weights and accompanying DVIDIA code/documentation are offered under [MIT](../../LICENSE). Upstream dependencies and robot assets retain their own licenses; this benchmark does not bundle them.

models/delta/README.md

@@ -0,0 +1,81 @@
+---
+license: mit
+language:
+- en
+tags:
+- robotics
+- simulation
+- numpy
+- research-candidate
+---
+
+# DVIDIA PiPER reference model — Cdelta
+
+**Research candidate; not a qualified skill or hardware controller.** This card documents the retained `delta` candidate from the separately registered training/inference integration benchmark. A separately registered four-case development smoke completed for this candidate and achieved 0/4 strict placements. It remains unqualified.
+
+The two candidates share the same 269-input, two-layer 64-unit tanh network, feature normalization, seeded initialization and optimizer recipe. Training uses 96,000 rows from 64 authored-teacher simulation demonstrations and 32 earlier mixed-control correction episodes, all within the original training split. Both candidates are retained; no model selection was performed.
+
+Inputs include an explicitly historical initial visual estimate, destination, elapsed time, joint/gripper measurements, encoder-derived tool pose, previous applied commands and 160 ms of causal observation history with availability masks and age. Current teacher commands, teacher phase, native object/contact truth and future observations are excluded. Exact float64 prior normal servo references are separate decoder state; they are not reconstructed from float32 features.
+
+The labels are reachable normal-reference endpoints computed by the original 20 limiter updates. Mixed-control teacher queries are counterfactual supervision; they are not relabeled as commands that were actually executed. This experiment does not recover actions from ordinary human videos.
+
+The decoder is `r + D*g(x)`. `r0` is the fixed reset reference; `r` is the actual retained previous normal servo reference. `D` is 0.03 rad per joint and 0.001 m for the single-finger jaw coordinate. Runtime projection uses native bounds and the original 20 normal reference updates. Low training error does not establish stable feedback behavior.
+
+## Recorded integration measurements
+
+| Measurement | Value |
+|---|---:|
+| Training rows | 96,000 |
+| Fit wall time | 5.097628 s |
+| Model JSON bytes | 538,021 |
+| Initial training loss | 8.72223173973e-05 |
+| Final training loss | 3.53132778774e-06 |
+| Open-loop callback median | 0.182729 ms |
+| Open-loop callback p95 | 0.207383 ms |
+| Serialized policy outputs | Bitwise identical at all 1,000 replay steps |
+
+One fit used seed 17, 50 epochs, batch size 512 and Adam learning rate 0.003. Loss is the mean squared unprojected endpoint error normalized by scales of 3 rad per joint and 0.035 m for the jaw; it is a training-set descriptive metric. The loss includes no task-success scorer.
+
+The 1,000-step replay follows the first protocol-ordered training teacher trajectory. Every next observation and reference remains the recorded teacher value. Predictions never enter simulation or feed the next state. Callback timing includes causal encoding, NumPy inference and reference projection; it excludes camera acquisition, rendering, physics and transport. Fit time includes source/derived-input validation and normalization. The shared host ran macOS 26.5.2, arm64, Python 3.12.9 and NumPy 2.5.3 with declared BLAS threads 1. Whole-process peak RSS across both fits and replays was 1,345,404,928 bytes; this is not a hardware minimum or model-only memory requirement.
+
+## Artifact and source identity
+
+- Artifact: [model.json](model.json), copied byte-for-byte from the retained integration result.
+- File SHA-256: `fece6d056e50a1bde41bdb9960bf7cbc81d983c4bfdb03a07632e30b712b1489`.
+- Canonical model content SHA-256: `4545be31f50267c9da29b211510c7a32df500678ba107c7a100604bb0c4c18ab`.
+- [Pinned source](https://github.com/Dvidia-Inference/dvidia-training/tree/78ea4fbb4b2733e7738eb425eeecd7cb67a52195); model source identity `12ea9757cb0d76573c2e67f80b56ab08271f95e43f18b344b6481cc9153d70bc`.
+- [Integration receipt](../../evidence/integration-result.json) and [training input identity](../../evidence/training-input-identity.json).
+
+The Python API is `dvidia_training.camera_action_reference_model.load_model` / `ModelPolicy` at the pinned source. Existing qualified-skill installers and arbitrary-arm deployment are not validated for this new model type. The model's `scientific_registration` is intentionally null; the external benchmark registration identifies this operational integration test.
+
+## Native evaluation and limitations
+
+Two earlier eight-attempt batches remain retained separately: import-guard refusals, then CoreGraphics renderer refusals. A tiny static-frame preflight isolated the graphics access requirement. A third, separately registered batch used the same models, source, cases, cadence and guards with scoped graphics access. All eight 20-second simulated attempts completed, four per candidate, with zero teacher queries or fallbacks. No training or parameter selection followed these outcomes.
+
+| Native development measurement | This candidate |
+|---|---:|
+| Complete independently checked attempts | 4/4 |
+| Strict placements | 0/4 |
+| Bilateral grasp contact seen | 4/4 |
+| Qualified lift seen | 0/4 |
+| Carry completed | 0/4 |
+| Episodes with unexpected contacts | 4/4 |
+| Unexpected-contact native ticks | 19,751 |
+| Available visual estimates | 2,148/4,000 (53.7%) |
+| Longest missing visual interval | 6.16 s |
+| Callback including storage guard p95 | 0.284297 ms |
+| Full offline boundary p95 | 22.049043 ms |
+
+The delta candidate made bilateral contact in all four episodes but never reached a qualified lift. All four had unexpected contacts and the strict trajectory scorer returned `invalid_state_value`: finite tool-centre coordinates exceeded its unchanged component envelope of ±0.5 m. These were not NaNs or solver warnings. Native metric consistency alone does not imply that the strict trajectory contract passed. [Exact diagnosis](../../evidence/native-smoke-diagnosis.json).
+
+These are four development cases per model, with no fixed/teacher/blind comparator, negative controls or uncertainty intervals in this engineering batch. They cannot establish generalization or relative superiority. Zero recorded retention-loss events cannot establish grip reliability when so few episodes reach the scorer's post-lift window.
+
+The shared-host callback distribution includes the storage guard; the model's diagnostic timer excludes it. Full-boundary wall timing includes file writes and alternate camera/tracker work. The simulator pauses for that work; this is not a real-time robot delivery measurement. The renderer warned that `ARB_clip_control` was unavailable for the requested depth mode, limiting depth accuracy. This backend was not depth-calibrated by the smoke test.
+
+[Exact eight-attempt audit](../../evidence/native-smoke-v3-audit.json), [all three batches](../../evidence/development-smoke-status.json), [renderer warnings](../../evidence/renderer-warnings.json) and [graphics preflight](../../evidence/graphics-preflight.json) retain the evidence and limitations.
+
+No confirmation-set evaluation, generalization claim, candidate selection, physical robot action or qualification has occurred. The frozen v0.1 result and the proposed 164-attempt screen remain unchanged. This model applies only to the bounded simulated PiPER block-placement setup and its exact sensing/reference contracts.
+
+## License
+
+DVIDIA candidate weights and accompanying DVIDIA code/documentation are offered under [MIT](../../LICENSE). Upstream dependencies and robot assets retain their own licenses; this benchmark does not bundle them.