Clarify compact architecture and temporal-ensemble inference

README.md

@@ -9,12 +9,7 @@ tags:
 
 # Bagel-Robotics-1 — checkpoint 1000
 
-Bagel-Robotics-1 is Bagel Labs' approach to inference-friendly compact world-action
-models (WAMs). Our method combines mask-based structured pruning and recovery
-training with effective selection of critically important segments from each task,
-using a focused subset of the training dataset. VPP2 is the base model used to
-validate this approach; reused implementation components are identified in the
-source notices below.
+Bagel-Robotics-1 combines mask-based structured pruning, recovery training and task-critical data selection to produce a compact WAM with nonuniform FFN widths and cross-attention head counts. Its inference pipeline predicts 32 actions, executes 16 before replanning, and applies 90/10 temporal ensembling. VPP2 is the validation base; reused source components are credited in the notices below.
 
 Inference release of the **exact checkpoint 1000** used in the Bagel-Robotics-1 evaluation campaign. **The repository is currently private.** Evaluator access is being arranged using a dedicated fine-grained token with read access only to this repository, delivered through an agreed private channel. The dedicated repository-only read credential is verified; assigned evaluator handoff remains pending. Public release is prepared for the later verified-leaderboard milestone and requires a bageldotcom administrator.
 
@@ -71,4 +66,4 @@ The technical report and training/pruning/export code are planned for **November
 
 ## Attribution and terms
 
-VPP2 is the base model used to validate the Bagel-Robotics-1 pruning, recovery-training and data-selection approach. Reused base-model inference components come from [VPP2](https://github.com/roboterax/video-prediction-policy-2)/FastWAM and Wan2.1/DiffSynth. See [THIRD_PARTY_NOTICES.md](THIRD_PARTY_NOTICES.md) and the adapter's retained MIT/Apache-2.0 notices. Model and auxiliary-asset upstream terms continue to apply; an inference-code license is not a new blanket license for all weights or datasets.
+VPP2 is the base model used to validate the Bagel-Robotics-1 pruning, recovery-training and data-selection approach. Reused source components come from [VPP2](https://github.com/roboterax/video-prediction-policy-2)/FastWAM and Wan2.1/DiffSynth. See [THIRD_PARTY_NOTICES.md](THIRD_PARTY_NOTICES.md) and the adapter's retained MIT/Apache-2.0 notices. Model and auxiliary-asset upstream terms continue to apply; an inference-code license is not a new blanket license for all weights or datasets.

THIRD_PARTY_NOTICES.md

@@ -1,6 +1,6 @@
 # Source and artifact notices
 
-- Base-model inference components adapted from [VPP2](https://github.com/roboterax/video-prediction-policy-2)/FastWAM retain the MIT notice in
+- Reused source components from [VPP2](https://github.com/roboterax/video-prediction-policy-2)/FastWAM retain the MIT notice in
   [LICENSE](inference/policy/bagel_robotics_1/LICENSE), copyright 2026 The FastWAM Authors. See [source-provenance.json](inference/policy/bagel_robotics_1/source-provenance.json)
   for the source identities and packaging changes.
 - Wan2.1/DiffSynth-derived model components retain Apache-2.0 terms; see

inference/README.md

@@ -1,9 +1,6 @@
 # Bagel-Robotics-1 inference adapter
 
-Bagel-Robotics-1 combines mask-based structured pruning and recovery training with
-effective selection of task-critical data to produce an inference-friendly compact
-WAM. VPP2 is the base model used to validate this approach; reused implementation
-components are identified in the policy's source notices.
+Bagel-Robotics-1 combines mask-based structured pruning, recovery training and task-critical data selection to produce a compact WAM with nonuniform FFN widths and cross-attention head counts. Its inference pipeline predicts 32 actions, executes 16 before replanning, and applies 90/10 temporal ensembling. VPP2 is the validation base; reused source components are credited in the policy's source notices.
 
 This folder is the standalone XPolicyLab adapter for Bagel-Robotics-1 checkpoint 1000. It contains inference code and evaluation instructions, with no training repository dependency. Model tensors download separately from `bageldotcom/Bagel-Robotics-1`, pinned to `2651631c5d08819088475e8778676f5cf6d9afcf` and verified against `policy/bagel_robotics_1/artifact-manifest.json`.
 

inference/SHA256SUMS

@@ -1,11 +1,11 @@
-8f9d97e666d57c8b73782c2a3dd65f40fac6bed9c95bc042481eb68bad9b55d2  README.md
+70f520e883d0d471c2317e606867f161285a239da572402b43670c901a5e6d07  README.md
 56f26e5b6ba87080c82703e102240fdfbd955c731fb2a044609ffff099795b3f  install_adapter.py
 6c9437916a0dc5b6ed5fd56b361a4a1894be910e6965c9570e8683cc74a40813  policy/bagel_robotics_1/.gitignore
 24cb153ee8bababbfc20fe7acd36768923fbcf2a07d0aa653f460282761ccf8f  policy/bagel_robotics_1/EVALUATION_RESULTS.md
 8379dddfbbb1154ad9cc57f008f1f8d4a8d26c3de99bd6b36aa4d8067888aa92  policy/bagel_robotics_1/LICENSE
-6bff9af9cf8910c5fdc423d7dc00bf416af27a9dc0ebae3bb88a9d7e00d1cc1b  policy/bagel_robotics_1/README.md
+df95fda0cfe78bf41a15d036166e2f5fba87e23cdb7e100315311bde5e99b6db  policy/bagel_robotics_1/README.md
 f91a4218b180cdef400540135a45e1c94b01e8b4ff1b2f5871c1d6ff328c97b5  policy/bagel_robotics_1/REEVALUATION.md
-de2bcb5caca65d423ec519ec5119e0da65501777ce77d39dcabf6f1c51776271  policy/bagel_robotics_1/THIRD_PARTY_NOTICES.md
+bf0d0498ea83fbc82593de12234892d4458099f36a9ff6d79c920add58640ab8  policy/bagel_robotics_1/THIRD_PARTY_NOTICES.md
 4a4d71a75e893be9ddf802b4e3c853f2010f8ccc277e7b7a5ce2ae1e17614116  policy/bagel_robotics_1/VALIDATION.md
 e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855  policy/bagel_robotics_1/__init__.py
 e7667cec0939b219dc1396dcf34c8fadcd95f509e6606df9757b6c909967b5f2  policy/bagel_robotics_1/artifact-manifest.json
@@ -28,7 +28,7 @@ c71d239df91726fc519c6eb72d318ec65820627232b2f796219e87dcf35d0ab4  policy/bagel_r
 56418464b2a88e7aea31d436ee42851dfababad97db9178abdce318b5a7e621b  policy/bagel_robotics_1/model.py
 2d1698608d28de6395f8c8794460b292f0233d2a4a5ef76a5d83152631ffe389  policy/bagel_robotics_1/requirements-transport.txt
 8379dddfbbb1154ad9cc57f008f1f8d4a8d26c3de99bd6b36aa4d8067888aa92  policy/bagel_robotics_1/runtime/LICENSE
-8d576d5b6a545bfac80f84c2869c866c65b96d6f9bd3bfa0173554033c561b03  policy/bagel_robotics_1/runtime/THIRD_PARTY_NOTICES.md
+43973f510e70c55382293f19573eb81c42e965f5286f52650ca920d35ad8df77  policy/bagel_robotics_1/runtime/THIRD_PARTY_NOTICES.md
 c71d239df91726fc519c6eb72d318ec65820627232b2f796219e87dcf35d0ab4  policy/bagel_robotics_1/runtime/licenses/Apache-2.0.txt
 2258964574a8c567f3f6f871e54ef3664218bed4c40ee73eee4ed8df304fab0b  policy/bagel_robotics_1/runtime/pyproject.toml
 e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855  policy/bagel_robotics_1/runtime/src/bagel_cp1000/__init__.py

inference/policy/bagel_robotics_1/README.md

@@ -2,7 +2,7 @@
 
 **Contributor:** Bagel Labs | **Paper:** Technical report in preparation | **arXiv:** Not yet available | **Original code:** [Bagel-Robotics-1 implementation](.)
 
-Bagel-Robotics-1 combines mask-based structured pruning, recovery training and selection of important task segments to produce an inference-friendly compact WAM. VPP2 is the base model used to validate the approach; reused inference components are bundled in `runtime/` with [source notices](THIRD_PARTY_NOTICES.md). This adapter supports RoboDojo ARX X5 end-effector inference.
+Bagel-Robotics-1 combines mask-based structured pruning, recovery training and task-critical data selection to produce a compact WAM with nonuniform FFN widths and cross-attention head counts. Its inference pipeline predicts 32 actions, executes 16 before replanning, and applies 90/10 temporal ensembling. VPP2 is the validation base; reused components in `runtime/` are credited in the [source notices](THIRD_PARTY_NOTICES.md).
 
 Shared conventions — argument meanings, checkpoint naming, split-machine deployment, `EVAL_ENV_TYPE` — are documented in the [XPolicyLab README](https://github.com/XPolicyLab/XPolicyLab/blob/main/README.md). Official results: [RoboDojo LeaderBoard](https://robodojo-benchmark.com/LeaderBoard).
 

inference/policy/bagel_robotics_1/THIRD_PARTY_NOTICES.md

@@ -1,6 +1,6 @@
 # Source and artifact notices
 
-- Base-model inference components adapted from [VPP2](https://github.com/roboterax/video-prediction-policy-2)/FastWAM retain the MIT notice in
+- Reused source components from [VPP2](https://github.com/roboterax/video-prediction-policy-2)/FastWAM retain the MIT notice in
   `LICENSE`, copyright 2026 The FastWAM Authors. See `source-provenance.json`
   for the source identities and packaging changes.
 - Wan2.1/DiffSynth-derived model components retain Apache-2.0 terms; see

inference/policy/bagel_robotics_1/runtime/THIRD_PARTY_NOTICES.md

@@ -1,6 +1,6 @@
 # Runtime source notices
 
-Base-model inference components adapted from [VPP2](https://github.com/roboterax/video-prediction-policy-2)/FastWAM retain the MIT notice in LICENSE.
+Reused source components from [VPP2](https://github.com/roboterax/video-prediction-policy-2)/FastWAM retain the MIT notice in LICENSE.
 Wan2.1/DiffSynth-derived components retain Apache-2.0 terms, included in
 licenses/Apache-2.0.txt. Source attribution is retained in individual modules
 where present. Source hashes and packaging changes accompany the adapter in