Skip to content

Publish shoe graspability diagnostic and full native evidence - #40

Merged
Tarzelf merged 1 commit into
mainfrom
feat/shoe-graspability-v03
Oct 15, 2026
Merged

Tarzelf merged 1 commit into
mainfrom
feat/shoe-graspability-v03

Conversation

@Tarzelf

@Tarzelf Tarzelf commented Oct 15, 2026

Copy link
Copy Markdown
Contributor

The earlier shoe study could not distinguish difficult geometry from the modeled gripper and attached tool. Publish a frozen direct-grasp diagnostic with three rigid shapes, two loads, two layout/friction combinations and mounted/absent rake conditions.

All 24 attempts are valid: the 30 g controls retain 8/12 lifts; every 300 g condition fails, including the block. Removing the rake gains one narrow-shoe result and loses one block result. Both retained current-shoe cases exceed nominal jaw travel under compliant simulator limits, which is disclosed separately from the unchanged retention scores.

Includes the unchanged 107-file native replay export, every full contact trace, source-bound independent audit, recorded jaw-limit diagnostic, root acceptance and honest reproduction/licensing notes. Adds research navigation and a scoped gzip mirror for the new Hugging Face release; earlier paper bodies and evidence remain unchanged.

Validation: build and link validation passed; 63 research tests passed. Independent audit verifies all 24 models/attempts without stepping physics. Root local browser QA loaded/scrubbed all 24 replays, matched eight displayed lifts, and reported no console errors/warnings. This is authored direct pickup with synthetic objects, not a learned shoe skill or physical hardware qualification.

@Tarzelf
Tarzelf merged commit 282b324 into main Oct 15, 2026
1 check passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant