Repository navigation
Publish paired arm benchmark and Skillspace v0.3 release - #11
Merged
Merged
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Question
What are the present Skillspace arm capability, DOF/gripper input contract, and evidence needed for future downloadable skills? Physical grounding.
Evidence
The Skillspace arm lab now exposes object dimensions, mass, contact parameters and actuator limits, binds installation/execution to exact runtime sources, and reports joint/gripper feedback plus bounded recovery failures. Publish its v0.3 source and a frozen paired evaluation: development completion rises from 14/20 to 17/20, but held-out completion declines from 30/32 to 29/32 and observed CPU throughput declines from 12.28× to 11.39×. The report preserves this regression and the uncalibrated, simulation-only scope.
The proposed future Skillspace inspector separates media, datasets, grounded scenarios, controllers and evaluation bindings; drafts never claim simulation or hardware qualification. Existing research evidence remains unchanged, and individual contributor attribution remains empty.
Reproduction
Source archive and full receipts are included under
topics/physical-grounding/runs/2026-10-07-arm-benchmark-v1/. Python 3.12.9, MuJoCo 3.15.0, NumPy 2.5.3, Apple M5 CPU, one native world.python -m unittest discover -s tests -q;python -m simlab.studio --offline --port 8240; individual frozen DEV reproduction documented in README. No learned-from-video policy, physical qualification, GPU measurements or air-gapped dependency installation claim.Checks
100 distinct Python tests covered: 99-test full run plus a targeted 26-test subset containing one additional test.
Native network-denied execution: 3/3 known layouts; extracted final archive: three visible development cases reproduced exactly, with all 85 manifest file hashes checked.
32 held-out cases per controller, six negative controls per controller, unchanged contact/success rules, zero captured engine warnings.
Research build/link validation and all 23 publication tests pass; dedicated research Worker dry-run passes.
Primary documentation is cited; Grok opinions are separated from evidence.
Release scan and deterministic build checks exclude private session paths, credentials and personal footage.
Source LICENSE and pinned dependency list are included; no model weights or third-party dataset is redistributed.
npm run checkand all 23 publication tests pass; dedicated research Worker dry-run passes.