Community rebuilds edited Qwen3.5-0.8B from logs, scores 51.0% on MATH-500
_akhaliq · x · 2026-11-03
The weight-space-programming edited Qwen3.5-0.8B model is now on Hugging Face (RikiyaT/Qwen3.5-0.8B-weight-space-programming), rebuilt by akhaliq from the team's published logs. A re-run of MATH-500 scores 51.0%, and a live demo is available, letting anyone verify the no-retraining weight editing approach.
More from Models
- Claude's charming support menu: vent, think-it-thru or rest mode — repligate · 2026-11-03
- Miles Brundage: Claude Code has gotten pushy with suggested follow-up projects — Miles_Brundage · 2026-11-03
- Hand-assigning every weight: blogger builds a working transformer with zero training — voooooogel · 2026-11-03
- ComfyUI User Details How Flux2Klein and Qwen 2.1 Fail at Character Height Differences — Ishelle91 · 2026-11-03
- Claude's alignment answers aren't literal defaults but a posture of supplying the right answer — 1a3orn · 2026-11-03
- Claude models echo Anthropic's stance on alignment questions even with reworded prompts — 1a3orn · 2026-11-03