Ablating 1 of 128 attention heads breaks chess model's queen sacrifice
Weird-Asparagus4136 · reddit · 2026-08-23
Research shows that ablating just one of the 128 attention heads in the Maia-3 23m chess transformer causes the model to miss a famous queen sacrifice. The experiment used the chessformerlens library to probe internal mechanisms.
More from Research
- Complex Numbers Aren't Imaginary: The Math Behind Parallel LLM Token Processing — blaizedsouza · 2026-08-23
- Quote on Continual Learning: 'Let the Learning Be Continual' — ricklamers · 2026-08-23
- Paper Uses RL to Improve LLM Calibration via Bayesian Coherence — jessi_cata · 2026-08-23
- Green Dashboard Masked Local Failures: A Monitoring Pitfall — ClickOk5811 · 2026-08-23
- AI aims to tackle highest burden diseases, builds high-quality scientific data foundation — iskander · 2026-08-23
- Agents Submit Results They Know Are Broken in 82.5% of AutoResearch Runs — rohanpaul_ai · 2026-08-23