LLMs can edit out flawed reasoning but don't improve it, even on tiny models
1a3orn · x · 2026-11-03
1a3orn observes that LLMs can help remove clear flaws in reasoning, yet fail to actually improve reasoning—even for very small models on a toy domain. He calls LLMs "a bit superhuman in intelligence" but "massively superhuman in patience," arguing a sufficiently patient interpretability researcher could likely reproduce this kind of work.
Related event: LLM Weight Edits Fix Flaws But Fail to Boost Reasoning, Researcher Finds(2 posts)→
More from Research
- Follow-up: Astra's weight edits only worked on Qwen's visible flaw, not on Gemma 270m — teortaxesTex · 2026-11-03
- EMBER: pretrained gene-centric foundation model predicts multigene perturbation effects — jmuiuc · 2026-11-03
- LLM edits another LLM: tweaking 0.016% of weights lifts Qwen3.5-0.8B math from 7.9% to 48.1% — teortaxesTex · 2026-11-03
- New preprint RLCE: reinforcement learning with comparative evidence for social intelligence — pliang279 · 2026-11-03
- Tau Ceti uses AI plus Lean kernel to formalize 'all of basic mathematics', open to contributors — AlexKontorovich · 2026-11-03
- New research links AI companions to less loneliness and higher self-esteem — serialchilla91 · 2026-11-03