Pre-training sets low-rank geometry; fine-tuning cannot fix fundamental mistakes
yunta_tsai · x · 2026-08-21
The core insight is that pre-training establishes the correct low-rank geometry, while RL fine-tunes the shape. A better low-rank geometry requires less fine-tuning. Conversely, fine-tuning cannot fix low-rank mistakes if the observations have gaping holes, as the optimizer cannot deduce the correct shape.
Related event: Pretraining sets the low-rank geometry; RL only reshapes it(2 posts)→
More from Research
- τ0-VLA: Hierarchical Robot Foundation Model with World-Model-Guided Test-Time Search — sii-research · 2026-08-22
- ITSMBench Open Sourced: 50 Tasks for Multi-turn Agent Evaluation — Shahules786 · 2026-08-22
- Recursion releases Nesso: 10x faster binding affinity model for drug discovery — anshulkundaje · 2026-08-22
- Interview with Owain Evans explores emergent misalignment and LLM persona safety — sjgadler · 2026-08-22
- NVIDIA & Berkeley open source tactile robotics method T-Rex — DrJimFan · 2026-08-22
- Lightwheel and Hugging Face release 100k-hour human manipulation dataset — vanstriendaniel · 2026-08-22