You Only Need Minimal RLVR Training: Extrapolating LLMs via Rank-1 Trajectories
- Published
- Source
- arXiv
- Paper number
- 208
- Field
- Machine Learning
- arXiv ID
- 2605.21468
Key points
- This paper demonstrates that RLVR weight trajectories are extremely low-rank and highly predictable.
Paper links
External research summaries. These are not HDATF publications or measured product results.