RESEARCH · RESEARCH · #1403
arXiv:2610.00331v1 — Mathematical transfer in LLMs favors reasoning approach over topic
A new arXiv preprint reports fine-tuning experiments that compare transfer from sources sharing a solution method (same-approach, SA) versus sources sharing the topic (same-topic, ST). Using two counterbalanced 2×2 designs (2,000 and 800 problems), five base models, and three training seeds, SA outperformed ST in all 40 pooled model–target comparisons, with model-level advantages roughly 8–16 percentage points (means 10.8 and 14.3); ST sources were more similar by embedding and lexical measures, so the advantage of SA runs opposite to surface statement similarity.
KEY POINTS
- A new arXiv preprint reports fine-tuning experiments that compare transfer from sources sharing a solution method (same-approach, SA) versus sources sharing the topic (same-topic, ST).
- Using two counterbalanced 2×2 designs (2,000 and 800 problems), five base models, and three training seeds, SA outperformed ST in all 40 pooled model–target comparisons, with model-level advantages roughly 8–16 percentage points (means 10.8 and 14.3); ST sources were more similar by embedding and lexical measures, so the advantage of SA runs opposite to surface statement similarity.
- Indicates that matching fine-tuning data by shared solution method produces stronger cross-topic mathematical transfer than matching by topic, guiding data selection for LLM math fine-tuning.
WHY IT MATTERS
Indicates that matching fine-tuning data by shared solution method produces stronger cross-topic mathematical transfer than matching by topic, guiding data selection for LLM math fine-tuning.