Skip to content
Research

The Format is the Filter: Scoring Format Fixes What an LLM Judge Can Express.

arXiv:2610.17158v1 Announce Type: new Abstract: When an LLM judge filters instruction-tuning data, its scoring format (binary verdict, 1--5 rating, 0--100 score, or comparison) is rarely justified or ablated. Holding the candidate pool, judge prompt, decoding

arXiv cs.AIResearch source·