1 paper · 2 filters
Minghan Li, Eric Gaussier, Juntao Li +1
Decoder-only LLM rerankers struggle with long documents: inference is costly and relevance signals can be diluted by irrelevant context. Motivated by an attention analysis indicati…