Count Evidence, Not Sentences: Tempered Evidence Fusion of LLM Judgments for Long-Text Value Measurement
Tempered Evidence Fusion (TEF) is proposed, a training-free rule that weights each sentence's log-odds by its normalized information gain, as derived from a generalized Bayesian posterior, making the fused score nearly vanish for uncertain sentences while preserving the Bayes-optimal weight of decisive evidence.
Yu-He Wu, Rui Qian, Guang-Yu Wang et al.
· 0 citations