CapCal de-biases generative listwise rerankers via content-agnostic placeholder-based bias estimation and entropy-adaptive logit rectification, yielding over 10-point NDCG gains on lightweight models across 10 benchmarks while retaining single-pass speed.
[{num}] {passage {num}} Search Query: {query}
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.AI 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Learning from Emptiness: De-biasing Listwise Rerankers with Content-Agnostic Probability Calibration
CapCal de-biases generative listwise rerankers via content-agnostic placeholder-based bias estimation and entropy-adaptive logit rectification, yielding over 10-point NDCG gains on lightweight models across 10 benchmarks while retaining single-pass speed.