Learning from Emptiness: De-biasing Listwise Rerankers with Content-Agnostic Probability Calibration

TL;DR

CapCal is a training-free method that reduces position bias in listwise rerankers by decoupling bias from relevance, significantly improving lightweight model performance across multiple benchmarks.

Contribution

It introduces a content-agnostic, training-free bias calibration framework that enhances reranking accuracy without additional inference latency.

Findings

CapCal outperforms existing training-free bias mitigation methods.

It achieves over 10 points NDCG gain on lightweight models.

CapCal surpasses permutation-based and data-augmentation baselines.

Abstract

Generative listwise reranking leverages global context for superior retrieval but is plagued by intrinsic position bias, where models exhibit structural sensitivity to input order independent of relevance. Existing mitigations present a dilemma: inference-time aggregation incurs prohibitive latency, while training-based methods often fail to eradicate ingrained priors, particularly in compact models. To resolve this dilemma, we propose CapCal (Content-Agnostic Probability Calibration), a training-free framework that mechanically decouples positional bias from ranking decisions. By estimating the bias distribution via content-free placeholders, CapCal rectifies output logits through an entropy-adaptive contrastive mechanism. Evaluations across 10 benchmarks confirm that CapCal achieves superior performance among training-free methods while preserving single-pass efficiency. Notably, it unlocks the latent potential of lightweight models (e.g., 0.6B), delivering absolute NDCG gains exceeding 10 points and outperforming both permutation-based aggregation and data-augmentation baselines.