将思维链用于文档重排序时，连续相关性信号被离散文本限制，导致排名精度难以提升。
Beyond Polarization: The Generative Constraint of Chain-of-Thought in Pointwise Reranking
- 通过强化学习、细粒度监督等方法修复模型偏差，但排名差距仍存
- 在320亿参数规模下，思维链模型始终落后于直接评分模型
- 揭示了点对点重排序中思维链的内在瓶颈，非简单训练问题
在点对点文档重排序任务中，思维链（Chain-of-Thought）模型通常表现不如直接打分模型。尽管现有诊断认为原因在于分类能力差、分数极化或校准失效，但针对性训练能否缩小这一差距尚不明确。我们的实证研究首先证实，该性能差距在高达320亿参数规模下依然稳定存在，排除了模型容量和数据量的干扰。随后，我们采用强化学习、细粒度监督与架构解耦等压力测试手段，主动修复各类偏差。尽管这些干预提升了分类准确率和绝对得分，相对排名差距依然持续存在。这表明，在点对点评分范式下，将连续相关性语义经由离散文本传递，会限制排名信号的分辨率，形成一个稳定且难以克服的瓶颈，而非可轻易修正的训练偏误。
原文摘要 · Abstract (English)
In pointwise document reranking, Chain-of-Thought models typically underperform direct scoring models. While existing diagnostics attribute this to inferior classification, score polarization, or calibration breakdown, whether targeted training can bridge this gap remains unclear. Our empirical study first confirms that this gap is stable across scales up to 32B parameters, ruling out model and data capacity confounders. We then apply stress tests utilizing reinforcement learning, fine-grained supervision, and architectural decoupling to explicitly repair these deviations. Although these interventions improve classification accuracy and absolute scores, the relative ranking gap persists. These findings suggest that, within the pointwise scoring paradigm, routing continuous relevance semantics through discrete text constrains ranking signal resolution, revealing a bottleneck that is stable and difficult to overcome under current standard methods, rather than an easily resolvable training bias.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。