根据问题难易动态调整大模型推理资源，简单题快算，复杂题深思。
Not All Queries Need Deep Thought: CoFiCot for Adaptive Coarse-to-fine Stateful Refinement
- 用语义熵、共识可靠性和推理深度多指标判断题目难度
- 简单题用快速聚合，复杂题进入上下文感知的纠错循环
- 纠错过程保留历史记录，避免逻辑碎片化，适合高精度推理场景
扩大测试时计算可提升大模型推理能力，但存在资源分配均一的悖论：简单任务过度修正，复杂任务修正不足。为此，我们提出 CoFiCot，一种粗粒度到细粒度自适应框架，动态匹配推理策略与问题难度。具体而言，设计多指标分类器，综合语义熵、共识可靠性与预测推理深度对查询进行分诊，实现差异化精炼阶段：简单查询采用高效聚合，复杂查询则转入上下文感知的纠错环。我们将纠错形式化为有状态的序列传播过程，每次修复严格依赖前序修正的验证历史。通过在状态依赖轨迹中集成过程奖励模型（PRMs），CoFiCot有效弥合细粒度错误定位与全局逻辑一致性之间的差距，防止无状态精炼方法常见的上下文碎片化问题。
原文摘要 · Abstract (English)
Scaling test-time computation enhances LLM reasoning ability but faces a uniform computation paradox. Allocating identical resources leads to over-correction on simple tasks and insufficient refinement on complex ones. To address this, we propose CoFiCot, a coarse-to-fine adaptive framework that dynamically tailors inference strategies to problem difficulty. Specifically, we implement a multi-metric classifier that triages queries by synthesizing semantic entropy, consensus reliability, and predicted reasoning depth . This enables a differentiated refinement stage that applies efficient aggregation for simple queries while routing complex ones to a context-aware correction loop . We formalize correction as a stateful sequential propagation process , where each repair is strictly conditioned on the verified history of prior rectifications. By integrating Process Reward Models (PRMs) within this state-dependent trajectory, CoFiCot effectively bridges the gap between granular error localization and global logical coherence, preventing the context fragmentation typical of stateless refinement methods.
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。