让大模型推理更快更准，通过动态调整思考过程的关键步骤。
Adaptive GoGI-Skip: Coupling Goal-Gradient Importance with Dynamic Uncertainty for Efficient Reasoning
- 结合目标梯度重要性和实时不确定性，动态跳过无关思考步骤。
- 在多个数学数据集上减少45%以上计算量，速度提升2倍且准确率不变。
- 适合追求高效推理的AI应用，尤其在资源受限场景下表现突出。
思维链（CoT）提示以牺牲推理速度为代价换取准确性。现有压缩方法存在权衡：静态梯度方法独立处理词元，破坏逻辑连贯性；基于不确定性的剪枝则忽略最终答案。本文提出Adaptive GoGI-Skip框架，通过非线性耦合目标梯度重要性（GoGI）与自适应动态跳过（ADS），解决这一矛盾。GoGI通过梯度敏感性量化每个词元对答案正确性的功能贡献；ADS利用运行时熵动态调节GoGI阈值，在高不确定性节点保留低梯度词元以维持结构连贯性。模型在7,472条MATH推理轨迹上训练，零样本迁移至AIME、GPQA和GSM8K，token消耗减少超过45%，推理速度最高提升2.0倍，且无准确率损失。结果表明，最优思考压缩需兼顾目的性目标与认知不确定性。
原文摘要 · Abstract (English)
Chain-of-Thought (CoT) prompting trades inference speed for reasoning accuracy. Existing compressors force a compromise as static gradient techniques treat tokens independently, severing sequential logic, while uncertainty-based pruning ignores the final answer. We introduce Adaptive GoGI-Skip, a framework that resolves this tension by non-linearly coupling Goal-Gradient Importance (GoGI) with Adaptive Dynamic Skipping (ADS). GoGI quantifies each token's functional contribution to answer correctness via gradient sensitivity. ADS leverages runtime entropy to dynamically modulate the GoGI threshold, preserving low-gradient tokens essential for structural coherence at high-uncertainty junctions. Trained on 7,472 MATH traces, our policy transfers zero-shot to AIME, GPQA, and GSM8K, reducing token volume by $>$45% and accelerating inference up to 2.0$\times$ without accuracy loss. These results suggest that thinking-optimal compression demands synergy between teleological goals and epistemic uncertainty.
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。