自适应GoGI-Skip：耦合目标梯度重要性与动态不确定性以实现高效推理
Adaptive GoGI-Skip: Coupling Goal-Gradient Importance with Dynamic Uncertainty for Efficient Reasoning
AI总结：
提出Adaptive GoGI-Skip框架，通过非线性耦合目标梯度重要性和自适应动态跳过，在保持推理准确性的同时减少45%以上令牌量，加速高达2倍。
摘要：
思维链提示以推理速度换取推理准确性。现有的压缩器由于静态梯度技术将令牌独立处理，切断序列逻辑，而基于不确定性的剪枝忽略最终答案，导致性能妥协。我们引入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）来解决这一矛盾。GoGI通过梯度敏感性量化每个令牌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对结构连贯性至关重要的低梯度令牌。在7,472个MATH轨迹上训练后，我们的策略零样本迁移到AIME、GPQA和GSM8K，令牌量减少超过45%，推理加速高达2.0倍，且无准确率损失。这些结果表明，最优压缩需要目的论目标与认知不确定性之间的协同。
英文摘要：
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.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
