基于置信度的分叉思考
Fork-Think with Confidence
AI总结：
提出一种先决策后思考的并行推理方法，通过模型置信度识别分叉点并采样多条路径聚合，在保持或提升推理性能的同时，将token消耗降低30%，运行时间减少57%。
摘要：
并行思考在无需重新训练的情况下，显著提升了LLM在推理任务上的性能。然而，现有方法遵循先思考后决策的范式，即先采样多条推理路径，这不可避免地导致过度生成，然后通过修剪或停止不必要的路径来补偿。相比之下，先决策后思考，即先识别可能产生理想生成的点，至今尚未得到充分探索。遵循这一范式，我们提出了Fork-think with confidence，该方法首先利用模型置信度在单一种子路径中识别分叉点，然后触发思考，采样多个延续并聚合它们以生成最终响应。我们在三个模型和三个推理基准上的实验表明，Fork-think将token消耗降低了高达30%，运行时间减少了高达57%，同时性能与并行思考相当或更优。我们的分析揭示，Fork-think能够识别对下游任务有意义的分叉点，并且在较晚位置采样可以产生显著更好的生成。最后，我们展示了将Fork-think与现有机制（如早停和加权投票）结合，可以进一步提升性能，并与现有最先进方法性能相当，且无需任何预热或离线训练。我们的结果确立了预确定分叉作为高效LLM推理的一个有前景的研究方向。
英文摘要：
Parallel thinking has enjoyed great success for boosting LLM performance on reasoning tasks without the need for any re-training. However, existing methods follow a think-first-then-decide paradigm, i.e., they first sample multiple reasoning paths, which inevitably leads to overgeneration, then prune or stop unnecessary paths to compensate. In contrast, decide-first-then-think, i.e., first identifying points that are likely to lead to desirable generations, has been underexplored so far. Following this paradigm, we propose Fork-think with confidence, that first identifies forking points using model confidence in a single seeding path, then triggers thinking, sampling multiple continuations and aggregating them for the final response. Our experiments across three models and three reasoning benchmarks show that Fork-think reduces the token consumption by up to 30% and run-time by up to 57%, while performing comparable to or better than parallel thinking. Our analysis reveals that Fork-think is able to identify forking points that are meaningful with respect to the downstream task and that sampling at later positions can lead to substantially better generations. Finally, we demonstrate how combining Fork-think with existing mechanisms such as early stopping and weighted voting can further boost the performance and perform comparably to existing state-of-the-art methods, without requiring any warm-up or offline training. Our results establish pre-determined forking as a promising research direction for efficient LLM reasoning.
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。
