一篇新研究论文探讨了语言模型和推理系统中的测试时缩放概念，认为过度采样会导致性能下降。该论文引入了“模态上限”和“相关性上限”来描述额外采样收益递减或产生负面结果的点。它表明，这些系统的瓶颈在于识别正确答案，而不是生成答案，并且对于大多数任务来说，几十次采样就足够了。 AI
影响 表明当前通过广泛采样提高AI性能的方法可能效率低下且适得其反。
排序理由 该集群包含一篇在arXiv上发表的学术论文，详细介绍了新的研究结果。
在 Hugging Face Daily Papers 阅读 →
AI 生成摘要 · Google Gemini · 来自 6 个来源。 我们如何撰写摘要 →
完整方法见我们的编辑标准。
arXiv:2607.01612v1 Announce Type: new Abstract: Training large language models (LLMs) with reinforcement learning (RL) has significantly advanced their performance on reasoning and question-answering tasks. However, prevailing RL reward designs typically prioritize response corre…
arXiv:2607.01179v1 Announce Type: cross Abstract: Scaling inference compute, by generating many parallel attempts per problem, is a costly but reliable lever for improving language model capabilities. By default these attempts are generated independently, wasting inference comput…
Scaling inference compute, by generating many parallel attempts per problem, is a costly but reliable lever for improving language model capabilities. By default these attempts are generated independently, wasting inference compute on redundant solutions. This waste seems unavoid…
Sampling-based reasoning systems face a trade-off between coverage and selection, where additional samples beyond a few dozen provide diminishing returns and can degrade performance.
arXiv:2606.28661v1 Announce Type: cross Abstract: People overthink; language models over-sample, and the extra effort can talk both into a worse answer. Reasoning systems answer a hard question by sampling it many times (test-time scaling), and the more they draw, the more often …
People overthink; language models over-sample, and the extra effort can talk both into a worse answer. Reasoning systems answer a hard question by sampling it many times (test-time scaling), and the more they draw, the more often a correct answer turns up somewhere, so coverage, …