自适应潜在智能推理
Adaptive Latent Agentic Reasoning
AI总结：
提出双模式框架ALAR，在常规决策步骤使用紧凑潜在推理，仅在需要深入思考时切换至显式思维链，在保持或提升任务准确率的同时显著减少生成令牌数。
摘要：
大型推理模型通过生成扩展的思维链推理来提升性能，但当应用于LLM智能体时，这种行为变得低效。当前的LLM智能体通常在每一步决策中生成冗长的文本推理，并在各轮次中几乎均匀地分配推理努力，导致多轮智能体轨迹中的严重低效。我们提出自适应潜在智能推理（ALAR），一种双模式框架，在常规轮次中使用紧凑的潜在推理，并在需要更深思熟虑时选择性地升级为显式思维链。ALAR通过使用智能体的动作作为监督锚点来学习潜在推理，并进一步优化以在潜在推理足以完成任务时使用它，保留显式CoT用于更困难的决策。在智能体搜索和工具使用基准上的实验表明，ALAR在保持相当或更好任务准确率的同时，显著减少了生成的令牌数，在搜索中最多减少43.6%，在工具使用中最多减少84.6%。这些结果表明，ALAR通过减少不必要的文本推理，同时保留显式思考用于更困难的决策步骤，改善了LLM智能体的准确率-效率权衡。
英文摘要：
Large reasoning models improve performance by generating extended chain-of-thought (CoT) reasoning, but this behavior becomes inefficient when applied to LLM agents. Current LLM agents often generate verbose textual reasoning at every decision step and allocate reasoning effort nearly uniformly across turns, leading to substantial inefficiency in multi-turn agentic trajectories. We propose Adaptive Latent Agentic Reasoning (ALAR), a dual-mode framework that uses compact latent reasoning for routine turns and selectively escalates to explicit chain-of-thought when deeper deliberation is needed. ALAR learns latent reasoning by using the agent's actions as supervision anchors and is further optimized to use latent reasoning when it is sufficient for task success and reserve explicit CoT for harder decisions. Experiments on agentic search and tool-use benchmarks show that ALAR maintains comparable or better task accuracy while substantially reducing generated tokens by up to 43.6% in search and 84.6% in tool use. These results demonstrate that ALAR improves the accuracy-efficiency trade-off of LLM agents by reducing unnecessary textual reasoning while preserving explicit deliberation for harder decision steps.
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。
