论文揭示多步LLM代理中控制框架导致的信念分歧，拷问基准测试的隐藏偏差。
arXiv:2607.04528v1 Announce Type: new Abstract: Software-agent benchmarks usually report whether an agent solves a task, but the agent reaches that ou…
新方法REVES通过修订与验证增强训练，破解LLM测试时缩放与多步推理的错位难题。
arXiv:2606.18910v1 Announce Type: new Abstract: Test-time scaling via sequential revision has emerged as a powerful paradigm for enhancing Large Langu…
用12,500条数据考验大模型在伊斯兰继承法中的复杂多步推理能力
arXiv:2603.07539v3 Announce Type: replace Abstract: Islamic inheritance law is challenging for large language models because solving inheritance cases…
多步LLM推理新突破：通过推理感知KV缓存共享与自信提前退出机制，大幅提升效率，已被ICML 2026 Workshop收录。
arXiv:2606.09937v1 Announce Type: cross Abstract: We introduce RKSC (Reasoning-Aware KV Cache Sharing), a training-free inference framework that elimi…
新型攻击揭示工具型LLM智能体因工件溯源缺口而存在的多步越狱漏洞，安全防御需突破单步文本隔离思维。
arXiv:2606.09084v1 Announce Type: cross Abstract: Tool-using LLM agents interact with the world through actions that persist state in artifacts (e.g.,…
将图结构融入LLM Agent策略优化，显著提升多步推理和任务完成能力。
arXiv:2510.26270v2 Announce Type: replace Abstract: Multi-step LLM agents in interactive environments represent a crucial step toward long-horizon dec…
通过视觉推理提升过程奖励建模精度，为复杂任务训练提供新思路。
arXiv:2508.03556v3 Announce Type: replace Abstract: Process Reward Model (PRM) is widely used in the post-training of Large Language Model (LLM) becau…
被ICML 2026收录，提出逐步置信度归因方法，精准诊断黑盒大模型的多步推理失败原因。
arXiv:2605.19228v1 Announce Type: cross Abstract: Large Language Models have achieved strong performance on reasoning tasks with objective answers by …
闭环验证推理突破复杂视觉生成，用可验证的多步推理解决规划幻觉问题，效果惊艳。
arXiv:2605.14876v1 Announce Type: cross Abstract: Despite rapid advancements, current text-to-image (T2I) models predominantly rely on a single-step g…
企业级AI智能体规模化实战：Netomi如何用GPT-4.1和GPT-5.2实现并发、治理与多步推理
How Netomi scales enterprise AI agents using GPT-4.1 and GPT-5.2—combining concurrency, governance, and multi-step reasoning for reliable production w…