用多会话强化学习重塑长上下文大模型的内存管理，ICLR 2026 Oral论文揭秘MemAgent。
arXiv:2507.02259v2 Announce Type: replace-cross Abstract: Despite improvements by length extrapolation, efficient attention and memory modules, handli…
ICLR2026最佳论文，用结构不确定性量化LLM逻辑推理一致性，为评估推理可靠性开辟新视角。
arXiv:2606.17312v1 Announce Type: new Abstract: Large language models can arrive at the same answer through reasoning paths that are unstable, contrad…
探讨仅用最后一层进行不确定性量化是否足够，挑战主流认知，对模型可靠部署有重要启发
arXiv:2605.30741v1 Announce Type: cross Abstract: Epistemic uncertainty quantification (UQ) for deep neural networks (DNNs) is a requirement for safe …
来自ICLR2026的研究：强化学习让推理更紧凑，监督微调则让推理更发散，揭示两种范式在大模型推理上的本质差异。
arXiv:2509.21128v2 Announce Type: replace Abstract: Large language models (LLMs) are typically trained by reinforcement learning (RL) with verifiable …
用反馈控制器实现精准激活引导，为提升大模型可控性提供新思路，ICLR2026论文。
arXiv:2510.04309v3 Announce Type: replace Abstract: Controlling the behaviors of large language models (LLM) is fundamental to their safety alignment …
8B参数模型专用生物实验protocol生成，步骤顺序不乱、剂量无幻觉，超越GPT-4o，已登ICLR2026。
超越GPT-4o