一种考虑成本校准前沿效用的预算感知LLM发现方法，平衡搜索效率与资源消耗，提升推理时发现性价比
arXiv:2607.26828v1 Announce Type: cross Abstract: Large language models increasingly support scientific and algorithmic discovery through inference-ti…
大语言模型秘密编码剩余输出长度，揭示其生成结构的可预测线性规律。
arXiv:2607.05316v1 Announce Type: cross Abstract: Large language models generate one token at a time, yet their responses show remarkably consistent l…
探索如何通过微调大语言模型提升量子推理能力，来自arXiv的前沿研究。
arXiv:2606.21974v1 Announce Type: cross Abstract: Large language models (LLMs) exhibit abilities beyond natural language modelling and text generation…
用LLM自动生成OpenSIL固件单元测试，结合库感知双打与迭代修复显著提升测试可靠性。
arXiv:2606.19725v1 Announce Type: cross Abstract: Validating changes in low-level C firmware is expensive because unit tests (UTs) are fragile under s…
被ICML Workshop收录，提出一种用于大语言模型的线性路由器，有望简化模型推理中的资源分配问题。
arXiv:2606.06098v1 Announce Type: new Abstract: Foundational Large Language Models (LLMs) demonstrate proficiency on a wide range of general tasks, an…
大模型指导多智能体合作强化学习，激励感知奖励设计新突破，提升协同效率与泛化能力。
arXiv:2603.24324v4 Announce Type: replace Abstract: Designing effective auxiliary rewards for cooperative multi-agent systems remains challenging, as …
评估大语言模型语义泛化能力的新方法，利用短语结构区分记忆与真正理解，直击预训练数据偏差挑战。
arXiv:2501.04661v3 Announce Type: replace-cross Abstract: The web-scale of pretraining data has created an important evaluation challenge: to disentan…
扩散模型与大语言模型联手攻克视觉语音识别，开辟唇读与无声交互新路径
arXiv:2605.28456v1 Announce Type: new Abstract: Existing Visual Speech Recognition (VSR) systems commonly rely on left-to-right autoregressive decodin…
大语言模型能否破解自我毁灭亚文化（如Jirai Kei）中的语义歧义？研究给出实证答案。
arXiv:2601.05004v2 Announce Type: replace Abstract: Self-destructive behaviors are linked to complex psychological states and can be challenging to di…
综述探讨基础模型在个性化联邦智能中的应用，汇总当前方法、挑战与未来方向。
arXiv:2505.06907v2 Announce Type: replace-cross Abstract: The rise of large language models (LLMs), such as ChatGPT, Gemini, and Grok, has reshaped th…
提出Ghosted Layers，无需训练即可恢复层剪枝后LLM的性能，通过激活对齐解决隐藏状态不匹配问题。
arXiv:2605.15491v1 Announce Type: cross Abstract: Layer pruning removes entire Transformer decoder blocks from large language models, but introduces a…
从被动回应到主动参与，最新研究教大模型何时该主动发言，突破传统对话边界。
arXiv:2508.18167v2 Announce Type: replace Abstract: Large Language Models (LLMs) have demonstrated remarkable capabilities in understanding and genera…
机器遗忘不仅能删除记忆，还能诱导模型产生可控的副作用行为与能力，颠覆了传统遗忘认知。
arXiv:2601.21702v3 Announce Type: replace Abstract: We consider Representation Misdirection (RM), a class of large language model (LLM) unlearning met…
多样信号集成监控比增加算力更能保障AI代理安全，这篇论文提出新思路。
arXiv:2605.15377v1 Announce Type: new Abstract: As AI systems are increasingly deployed in autonomous agentic settings at scale, it is important to en…
多轮对话中，长距离依赖与信息稀疏是老大难，这篇论文提出的Self-Recall Thinking框架，让模型在推理时主动召回历史关键轮次，无需外部模块，4.7% F1提升+14.7%延迟缩减，平衡了准确与效率，思路清奇。
arXiv:2605.15102v1 Announce Type: cross Abstract: Large language model (LLM) based multi-turn dialogue systems often struggle to track dependencies ac…
大模型在纵向电子病历推理中常因缺乏精细化统计导致幻觉，COTCAgent通过概率化思维链补全实现预防性咨询，为临床决策支持提供新范式。
arXiv:2605.15016v1 Announce Type: cross Abstract: As large language models empower healthcare, intelligent clinical decision support has developed rap…