cs.AI（2026-02-27）

📊 共 27 篇论文 | 🔗 2 篇有代码

🎯 兴趣领域导航

支柱九：具身大模型 (Embodied Foundation Models) (16 🔗1) 支柱二：RL算法与架构 (RL & Architecture) (11 🔗1)

🔬 支柱九：具身大模型 (Embodied Foundation Models) (16 篇)

#题目一句话要点标签🔗⭐
1 Uncertainty Quantification for Multimodal Large Language Models with Incoherence-adjusted Semantic Volume 提出UMPIRE框架，通过不一致性调整的语义体积量化多模态大语言模型的不确定性 large language model multimodal
2 Unlocking Cognitive Capabilities and Analyzing the Perception-Logic Trade-off 提出MERaLiON2-Omni以解决东南亚多模态感知与推理的挑战 large language model multimodal instruction following
3 Reasoning-Driven Multimodal LLM for Domain Generalization 提出RD-MLDG框架，利用多模态LLM的推理能力提升领域泛化性能 large language model multimodal
4 Ask don't tell: Reducing sycophancy in large language models 提出一种简单有效的输入干预方法，显著降低大语言模型中的谄媚现象 large language model
5 PseudoAct: Leveraging Pseudocode Synthesis for Flexible Planning and Action Control in Large Language Model Agents PseudoAct：利用伪代码生成提升LLM Agent的规划能力和动作控制 large language model
6 MMKG-RDS: Reasoning Data Synthesis via Deep Mining of Multimodal Knowledge Graphs 提出MMKG-RDS框架，通过深度挖掘多模态知识图谱合成推理数据，提升领域模型推理能力。 multimodal ✅
7 SleepLM: Natural-Language Intelligence for Human Sleep 提出SleepLM，通过自然语言智能实现人类睡眠的对齐、解读和交互。 foundation model multimodal
8 Artificial Agency Program: Curiosity, compression, and communication in agents 提出人工代理程序(AAP)，通过好奇心驱动的智能体学习，构建现实嵌入、资源受限的AI系统。 multimodal
9 Data Driven Optimization of GPU efficiency for Distributed LLM Adapter Serving 提出一种数据驱动的GPU优化方法，用于分布式LLM适配器服务，提升资源效率。 large language model
10 Enhancing Continual Learning for Software Vulnerability Prediction: Addressing Catastrophic Forgetting via Hybrid-Confidence-Aware Selective Replay for Temporal LLM Fine-Tuning 提出Hybrid-CASR方法，解决LLM在软件漏洞预测中灾难性遗忘问题。 large language model
11 LemmaBench: A Live, Research-Level Benchmark to Evaluate LLM Capabilities in Mathematics 提出LemmaBench，一个评估LLM数学能力的实时研究级基准 large language model
12 HotelQuEST: Balancing Quality and Efficiency in Agentic Search HotelQuEST：兼顾质量与效率的Agentic搜索评测基准 large language model
13 SLA-Aware Distributed LLM Inference Across Device-RAN-Cloud 提出面向Device-RAN-Cloud异构环境的SLA感知分布式LLM推理方案。 embodied AI
14 ODAR: Principled Adaptive Routing for LLM Reasoning via Active Inference 提出ODAR-Expert，通过主动推理进行LLM推理的自适应路由，优化计算效率。 large language model
15 AudioCapBench: Quick Evaluation on Audio Captioning across Sound, Music, and Speech AudioCapBench：一个用于快速评估跨声音、音乐和语音的音频字幕生成能力的基准。 multimodal
16 ReDON: Recurrent Diffractive Optical Neural Processor with Reconfigurable Self-Modulated Nonlinearity 提出ReDON：一种具有可重构自调制非线性的循环衍射光学神经网络处理器 large language model

🔬 支柱二：RL算法与架构 (RL & Architecture) (11 篇)

#题目一句话要点标签🔗⭐
17 EMO-R3: Reflective Reinforcement Learning for Emotional Reasoning in Multimodal Large Language Models 提出EMO-R3框架，提升多模态大语言模型在视觉情感理解中的推理能力。 reinforcement learning large language model multimodal
18 Pessimistic Auxiliary Policy for Offline Reinforcement Learning 提出悲观辅助策略，解决离线强化学习中的过估计问题 reinforcement learning offline RL offline reinforcement learning
19 DARE-bench: Evaluating Modeling and Instruction Fidelity of LLMs in Data Science DARE-bench：评估LLM在数据科学中建模和指令遵循的基准 reinforcement learning large language model instruction following
20 ProductResearch: Training E-Commerce Deep Research Agents via Multi-Agent Synthetic Trajectory Distillation 提出ProductResearch框架，通过多智能体合成轨迹蒸馏训练电商深度研究Agent distillation large language model
21 RF-Agent: Automated Reward Function Design via Language Agent Tree Search 提出RF-Agent，利用语言代理树搜索自动设计强化学习奖励函数 reward design large language model ✅
22 Construct, Merge, Solve & Adapt with Reinforcement Learning for the min-max Multiple Traveling Salesman Problem 提出基于强化学习的构造-合并-求解-适应算法RL-CMSA，解决最小-最大多旅行商问题。 reinforcement learning
23 Learning Flexible Job Shop Scheduling under Limited Buffers and Material Kitting Constraints 提出基于异构图网络的DRL方法，解决有限缓冲和物料配套约束下的柔性作业车间调度问题 reinforcement learning deep reinforcement learning DRL
24 Portfolio Reinforcement Learning with Scenario-Context Rollout 提出宏观条件情景上下文展开的强化学习方法，提升投资组合在市场剧变下的鲁棒性。 reinforcement learning
25 The Auton Agentic AI Framework Auton：用于自主Agent系统构建、执行和治理的通用AI框架 reinforcement learning large language model
26 Green or Fast? Learning to Balance Cold Starts and Idle Carbon in Serverless Computing LACE-RL：基于强化学习的Serverless冷启动与碳排放平衡管理框架 reinforcement learning deep reinforcement learning
27 RUMAD: Reinforcement-Unifying Multi-Agent Debate RUMAD：提出基于强化学习的多智能体辩论框架，提升效率与泛化性 reinforcement learning PPO

⬅️ 返回 cs.AI 首页 · 🏠 返回主页