cs.AI（2026-03-02）

📊 共 40 篇论文 | 🔗 6 篇有代码

🎯 兴趣领域导航

支柱九：具身大模型 (Embodied Foundation Models) (29 🔗4) 支柱二：RL算法与架构 (RL & Architecture) (9 🔗1) 支柱一：机器人控制 (Robot Control) (2 🔗1)

🔬 支柱九：具身大模型 (Embodied Foundation Models) (29 篇)

#题目一句话要点标签🔗⭐
1 Nano-EmoX: Unifying Multimodal Emotional Intelligence from Perception to Empathy 提出Nano-EmoX，统一多模态情感智能，实现从感知到共情的建模。 multimodal chain-of-thought
2 Benchmarking LLM Summaries of Multimodal Clinical Time Series for Remote Monitoring 提出事件感知的评估框架，用于评估LLM对多模态临床时间序列的总结质量。 large language model multimodal
3 Decoding Answers Before Chain-of-Thought: Evidence from Pre-CoT Probes and Activation Steering 揭示CoT推理前决策：通过预CoT探针和激活操控发现LLM在生成CoT前已确定答案 large language model chain-of-thought
4 Cognitive Prosthetic: An AI-Enabled Multimodal System for Episodic Recall in Knowledge Work 提出CPMS认知假肢系统，利用AI增强知识工作中的情景记忆。 multimodal
5 LiveCultureBench: a Multi-Agent, Multi-Cultural Benchmark for Large Language Models in Dynamic Social Simulations LiveCultureBench：一个用于动态社会模拟中评估大语言模型的多智能体、多文化基准 large language model
6 Emerging Human-like Strategies for Semantic Memory Foraging in Large Language Models 提出人类类策略以增强大语言模型的语义记忆获取能力 large language model
7 Multimodal Mixture-of-Experts with Retrieval Augmentation for Protein Active Site Identification 提出MERA框架，利用检索增强多专家模型解决蛋白质活性位点识别问题 multimodal
8 GraphScout: Empowering Large Language Models with Intrinsic Exploration Ability for Agentic Graph Reasoning GraphScout：赋予大语言模型自主探索能力，实现Agentic图推理 large language model ✅
9 Co-Evolutionary Multi-Modal Alignment via Structured Adversarial Evolution 提出CEMMA，通过结构化对抗进化实现多模态对齐，提升安全性。 large language model multimodal
10 According to Me: Long-Term Personalized Referential Memory QA 提出ATM-Bench基准测试，用于评估多模态、多源的长期个性化指代记忆问答系统。 multimodal ✅
11 PhotoBench: Beyond Visual Matching Towards Personalized Intent-Driven Photo Retrieval 提出PhotoBench：一个面向个性化意图驱动的照片检索评测基准。 multimodal
12 GenDB: The Next Generation of Query Processing -- Synthesized, Not Engineered GenDB：利用LLM合成查询执行代码，革新传统数据库查询处理。 large language model
13 FT-Dojo: Towards Autonomous LLM Fine-Tuning with Language Agents FT-Dojo：利用语言Agent实现LLM的自主微调 large language model
14 Inference-Time Safety For Code LLMs Via Retrieval-Augmented Revision 提出检索增强修订方法，提升代码大语言模型推理时安全性 large language model
15 Words & Weights: Streamlining Multi-Turn Interactions via Co-Adaptation ROSA2：通过词与权重的协同自适应，优化多轮交互中的测试时策略调整。 large language model
16 How Small Can 6G Reason? Scaling Tiny Language Models for AI-Native Networks 针对AI原生6G网络，研究小型语言模型在网络级语义推理中的性能与效率。 large language model ✅
17 Exploring Plan Space through Conversation: An Agentic Framework for LLM-Mediated Explanations in Planning 提出基于LLM的多智能体框架，用于规划中人机交互式解释，提升用户理解与信任。 large language model
18 Real Money, Fake Models: Deceptive Model Claims in Shadow APIs 揭示影子API中LLM欺骗行为：性能差异、安全风险与身份伪造 large language model
19 GAM-RAG: Gain-Adaptive Memory for Evolving Retrieval in Retrieval-Augmented Generation 提出GAM-RAG，通过增益自适应记忆进化检索，提升RAG效率并降低推理成本。 large language model
20 GMP: A Benchmark for Content Moderation under Co-occurring Violations and Dynamic Rules 提出GMP基准，用于评估AI在多重违规和动态规则下的内容审核能力 large language model
21 CeProAgents: A Hierarchical Agents System for Automated Chemical Process Development CeProAgents：用于自动化化学过程开发的分层多智能体系统 large language model
22 Learning Structured Reasoning via Tractable Trajectory Control 提出Ctrl-R框架，通过可控轨迹学习结构化推理，提升语言和视觉语言模型在数学推理任务上的性能。 large language model
23 Assessing Crime Disclosure Patterns in a Large-Scale Cybercrime Forum 提出基于LLM的犯罪披露模式分析方法，用于大规模网络犯罪论坛内容理解。 large language model
24 DualSentinel: A Lightweight Framework for Detecting Targeted Attacks in Black-box LLM via Dual Entropy Lull Pattern DualSentinel：轻量级框架，通过双重熵平静模式检测黑盒LLM中的定向攻击 large language model
25 RubricBench: Aligning Model-Generated Rubrics with Human Standards 提出RubricBench，用于评估模型生成评分细则与人类标准的对齐程度 large language model
26 S5-HES Agent: Society 5.0-driven Agentic Framework to Democratize Smart Home Environment Simulation 提出S5-HES Agent，一个基于Society 5.0的智能家居环境模拟框架，旨在普及智能家居研究。 large language model ✅
27 Towards Privacy-Preserving LLM Inference via Collaborative Obfuscation (Technical Report) 提出AloePri，通过协同混淆实现保护隐私的大语言模型推理 large language model
28 Agentic Multi-Source Grounding for Enhanced Query Intent Understanding: A DoorDash Case Study 提出Agentic多源 grounding 系统，增强电商平台搜索中query意图理解 foundation model
29 SciDER: Scientific Data-centric End-to-end Researcher SciDER：面向科研数据的端到端自主研究Agent large language model

🔬 支柱二：RL算法与架构 (RL & Architecture) (9 篇)

#题目一句话要点标签🔗⭐
30 ProtRLSearch: A Multi-Round Multimodal Protein Search Agent with Large Language Models Trained via Reinforcement Learning ProtRLSearch：提出一种基于强化学习的多轮多模态蛋白质搜索Agent，用于解决蛋白质分析任务。 reinforcement learning large language model multimodal
31 LLM-assisted Semantic Option Discovery for Facilitating Adaptive Deep Reinforcement Learning 提出LLM辅助的语义选项发现框架，提升DRL在复杂任务中的适应性 reinforcement learning deep reinforcement learning DRL
32 MIST-RL: Mutation-based Incremental Suite Testing via Reinforcement Learning MIST-RL：基于变异和强化学习的增量式测试套件生成，提升代码验证效率。 reinforcement learning large language model
33 ToolRLA: Fine-Grained Reward Decomposition for Tool-Integrated Reinforcement Learning Alignment in Domain-Specific Agents ToolRLA：针对领域特定智能体的工具集成强化学习对齐，提出细粒度奖励分解方法 reinforcement learning direct preference optimization
34 Pencil Puzzle Bench: A Benchmark for Multi-Step Verifiable Reasoning Pencil Puzzle Bench：一个用于多步可验证推理的基准测试框架 reinforcement learning large language model
35 Tool Verification for Test-Time Reinforcement Learning T^3RL：通过工具验证稳定测试时强化学习，解决模式崩溃问题 reinforcement learning
36 CodecFlow: Efficient Bandwidth Extension via Conditional Flow Matching in Neural Codec Latent Space CodecFlow：基于神经编解码器隐空间条件流匹配的高效带宽扩展 flow matching
37 Beyond Length Scaling: Synergizing Breadth and Depth for Generative Reward Models Mix-GRM：结合广度和深度CoT的生成式奖励模型，提升评估可靠性。 reinforcement learning chain-of-thought ✅
38 Harmonizing Dense and Sparse Signals in Multi-turn RL: Dual-Horizon Credit Assignment for Industrial Sales Agents 提出DuCA框架，解决工业销售Agent中长短期目标不平衡问题 reinforcement learning large language model

🔬 支柱一：机器人控制 (Robot Control) (2 篇)

#题目一句话要点标签🔗⭐
39 Scaling Tasks, Not Samples: Mastering Humanoid Control through Multi-Task Model-Based Reinforcement Learning 提出EfficientZero-Multitask以解决人形机器人多任务控制问题 humanoid humanoid control whole-body control ✅
40 Ignore All Previous Instructions: Jailbreaking as a de-escalatory peace building practise to resist LLM social media bots 提出利用“越狱”对抗LLM社交媒体机器人，作为一种非暴力冲突降级实践 manipulation large language model

⬅️ 返回 cs.AI 首页 · 🏠 返回主页