cs.AI（2026-05-27）

📊 共 32 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) (21 🔗4) 支柱二：RL算法与架构 (RL & Architecture) (9 🔗2) 支柱三：空间感知与语义 (Perception & Semantics) (1) 支柱八：物理动画 (Physics-based Animation) (1)

🔬 支柱九：具身大模型 (Embodied Foundation Models) (21 篇)

#题目一句话要点标签🔗⭐
1 Adaptive Multimodal Agents-Based Framework for Automatic Workflow Execution 提出自适应多模态Agent框架，用于自动工作流执行，提升非稳态场景下的可靠性。 multimodal
2 A Conflict-Aware Penalty and Statistical Loss Framework for Balancing Modalities and Enhancing Stability in Multimodal Sentiment Analysis 提出冲突感知惩罚与统计损失框架，平衡多模态信息并提升多模态情感分析稳定性。 multimodal
3 GS-FUSE: Granger-Supervised Gated Fusion and Multi-Granularity Alignment for Event-Driven Financial Forecasting GS-Fuse：基于Granger因果监督门控融合和多粒度对齐的事件驱动金融预测框架 large language model foundation model multimodal
4 Diffusion Large Language Models for Visual Speech Recognition 提出基于扩散大语言模型的视觉语音识别框架DLLM-VSR，解决传统自回归解码的局限性。 large language model
5 Explaining is Harder Than Predicting Alone: Evaluating Concept-based Explanations of MLLMs as ICL Visual Classifiers 评估MLLM作为ICL视觉分类器的概念解释能力，揭示解释比预测更难 large language model multimodal chain-of-thought
6 CyberJurors: A Multi-Agent Simulation Task for E-Commerce Disputes Verdict 提出CyberJurors多智能体框架，解决电商纠纷判决任务，模拟众裁陪审团决策。 multimodal chain-of-thought ✅
7 Revealing Algorithmic Deductive Circuits for Logical Reasoning 揭示LLM逻辑推理的算法演绎回路，定位关键注意力头 large language model chain-of-thought
8 Do Agents Need Semantic Metadata? A Comparative Study in Agentic Data Retrieval 对比实验揭示：在Agent数据检索中，语义元数据对保证数据质量至关重要 large language model
9 Multi-Adapter Representation Interventions via Energy Calibration 提出MARI：通过能量校准的多适配器表征干预，提升大语言模型对齐效果。 large language model ✅
10 The Importance of Being Statistically Earnest: A Critical Re-evaluation of GSM-Symbolic 重新评估GSM-Symbolic基准，揭示LLM推理能力评估的统计学陷阱 large language model
11 An LLM-Based Assistance System for Intuitive and Flexible Capability-Based Planning 提出基于LLM的辅助系统，提升基于能力的规划在工业自动化中的可访问性和适应性。 large language model
12 Blind PRNG Hijacking: An Undetectable Integrity-Preserving Attack Against LLM Watermarking 提出SeedHijack攻击，针对LLM水印的PRNG供应链盲篡改，实现完整性保持和正交检测。 large language model
13 Satisfiability Solving with LLMs: A Matched-Pair Evaluation of Reasoning Capability 提出基于配对公式和ADR的评估方法，更可靠地评估LLM在SAT问题上的推理能力 large language model
14 MUSE: Benchmarking Manufacturable, Functional, and Assemblable Text-to-CAD Generation MUSE：面向可制造、功能化和可组装的文本驱动CAD生成基准测试 large language model ✅
15 Tree of Thoughts as a Classical Heuristic Search Problem: Formal Foundations and Design Patterns 将思维树（ToT）框架形式化为经典启发式搜索问题，并提出设计模式 large language model
16 Let Relations Speak: An End-to-End LLM-GNN Soft Prompt Framework for Fraud Detection 提出LLM-GNN软提示框架LGSPF，用于解决欺诈检测中多关系复杂性和文本信息缺失问题。 large language model
17 Do LLMs Favor Their Providers? Measuring Vertical Integration Bias in Code Generation VIBench：评估代码生成中大型语言模型对提供商的垂直整合偏见 large language model
18 Efficient and Scalable Provenance Tracking for LLM-Generated Code Snippets 提出HYBRIDSOURCETRACKER，实现LLM生成代码片段的高效可扩展溯源跟踪 large language model
19 From Learning Resources to Competencies: LLM-Based Tagging with Evidence and Graph Constraints 提出基于LLM和图约束的标签方法，实现学习资源到能力模型的自动对齐。 large language model
20 HRBench: Benchmarking and Understanding Thinking-Mode Switch Strategies in Hybrid-Reasoning LLMs HRBench：混合推理LLM中思维模式切换策略的基准测试与理解 large language model ✅
21 A Fixed-Budget, Cluster-Aware Standard for LLM-as-a-Judge Evaluation: A Multi-Hop RAG Stress Test 提出一种固定预算、聚类感知的LLM评判标准，用于多跳RAG系统的压力测试。 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🔗⭐
22 Modeling Vehicle-Type-Specific Pedestrian Crash Avoidance Behavior in Safety-Critical Interactions Using Smooth-Mamba Deep Reinforcement Learning 提出SMamba-DDPG框架，建模自动驾驶和人类驾驶车辆下行人避撞行为差异 reinforcement learning deep reinforcement learning Mamba
23 Utility-Aware Multimodal Contrastive Learning for Product Image Generation 提出效用感知多模态对比学习框架，提升电商产品图生成质量与销量。 contrastive learning multimodal
24 DREAM-R: Multimodal Speculative Reasoning with RL-Based Refined Drafting, Precise Verification, and Fully Parallel Execution DREAM-R：通过强化学习优化草稿、精确验证和全并行执行，加速多模态推理性生成。 reinforcement learning multimodal
25 Efficient Post-training of LLMs for Code Generation With Offline Reinforcement Learning 提出离线强化学习方法，高效后训练代码生成大语言模型 reinforcement learning offline RL offline reinforcement learning
26 CaMBRAIN: Real-time, Continuous EEG Inference with Causal State Space Models CaMBRAIN：基于因果状态空间模型的实时、连续脑电图推断 Mamba SSM state space model
27 Bandwidth-Efficient and Privacy-Preserving Edge-Cloud Many-to-Many Speech Translation 提出ESRT边缘云协同框架，解决语音翻译中的隐私、带宽和语言偏见问题。 curriculum learning large language model multimodal ✅
28 TRACER: Turn-level Regret Matching with Inner Reinforcement Credit for Cooperative Multi-LLM Reasoning 提出TRACER框架，解决多智能体LLM协同推理中的奖励稀疏和策略震荡问题 reinforcement learning large language model ✅
29 DenoiseRL: Bootstrapping Reasoning Models to Recover from Noisy Prefixes DenoiseRL：通过从噪声前缀中恢复，引导推理模型提升性能 reinforcement learning large language model
30 Mechanistically Interpreting the Role of Sample Difficulty in RLVR for LLMs 通过难度分析改进RLVR，提升LLM在数学和编程推理中的性能 reinforcement learning large language model

🔬 支柱三：空间感知与语义 (Perception & Semantics) (1 篇)

#题目一句话要点标签🔗⭐
31 CubePart: An Open-Vocabulary Part-Controllable 3D Generator CubePart：提出一种可控部件的开放词汇3D生成框架，适用于游戏和模拟。 open-vocabulary open vocabulary

🔬 支柱八：物理动画 (Physics-based Animation) (1 篇)

#题目一句话要点标签🔗⭐
32 Thermodynamic properties of chemically disordered compounds via AI-driven estimation of partition function with the PULSE method 提出基于AI的PULSE方法，高效估计化学无序化合物的热力学性质 PULSE

⬅️ 返回 cs.AI 首页 · 🏠 返回主页