cs.AI（2026-02-14）

📊 共 24 篇论文 | 🔗 4 篇有代码

🎯 兴趣领域导航

支柱九：具身大模型 (Embodied Foundation Models) (16 🔗3) 支柱二：RL算法与架构 (RL & Architecture) (6 🔗1) 支柱八：物理动画 (Physics-based Animation) (1) 支柱一：机器人控制 (Robot Control) (1)

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

#	题目	一句话要点	标签	🔗	⭐
1	Diagnosing Pathological Chain-of-Thought in Reasoning Models	提出一套评估指标，用于诊断推理模型中思维链（CoT）的病态现象	chain-of-thought
2	StackingNet: Collective Inference Across Independent AI Foundation Models	StackingNet：通过跨独立AI基础模型的集体推理实现性能提升	foundation model
3	From What to How: Bridging User Requirements with Software Development Using Large Language Models	提出DesBench基准，评估大语言模型在软件设计任务中的能力	large language model
4	Attention in Constant Time: Vashista Sparse Attention for Long-Context Decoding with Exponential Guarantees	提出Vashista稀疏注意力以解决长上下文解码效率问题	large language model
5	RDBLearn: Simple In-Context Prediction Over Relational Databases	RDBLearn：关系数据库上的简单上下文学习预测	foundation model	✅
6	Multi-Modal Sensing and Fusion in mmWave Beamforming for Connected Vehicles: A Transformer Based Framework	提出基于Transformer的多模态融合毫米波波束赋形框架，降低车联网环境下的波束训练开销。	multimodal
7	Evaluating LLM-Generated ACSL Annotations for Formal Verification	评估LLM生成的ACSL注解在形式化验证中的有效性	large language model
8	DTBench: A Synthetic Benchmark for Document-to-Table Extraction	DTBench：一个用于文档到表格抽取任务的合成基准测试，着重评估LLM的结构化数据生成能力。	large language model	✅
9	OneLatent: Single-Token Compression for Visual Latent Reasoning	OneLatent：通过单token压缩视觉潜在推理，降低CoT推理成本。	chain-of-thought
10	Can a Lightweight Automated AI Pipeline Solve Research-Level Mathematical Problems?	轻量级AI自动流程解决研究级数学难题：基于引用验证优化	large language model
11	PhGPO: Pheromone-Guided Policy Optimization for Long-Horizon Tool Planning	提出PhGPO，利用信息素引导策略优化，解决长时程工具规划问题	large language model
12	AllMem: A Memory-centric Recipe for Efficient Long-context Modeling	AllMem：一种以内存为中心的方案，用于高效的长文本建模。	large language model
13	MAS-on-the-Fly: Dynamic Adaptation of LLM-based Multi-Agent Systems at Test Time	MASFly：测试时动态自适应LLM多智能体系统框架	large language model
14	Guided Collaboration in Heterogeneous LLM-Based Multi-Agent Systems via Entropy-Based Understanding Assessment and Experience Retrieval	提出基于熵的多智能体协作框架，解决异构LLM系统中认知失配问题。	large language model
15	Rubrics as an Attack Surface: Stealthy Preference Drift in LLM Judges	揭示LLM评判器中基于规则的隐蔽偏好漂移攻击，并提出RIPD风险	large language model	✅
16	Who Do LLMs Trust? Human Experts Matter More Than Other LLMs	LLM更信任谁？人类专家比其他LLM更重要	large language model

🔬 支柱二：RL算法与架构 (RL & Architecture) (6 篇)

#	题目	一句话要点	标签	🔗	⭐
17	No Need to Train Your RDB Foundation Model	提出一种无需训练的关系数据库(RDB)基础模型，实现跨表预测任务的零样本迁移。	predictive model foundation model	✅
18	Building Autonomous GUI Navigation via Agentic-Q Estimation and Step-Wise Policy Optimization	提出Agentic-Q估计和步进式策略优化，提升GUI自主导航能力	reinforcement learning large language model multimodal
19	GSRM: Generative Speech Reward Model for Speech RLHF	提出GSRM：一种用于语音RLHF的生成式语音奖励模型，提升语音自然度评估与生成。	RLHF chain-of-thought
20	From Pixels to Policies: Reinforcing Spatial Reasoning in Language Models for Content-Aware Layout Design	LaySPA：强化语言模型空间推理，实现内容感知布局设计	reinforcement learning policy learning large language model
21	OpAgent: Operator Agent for Web Navigation	OpAgent：用于Web导航的在线增强学习操作代理，实现71.6%的SOTA成功率	reinforcement learning offline reinforcement learning instruction following
22	AuTAgent: A Reinforcement Learning Framework for Tool-Augmented Audio Reasoning	AuTAgent：强化学习驱动的工具增强音频推理框架	reinforcement learning

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

#	题目	一句话要点	标签	🔗	⭐
23	AISA: Awakening Intrinsic Safety Awareness in Large Language Models against Jailbreak Attacks	AISA：通过唤醒大语言模型内在安全意识防御越狱攻击	spatiotemporal large language model

🔬 支柱一：机器人控制 (Robot Control) (1 篇)

#	题目	一句话要点	标签	🔗	⭐
24	Enabling Option Learning in Sparse Rewards with Hindsight Experience Replay	提出MOC-2HER，通过双目标逆向经验回放解决稀疏奖励下的机械臂操作学习问题	manipulation reinforcement learning

⬅️ 返回 cs.AI 首页 · 🏠 返回主页