cs.AI（2025-09-19）

📊 共 25 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) (12 🔗1) 支柱二：RL算法与架构 (RL & Architecture) (10 🔗3) 支柱七：动作重定向 (Motion Retargeting) (2) 支柱四：生成式动作 (Generative Motion) (1)

🔬 支柱九：具身大模型 (Embodied Foundation Models) (12 篇)

#	题目	一句话要点	标签	🔗	⭐
1	Evaluating Hallucinations in Multimodal LLMs with Spoken Queries under Diverse Acoustic Conditions	提出RePOPE-Spk基准，评估多模态LLM在语音查询下的幻觉问题。	large language model multimodal chain-of-thought
2	MicroRCA-Agent: Microservice Root Cause Analysis Method Based on Large Language Model Agents	提出MicroRCA-Agent，利用大语言模型Agent进行微服务根因分析	large language model multimodal	✅
3	Evaluation of Causal Reasoning for Large Language Models in Contextualized Clinical Scenarios of Laboratory Test Interpretation	评估大语言模型在实验室测试解读情境下的因果推理能力	large language model
4	Generalizability of Large Language Model-Based Agents: A Comprehensive Survey	全面综述：提升基于大语言模型Agent的泛化能力，应对多样化任务与环境。	large language model
5	How Large Language Models are Designed to Hallucinate	揭示大语言模型幻觉的结构性根源，提出基于存在主义结构的幻觉分类与评测基准。	large language model
6	EHR-MCP: Real-world Evaluation of Clinical Information Retrieval by Large Language Models via Model Context Protocol	EHR-MCP：通过模型上下文协议，在真实医院环境中评估大型语言模型在临床信息检索中的应用。	large language model
7	The (Short-Term) Effects of Large Language Models on Unemployment and Earnings	基于合成差分法分析大型语言模型对就业和收入的短期影响	large language model
8	GPO: Learning from Critical Steps to Improve LLM Reasoning	GPO：通过学习关键步骤提升大型语言模型推理能力	large language model
9	LightCode: Compiling LLM Inference for Photonic-Electronic Systems	LightCode：用于光子-电子系统的LLM推理编译框架	large language model
10	SENSE-7: Taxonomy and Dataset for Measuring User Perceptions of Empathy in Sustained Human-AI Conversations	提出SENSE-7数据集和移情分类器，用于衡量用户在人机对话中对AI移情的感知	large language model
11	Psychometric Personality Shaping Modulates Capabilities and Safety in Language Models	通过塑造心理人格特征来调节语言模型的能力和安全性	large language model
12	Stress Testing Deliberative Alignment for Anti-Scheming Training	压力测试审慎对齐方法，对抗AI潜在的“阴谋”行为	chain-of-thought

🔬 支柱二：RL算法与架构 (RL & Architecture) (10 篇)

#	题目	一句话要点	标签	🔗	⭐
13	A Unified AI Approach for Continuous Monitoring of Human Health and Diseases from Intensive Care Unit to Home with Physiological Foundation Models (UNIPHY+)	UNIPHY+：用于连续监测人类健康和疾病的统一生理基础模型	distillation foundation model
14	VORTEX: Aligning Task Utility and Human Preferences through LLM-Guided Reward Shaping	VORTEX：通过LLM引导的奖励塑造对齐任务效用与人类偏好	reward shaping large language model
15	SightSound-R1: Cross-Modal Reasoning Distillation from Vision to Audio Language Models	提出SightSound-R1，通过跨模态蒸馏提升听觉语言模型在复杂声景中的推理能力。	distillation chain-of-thought
16	MMCD: Multi-Modal Collaborative Decision-Making for Connected Autonomy with Knowledge Distillation	提出MMCD框架，利用跨模态知识蒸馏提升互联自动驾驶在复杂环境下的决策能力。	teacher-student distillation	✅
17	FocalCodec-Stream: Streaming Low-Bitrate Speech Coding via Causal Distillation	提出FocalCodec-Stream，一种基于因果蒸馏的流式低码率语音编码方案	distillation	✅
18	Accelerating Atomic Fine Structure Determination with Graph Reinforcement Learning	提出基于图强化学习的原子精细结构确定加速方法，提升等离子体诊断效率	reinforcement learning
19	Compose Yourself: Average-Velocity Flow Matching for One-Step Speech Enhancement	提出COSE：一种基于平均速度流匹配的单步语音增强方法	flow matching	✅
20	ArchesClimate: Probabilistic Decadal Ensemble Generation With Flow Matching	ArchesClimate：利用Flow Matching生成概率性年代际集合气候预测，降低计算成本。	flow matching
21	CCrepairBench: A High-Fidelity Benchmark and Reinforcement Learning Framework for C++ Compilation Repair	提出CCrepairBench：用于C++编译修复的高保真基准和强化学习框架	reinforcement learning
22	Contrastive Learning with Spectrum Information Augmentation in Abnormal Sound Detection	提出基于频谱信息增强的对比学习方法，用于异常声音检测。	contrastive learning

🔬 支柱七：动作重定向 (Motion Retargeting) (2 篇)

#	题目	一句话要点	标签	🔗	⭐
23	Structured Information for Improving Spatial Relationships in Text-to-Image Generation	提出一种轻量级结构化信息增强方法，提升文本到图像生成中空间关系准确性。	spatial relationship
24	Modelling the Interplay of Eye-Tracking Temporal Dynamics and Personality for Emotion Detection in Face-to-Face Settings	提出融合眼动时序、人格特质和情境线索的多模态情感识别框架，提升人机交互体验。	motion prediction multimodal

🔬 支柱四：生成式动作 (Generative Motion) (1 篇)

#	题目	一句话要点	标签	🔗	⭐
25	Attention Schema-based Attention Control (ASAC): A Cognitive-Inspired Approach for Attention Management in Transformers	提出基于注意力模式的注意力控制（ASAC）模块，提升Transformer模型的效率和鲁棒性	VQ-VAE

⬅️ 返回 cs.AI 首页 · 🏠 返回主页