cs.CL（2025-05-10）

📊 共 12 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) (8 🔗3) 支柱二：RL算法与架构 (RL & Architecture) (3 🔗1) 支柱六：视频提取与匹配 (Video Extraction) (1)

🔬 支柱九：具身大模型 (Embodied Foundation Models) (8 篇)

#题目一句话要点标签🔗⭐
1 Integrating Video and Text: A Balanced Approach to Multimodal Summary Generation and Evaluation 提出一种平衡视频与文本的多模态摘要生成与评估方法，提升视觉信息相关性。 multimodal
2 People Are Highly Cooperative with Large Language Models, Especially When Communication Is Possible or Following Human Interaction 研究表明，人们与大型语言模型（LLM）的合作意愿高，尤其在可沟通或人机交互后。 large language model
3 Recovering Event Probabilities from Large Language Model Embeddings via Axiomatic Constraints 提出基于公理约束的VAE方法，从LLM嵌入中恢复更符合概率论的事件概率 large language model
4 Think in Safety: Unveiling and Mitigating Safety Alignment Collapse in Multimodal Large Reasoning Model 提出安全导向思维调优方法，提升多模态大模型推理的安全性与鲁棒性 multimodal ✅
5 Utilizing LLMs to Investigate the Disputed Role of Evidence in Electronic Cigarette Health Policy Formation in Australia and the UK 利用大型语言模型分析电子烟政策文件中证据的角色差异 large language model
6 SCAN: Structured Capability Assessment and Navigation for LLMs 提出SCAN框架，用于对LLM能力进行结构化评估与导航，实现细粒度能力分析。 large language model ✅
7 Signals from the Floods: AI-Driven Disaster Analysis through Multi-Source Data Fusion 融合社交媒体与公共调查数据，提出AI驱动的灾害分析方法。 large language model
8 MacRAG: Compress, Slice, and Scale-up for Multi-Scale Adaptive Context RAG MacRAG：一种多尺度自适应上下文RAG框架，用于压缩、切分和扩展长文档多跳问答。 large language model ✅

🔬 支柱二：RL算法与架构 (RL & Architecture) (3 篇)

#题目一句话要点标签🔗⭐
9 Gated Attention for Large Language Models: Non-linearity, Sparsity, and Attention-Sink-Free 提出门控注意力机制，提升大语言模型非线性、稀疏性和长文本外推能力。 state space model linear attention large language model ✅
10 REFINE-AF: A Task-Agnostic Framework to Align Language Models via Self-Generated Instructions using Reinforcement Learning from Automated Feedback 提出REFINE-AF框架，通过自生成指令和强化学习对小型语言模型进行任务无关对齐。 reinforcement learning large language model
11 xGen-small Technical Report xGen-small：面向长文本应用的4B/9B Transformer解码器模型 reinforcement learning preference learning

🔬 支柱六：视频提取与匹配 (Video Extraction) (1 篇)

#题目一句话要点标签🔗⭐
12 Boosting Neural Language Inference via Cascaded Interactive Reasoning 提出级联交互推理网络CIRN，通过多层级交互提升自然语言推理性能。 feature matching

⬅️ 返回 cs.CL 首页 · 🏠 返回主页