cs.CL（2026-01-15）

📊 共 25 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) (17 🔗2) 支柱二：RL算法与架构 (RL & Architecture) (8 🔗4)

🔬 支柱九：具身大模型 (Embodied Foundation Models) (17 篇)

#题目一句话要点标签🔗⭐
1 Role-Playing Agents Driven by Large Language Models: Current Status, Challenges, and Future Trends 综述性论文：大型语言模型驱动的角色扮演Agent现状、挑战与未来趋势 large language model multimodal
2 SIN-Bench: Tracing Native Evidence Chains in Long-Context Multimodal Scientific Interleaved Literature 提出FITO范式以解决多模态科学文献理解问题 large language model multimodal
3 GeoSteer: Faithful Chain-of-Thought Steering via Latent Manifold Gradients GeoSteer：通过隐空间流形梯度提升LLM的忠实思维链推理 large language model chain-of-thought
4 Contextual StereoSet: Stress-Testing Bias Alignment Robustness in Large Language Models 提出Contextual StereoSet，用于压力测试大语言模型在不同上下文中的偏见对齐鲁棒性。 large language model
5 OctoBench: Benchmarking Scaffold-Aware Instruction Following in Repository-Grounded Agentic Coding OctoBench：评估代码仓库环境下的具身智能体对脚手架指令的遵循能力 instruction following
6 Detecting Winning Arguments with Large Language Models and Persuasion Strategies 利用大型语言模型和说服策略检测论辩文本中的胜方 large language model
7 Credit C-GPT: A Domain-Specialized Large Language Model for Conversational Understanding in Vietnamese Debt Collection 提出Credit C-GPT：一个越南语催收场景的领域专用大型语言模型 large language model
8 MoST: Mixing Speech and Text with Modality-Aware Mixture of Experts MoST：通过模态感知专家混合模型融合语音和文本 large language model multimodal ✅
9 Grounding Agent Memory in Contextual Intent 提出STITCH，通过上下文意图索引记忆，解决长时交互中记忆检索的歧义性问题。 large language model
10 Loop as a Bridge: Can Looped Transformers Truly Link Representation Space and Natural Language Outputs? 研究循环Transformer能否通过迭代提升表征空间与自然语言输出的关联性 large language model
11 AWED-FiNER: Agents, Web applications, and Expert Detectors for Fine-grained Named Entity Recognition across 36 Languages for 6.6 Billion Speakers AWED-FiNER：为66亿用户提供36种语言的细粒度命名实体识别 large language model ✅
12 DR-Arena: an Automated Evaluation Framework for Deep Research Agents DR-Arena：提出一个全自动的深度研究Agent评估框架，解决现有基准测试的局限性。 large language model
13 The Assistant Axis: Situating and Stabilizing the Default Persona of Language Models 提出助手轴概念，稳定大型语言模型默认人格并抑制有害行为。 large language model
14 Unlocking Implicit Experience: Synthesizing Tool-Use Trajectories from Text 提出GEM：一种从文本合成工具使用轨迹的方法，提升LLM多轮交互能力。 large language model
15 The Straight and Narrow: Do LLMs Possess an Internal Moral Path? 利用道德基础理论，通过干预LLM内部道德表征提升其道德对齐性 large language model
16 HUMANLLM: Benchmarking and Reinforcing LLM Anthropomorphism via Human Cognitive Patterns HUMANLLM：通过人类认知模式基准测试并强化LLM的拟人化能力 large language model
17 CALM-IT: Generating Realistic Long-Form Motivational Interviewing Dialogues with Dual-Actor Conversational Dynamics Tracking CALM-IT：通过双角色会话动态跟踪生成逼真的长程动机访谈对话 large language model

🔬 支柱二：RL算法与架构 (RL & Architecture) (8 篇)

#题目一句话要点标签🔗⭐
18 PERM: Psychology-grounded Empathetic Reward Modeling for Large Language Models 提出PERM：一种心理学驱动的共情奖励建模方法，提升大语言模型的共情能力 reinforcement learning large language model ✅
19 An Efficient Long-Context Ranking Architecture With Calibrated LLM Distillation: Application to Person-Job Fit 提出一种基于校准LLM蒸馏的高效长文本排序架构，用于人岗匹配。 distillation large language model
20 HOMURA: Taming the Sand-Glass for Time-Constrained LLM Translation via Reinforcement Learning 提出HOMURA，通过强化学习解决LLM翻译中时间约束下的跨语言冗余问题 reinforcement learning large language model
21 Long-Chain Reasoning Distillation via Adaptive Prefix Alignment 提出自适应前缀对齐蒸馏方法P-ALIGN，提升小模型长链推理能力 distillation large language model ✅
22 MatchTIR: Fine-Grained Supervision for Tool-Integrated Reasoning via Bipartite Matching MatchTIR：通过二分图匹配实现工具集成推理的细粒度监督 reinforcement learning large language model ✅
23 Be Your Own Red Teamer: Safety Alignment via Self-Play and Reflective Experience Replay 提出Safety Self-Play，通过自博弈和经验回放提升LLM安全性对齐。 reinforcement learning large language model
24 Boundary-Aware NL2SQL: Integrating Reliability through Hybrid Reward and Data Synthesis 提出BAR-SQL框架，通过混合奖励和数据合成提升NL2SQL的可靠性和边界感知能力。 reinforcement learning chain-of-thought ✅
25 Skill-Aware Data Selection and Fine-Tuning for Data-Efficient Reasoning Distillation 提出技能感知的数据选择与微调方法，实现数据高效的推理蒸馏 distillation

⬅️ 返回 cs.CL 首页 · 🏠 返回主页