cs.CL（2026-01-27）

📊 共 17 篇论文 | 🔗 2 篇有代码

🎯 兴趣领域导航

支柱九：具身大模型 (Embodied Foundation Models) (14 🔗2) 支柱三：空间感知与语义 (Perception & Semantics) (2) 支柱二：RL算法与架构 (RL & Architecture) (1)

🔬 支柱九：具身大模型 (Embodied Foundation Models) (14 篇)

#	题目	一句话要点	标签	🔗	⭐
1	DiaDem: Advancing Dialogue Descriptions in Audiovisual Video Captioning for Multimodal Large Language Models	DiaDem：一种用于多模态大语言模型的音视频视频字幕生成模型，提升对话描述的准确性。	large language model multimodal
2	DREAMSTATE: Diffusing States and Parameters for Recurrent Large Language Models	DREAMSTATE：通过扩散模型编辑循环神经网络状态与参数，提升RNN的上下文感知能力。	large language model	✅
3	How Do Transformers Learn to Associate Tokens: Gradient Leading Terms Bring Mechanistic Interpretability	通过梯度主导项近似，揭示Transformer学习Token关联的机制可解释性	large language model
4	TokenSeek: Memory Efficient Fine Tuning via Instance-Aware Token Ditching	TokenSeek：通过实例感知的Token丢弃实现内存高效的微调	large language model
5	Up to 36x Speedup: Mask-based Parallel Inference Paradigm for Key Information Extraction in MLLMs	提出基于掩码的并行推理范式PIP，加速MLLM在关键信息抽取任务中的应用。	large language model
6	When Benchmarks Leak: Inference-Time Decontamination for LLMs	提出DeconIEP，通过输入嵌入扰动实现大语言模型推理时去污染，提升评测可靠性。	large language model
7	Identifying and Transferring Reasoning-Critical Neurons: Improving LLM Inference Reliability via Activation Steering	提出AdaRAS，通过激活干预提升LLM推理可靠性，无需额外训练。	large language model
8	Zero-Shot Stance Detection in the Wild: Dynamic Target Generation and Multi-Target Adaptation	提出DGTA框架，解决现实场景下零样本立场检测中动态目标生成与多目标适应问题	large language model
9	Decompose-and-Formalise: Recursively Verifiable Natural Language Inference	提出分解与形式化框架，提升自然语言推理中神经符号推理的可验证性和局部修正能力。	large language model
10	Yunque DeepResearch Technical Report	Yunque DeepResearch：提出层级模块化深度研究框架，提升自主Agent复杂任务处理能力	large language model
11	Formula-One Prompting: Adaptive Reasoning Through Equations For Applied Mathematics	提出Formula-One Prompting，通过公式推导实现应用数学问题自适应推理。	chain-of-thought
12	MetaGen: Self-Evolving Roles and Topologies for Multi-Agent LLM Reasoning	MetaGen：面向多智能体LLM推理的自进化角色与拓扑框架	large language model
13	DART: Diffusion-Inspired Speculative Decoding for Fast LLM Inference	DART：受扩散模型启发的推测解码加速LLM推理	large language model	✅
14	PsyProbe: Proactive and Interpretable Dialogue through User State Modeling for Exploratory Counseling	PsyProbe：通过用户状态建模实现探索性咨询中主动且可解释的对话	large language model

🔬 支柱三：空间感知与语义 (Perception & Semantics) (2 篇)

#	题目	一句话要点	标签	🔗	⭐
15	Dynamic Multi-Expert Projectors with Stabilized Routing for Multilingual Speech Recognition	提出SMEAR-MoE，一种用于多语种语音识别的稳定路由动态多专家投影器。	semantic mapping semantic map large language model
16	Riddle Quest : The Enigma of Words	提出基于类比的谜语生成与评估流程，用于分析语言模型的推理能力和歧义处理	semantic map large language model

🔬 支柱二：RL算法与架构 (RL & Architecture) (1 篇)

#	题目	一句话要点	标签	🔗	⭐
17	KG-CRAFT: Knowledge Graph-based Contrastive Reasoning with LLMs for Enhancing Automated Fact-checking	KG-CRAFT：利用知识图谱对比推理增强LLM的自动事实核查能力	distillation large language model

⬅️ 返回 cs.CL 首页 · 🏠 返回主页