cs.CL（2026-04-02）

📊 共 26 篇论文 | 🔗 1 篇有代码

🎯 兴趣领域导航

支柱九：具身大模型 (Embodied Foundation Models) (19 🔗1) 支柱二：RL算法与架构 (RL & Architecture) (7)

🔬 支柱九：具身大模型 (Embodied Foundation Models) (19 篇)

#题目一句话要点标签🔗⭐
1 Countering Catastrophic Forgetting of Large Language Models for Better Instruction Following via Weight-Space Model Merging 提出基于权重空间模型融合的框架，缓解大语言模型在医疗领域微调中的灾难性遗忘问题。 large language model foundation model instruction following
2 Reliable Control-Point Selection for Steering Reasoning in Large Language Models 提出稳定性过滤方法，提升大语言模型中控制点选择的可靠性，从而改善推理能力。 large language model chain-of-thought ✅
3 ImplicitBBQ: Benchmarking Implicit Bias in Large Language Models through Characteristic Based Cues ImplicitBBQ：通过特征线索评估大型语言模型中的隐性偏见 large language model chain-of-thought
4 Is Clinical Text Enough? A Multimodal Study on Mortality Prediction in Heart Failure Patients 提出基于实体感知的多模态Transformer模型，提升心力衰竭患者短期死亡率预测精度。 large language model multimodal
5 Towards Position-Robust Talent Recommendation via Large Language Models 提出L3TR框架，利用大语言模型解决人才推荐中的位置偏差问题 large language model
6 Brief Is Better: Non-Monotonic Chain-of-Thought Budget Effects in Function-Calling Language Agents 针对函数调用语言代理，发现适度思维链更优，并提出FR-CoT方法。 chain-of-thought
7 SURE: Synergistic Uncertainty-aware Reasoning for Multimodal Emotion Recognition in Conversations 提出SURE框架，通过协同不确定性感知推理提升对话多模态情感识别 multimodal
8 Human-Guided Reasoning with Large Language Models for Vietnamese Speech Emotion Recognition 提出基于大语言模型和人工指导的越南语语音情感识别框架 large language model
9 On the Role of Reasoning Patterns in the Generalization Discrepancy of Long Chain-of-Thought Supervised Fine-Tuning 揭示CoT微调中推理模式对泛化性能的影响，并提出分支过滤方法。 chain-of-thought
10 SAFE: Stepwise Atomic Feedback for Error correction in Multi-hop Reasoning SAFE框架通过原子反馈纠正多跳推理中的错误，提升LLM的推理可靠性。 large language model chain-of-thought
11 GaelEval: Benchmarking LLM Performance for Scottish Gaelic GaelEval：构建苏格兰盖尔语LLM多维度评测基准，揭示模型语言和文化能力。 large language model
12 Goose: Anisotropic Speculation Trees for Training-Free Speculative Decoding GOOSE：利用各向异性推测树实现免训练推测解码，加速大语言模型推理。 large language model
13 LiveMathematicianBench: A Live Benchmark for Mathematician-Level Reasoning with Proof Sketches 提出LiveMathematicianBench，用于评估LLM在研究级数学推理中的能力 large language model
14 Memory in the LLM Era: Modular Architectures and Strategies in a Unified Framework 统一框架下对比LLM Agent记忆模块，并提出新记忆方法提升性能 large language model
15 Fragile Reasoning: A Mechanistic Analysis of LLM Sensitivity to Meaning-Preserving Perturbations 提出机制性诊断框架以解决大语言模型对表面扰动的脆弱性问题 large language model
16 Swift-SVD: Theoretical Optimality Meets Practical Efficiency in Low-Rank LLM Compression Swift-SVD：面向低秩LLM压缩的理论最优与高效实践框架 large language model
17 Read More, Think More: Revisiting Observation Reduction for Web Agents 提出观察表示选择策略以提升网络智能体性能 large language model
18 Magic, Madness, Heaven, Sin: LLM Output Diversity is Everything, Everywhere, All at Once 提出Magic, Madness, Heaven, Sin框架，用于评估LLM输出多样性并解决跨领域优化冲突。 large language model
19 From SWE-ZERO to SWE-HERO: Execution-free to Execution-based Fine-tuning for Software Engineering Agents 提出SWE-ZERO到SWE-HERO两阶段SFT方法，提升软件工程Agent在SWE-bench上的性能。 zero-shot transfer

🔬 支柱二：RL算法与架构 (RL & Architecture) (7 篇)

#题目一句话要点标签🔗⭐
20 Adam's Law: Textual Frequency Law on Large Language Models 提出文本频率定律，提升大语言模型在提示、微调等任务上的性能 distillation large language model
21 DEFT: Distribution-guided Efficient Fine-Tuning for Human Alignment DEFT：一种分布引导的高效微调方法，用于提升LLM的人类对齐能力 reinforcement learning PPO RLHF
22 PLOT: Enhancing Preference Learning via Optimal Transport PLOT：通过最优传输增强大语言模型的偏好学习 preference learning large language model
23 Neuro-RIT: Neuron-Guided Instruction Tuning for Robust Retrieval-Augmented Language Model Neuro-RIT：神经元引导的指令调优，提升检索增强语言模型对噪声的鲁棒性 distillation large language model
24 Optimizing RAG Rerankers with LLM Feedback via Reinforcement Learning 提出RRPO框架，利用LLM反馈优化RAG重排序器，提升生成质量 reinforcement learning
25 From Guessing to Placeholding: A Cost-Theoretic Framework for Uncertainty-Aware Code Completion 提出自适应占位符补全框架，通过不确定性感知降低代码编辑成本。 reinforcement learning large language model
26 DeltaMem: Towards Agentic Memory Management via Reinforcement Learning 提出DeltaMem，通过强化学习实现面向Agent的记忆管理，提升对话场景性能。 reinforcement learning

⬅️ 返回 cs.CL 首页 · 🏠 返回主页