cs.CL（2026-03-05）

📊 共 32 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) (27 🔗4) 支柱二：RL算法与架构 (RL & Architecture) (5)

🔬 支柱九：具身大模型 (Embodied Foundation Models) (27 篇)

#题目一句话要点标签🔗⭐
1 C2-Faith: Benchmarking LLM Judges for Causal and Coverage Faithfulness in Chain-of-Thought Reasoning 提出C2-Faith基准以评估链式推理中的因果与覆盖忠实性 large language model chain-of-thought
2 TSEmbed: Unlocking Task Scaling in Universal Multimodal Embeddings TSEmbed：通过解耦任务目标实现通用多模态嵌入的任务扩展 large language model multimodal
3 IF-RewardBench: Benchmarking Judge Models for Instruction-Following Evaluation 提出IF-RewardBench，用于全面评估指令跟随评估中判别模型的性能。 large language model instruction following ✅
4 Reasoning Theater: Disentangling Model Beliefs from Chain-of-Thought 揭示推理模型中的表演性CoT：通过激活探测解耦模型信念与思维链 chain-of-thought
5 Detection of Illicit Content on Online Marketplaces using Large Language Models 利用大型语言模型检测在线市场中的非法内容 large language model
6 FireBench: Evaluating Instruction Following in Enterprise and API-Driven LLM Applications FireBench：评估企业和API驱动的LLM应用中的指令遵循能力 instruction following
7 From Unfamiliar to Familiar: Detecting Pre-training Data via Gradient Deviations in Large Language Models GDS：基于梯度偏差的大语言模型预训练数据检测方法 large language model
8 An Exploration-Analysis-Disambiguation Reasoning Framework for Word Sense Disambiguation with Low-Parameter LLMs 提出探索-分析-消歧推理框架，利用低参数LLM实现媲美GPT-4-Turbo的词义消歧。 large language model chain-of-thought
9 Balancing Coverage and Draft Latency in Vocabulary Trimming for Faster Speculative Decoding 提出词汇表裁剪方法，平衡覆盖率与延迟，加速推测解码。 large language model
10 Feature Resemblance: On the Theoretical Understanding of Analogical Reasoning in Transformers 提出特征相似性理论以理解变换器中的类比推理 large language model
11 Leveraging LLM Parametric Knowledge for Fact Checking without Retrieval 提出INTRA方法，利用LLM内部知识进行无需检索的事实核查，实现更强的泛化能力。 large language model
12 Stacked from One: Multi-Scale Self-Injection for Context Window Extension 提出SharedLLM，通过多尺度自注入扩展LLM上下文窗口至128K tokens。 large language model
13 FlashAttention-4: Algorithm and Kernel Pipelining Co-Design for Asymmetric Hardware Scaling FlashAttention-4：面向非对称硬件扩展的算法与Kernel流水线协同设计 large language model
14 Progressive Residual Warmup for Language Model Pretraining 提出渐进残差预热(ProRes)方法，加速并稳定Transformer语言模型预训练。 large language model ✅
15 Oral to Web: Digitizing 'Zero Resource'Languages of Bangladesh 构建孟加拉国少数民族语言多模态平行语料库，助力濒危语言数字化。 multimodal
16 VietJobs: A Vietnamese Job Advertisement Dataset VietJobs：首个大规模越南语招聘广告数据集，为NLP和劳动力市场分析提供基准。 large language model ✅
17 Sparse-BitNet: 1.58-bit LLMs are Naturally Friendly to Semi-Structured Sparsity 提出Sparse-BitNet以提升大语言模型的稀疏性与低比特量化效率 large language model ✅
18 NeuronMoE: Neuron-Guided Mixture-of-Experts for Efficient Multilingual LLM Extension NeuronMoE：神经元引导的MoE用于高效多语言LLM扩展 large language model
19 HiFlow: Hierarchical Feedback-Driven Optimization for Constrained Long-Form Text Generation HiFlow：用于约束长文本生成的分层反馈驱动优化框架 large language model
20 VRM: Teaching Reward Models to Understand Authentic Human Preferences 提出VRM，通过变分推理学习奖励模型以理解真实人类偏好 large language model
21 AILS-NTUA at SemEval-2026 Task 10: Agentic LLMs for Psycholinguistic Marker Extraction and Conspiracy Endorsement Detection 提出基于Agentic LLM的心理语言学标记提取与阴谋论认可检测方法 chain-of-thought
22 Beyond the Context Window: A Cost-Performance Analysis of Fact-Based Memory vs. Long-Context LLMs for Persistent Agents 对比长上下文LLM与事实型记忆系统，为持久Agent选择提供成本-性能分析。 large language model
23 The Fragility Of Moral Judgment In Large Language Models 揭示大语言模型道德判断的脆弱性：叙事形式和任务设计显著影响判断结果 large language model
24 Safer Reasoning Traces: Measuring and Mitigating Chain-of-Thought Leakage in LLMs 研究CoT推理中LLM的PII泄露问题，并提出轻量级推理时门控方法缓解泄露风险。 chain-of-thought
25 NERdME: a Named Entity Recognition Dataset for Indexing Research Artifacts in Code Repositories 提出NERdME数据集，用于识别代码仓库中研究成果的命名实体。 large language model
26 Towards Robust Retrieval-Augmented Generation Based on Knowledge Graph: A Comparative Analysis 基于知识图谱的检索增强生成：提升RAG系统在噪声环境下的鲁棒性 large language model
27 Oral to Web: Digitizing 'Zero Resource'Languages of Bangladesh 构建孟加拉国少数民族语言多模态平行语料库，助力濒危语言数字化。 multimodal

🔬 支柱二：RL算法与架构 (RL & Architecture) (5 篇)

#题目一句话要点标签🔗⭐
28 SarcasmMiner: A Dual-Track Post-Training Framework for Robust Audio-Visual Sarcasm Reasoning SarcasmMiner：提出双轨后训练框架，增强音视频反讽推理的鲁棒性 reinforcement learning distillation foundation model
29 LBM: Hierarchical Large Auto-Bidding Model via Reasoning and Acting 提出LBM：一种分层大型自动竞价模型，通过推理和行动提升广告效果 reinforcement learning offline reinforcement learning large language model
30 When Weak LLMs Speak with Confidence, Preference Alignment Gets Stronger 利用弱LLM置信度加权，显著提升偏好对齐效果 DPO large language model
31 DiSCTT: Consensus-Guided Self-Curriculum for Efficient Test-Time Adaptation in Reasoning 提出DiSCTT，通过共识引导的自课程学习实现推理中高效的测试时自适应 reinforcement learning large language model
32 Structured Multidimensional Representation Learning for Large Language Models 提出基于L-product的张量Transformer，有效降低大语言模型参数冗余并提升泛化能力。 representation learning large language model

⬅️ 返回 cs.CL 首页 · 🏠 返回主页