cs.CL（2026-09-01）

📊 共 32 篇论文 | 🔗 3 篇有代码

🎯 兴趣领域导航

支柱九：具身大模型 (Embodied Foundation Models) (19 🔗3) 支柱二：RL算法与架构 (RL & Architecture) (12) 支柱三：空间感知与语义 (Perception & Semantics) (1)

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

#题目一句话要点标签🔗⭐
1 EM^2Mem: Event-Centric Multimodal Memory for Large Language Models 提出EM^2Mem框架以解决长视频问答中的多模态记忆问题 large language model multimodal ✅
2 Some Emotions Run Deeper: Layer-wise Probing and Causal Intervention in Large Language Models 通过层级探测与因果干预分析大语言模型中的情感表达 large language model
3 SDARE-Bench: Evaluating Large Language Models on Conversational Stigma Detection and Response in Dyadic and Group Dialogue 提出SDARE-Bench以解决对话中的污名检测与响应生成问题 large language model
4 Joint Training Is Not Enough: Conditioned Cross-Granularity Training for Multimodal Document Understanding 提出条件交叉粒度训练以提升多模态文档理解效果 multimodal
5 Can Large Language Models Forecast What Researchers Study Next? 提出IdeaForecastBench以评估研究思想的预测能力 large language model
6 From Confusion to Clarity: Confusion-Aware Retrieval and Knowledge Injection for Text Classification 提出混淆感知检索与知识注入以解决文本分类问题 large language model
7 CHARM: Character Hallucination for Multicultural Role Play Benchmark 提出CHARM基准以解决角色扮演中的角色幻觉问题 large language model
8 StateSwap: Probing Support-Elimination Hidden States in Multiple-Choice Questions 提出StateSwap以解决多选题中支持与消除框架的表征差异问题 large language model
9 Disclosure-Gated User Simulation for Companion-Agent Evaluation 提出披露门控用户模拟以解决伴随代理评估中的合作性问题 large language model
10 Calibration is the Bottleneck: An Action-Class Diagnostic of Multi-Turn Tool-Calling 提出行动类别诊断框架以解决多轮工具调用中的校准问题 large language model
11 From Production Traffic to Post-Training: Building a Self-Hosted LLM That Covers the Corporate Request Mix 提出自托管LLM以解决企业数据驻留问题 instruction following
12 A systematic Approach to constructing a Chance-and-Risk Matrix for Semiconductor Supply Chains 提出一种系统化方法构建半导体供应链的风险与机会矩阵 large language model
13 PCoMoE: Shifting MoE Inference from Monolithic Expert Selection to Fine-Grained Path Composition 提出PCoMoE以解决MoE推理中的专家选择问题 large language model ✅
14 TEIDAN: A Multilingual Multiparty Dialogue Corpus 提出TEIDAN多语言多方对话语料库以支持跨语言研究 multimodal
15 A Unified Mechanistic Analysis of Knowledge- and Safety-Based Refusals 提出统一机制分析以解决知识与安全拒绝问题 large language model
16 Compile, Don't Memorize: A Context Compilation Architecture (CCA) for In-Context Learning 提出上下文编译架构以解决长上下文学习的脆弱性问题 large language model ✅
17 Trust Your Guide Only When Certain: Uncertainty-Aware Sparse Alignment at Inference Time 提出TUSA以解决推理时稀疏对齐中的不确定性问题 large language model
18 Aligned but Flattened: Analyzing the Trade-off between Cultural Alignment and Diversity in LLMs 提出文化对齐与多样性权衡分析框架以解决LLMs的文化扁平化问题 large language model
19 Skill Following: Evaluating Actual Skill Use in Retrieval-Enabled LLM Agents 提出技能跟随以评估检索启用LLM代理的实际技能使用 large language model

🔬 支柱二：RL算法与架构 (RL & Architecture) (12 篇)

#题目一句话要点标签🔗⭐
20 Same Semantics, Different Outcome: On the Modality Robustness of Multimodal LLMs under Knowledge Conflict 研究多模态大语言模型在知识冲突下的鲁棒性问题 direct preference optimization large language model multimodal
21 CaRL-EM: Cost-Aware Reinforcement Learning for Entity Matching with LLMs 提出CaRL-EM以解决实体匹配中的成本意识问题 reinforcement learning large language model zero-shot transfer
22 Instella-MoE Technical Report 提出Instella-MoE以提升大规模语言模型训练效率 reinforcement learning direct preference optimization distillation
23 PersuaRL: Reinforcement Learning-Driven Multi-Expert Selection for Persuasive Dialogue Generation in Insurance 提出PersuaRL以解决保险领域对话生成的说服力不足问题 reinforcement learning large language model
24 VerTox: Verifiable Reward-Guided Corpus Poisoning Against Neural Ranking Models 提出VerTox框架以解决神经排名模型的语料中毒问题 reinforcement learning reward shaping large language model
25 SFAD: Speculative Factuality-Aware Decoding 提出SFAD框架以解决大语言模型的上下文可信性问题 reinforcement learning direct preference optimization large language model
26 Context-Grounding Gains Are Mediated by Pre-existing Machinery: Auditing GRPO, SFT, and DPO 提出审计GRPO、SFT和DPO以增强语言模型的上下文理解能力 DPO
27 The Rise of Verbal Reinforcement Learning 提出语言强化学习以提升语言代理的反馈能力 reinforcement learning
28 Knowledge Distillation During Mid-Training Favors Reasoning over Factual Recall 提出Switch Distillation以解决中期训练中的知识蒸馏问题 distillation
29 From Rollouts to Recipes: Self-Contained Post-Training for LLMs 提出自适应后训练框架以优化大语言模型的样本处理 distillation large language model
30 OUTLETS: Output-Length Prediction from Speculative Decoding Backbones 提出OUTLETS以解决大语言模型输出长度预测问题 MAE large language model
31 A Dataset for Modeling Iterative Problem-Solving 构建CodeInsight数据集以建模迭代问题解决过程 state space model TAMP

🔬 支柱三：空间感知与语义 (Perception & Semantics) (1 篇)

#题目一句话要点标签🔗⭐
32 Inspicio: Open-Vocabulary, LLM-Based Sense Retrieval for Historical Languages 提出Inspicio以解决历史语言的词义消歧问题 open-vocabulary open vocabulary

⬅️ 返回 cs.CL 首页 · 🏠 返回主页