cs.CL（2026-03-02）

📊 共 24 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) (19 🔗1) 支柱二：RL算法与架构 (RL & Architecture) (5 🔗3)

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

#题目一句话要点标签🔗⭐
1 MMR-Life: Piecing Together Real-life Scenes for Multimodal Multi-image Reasoning 提出MMR-Life基准，评估多模态大语言模型在真实场景下的多图推理能力。 large language model multimodal
2 Toward Graph-Tokenizing Large Language Models with Reconstructive Graph Instruction Tuning 提出RGLM，通过重构图指令微调增强图Token化大语言模型中的图文对齐。 large language model foundation model
3 Extracting Training Dialogue Data from Large Language Model based Task Bots 针对LLM驱动的任务型对话系统，提出新型数据提取攻击方法以评估和缓解隐私风险。 large language model
4 nchellwig at SemEval-2026 Task 3: Self-Consistent Structured Generation (SCSG) for Dimensional Aspect-Based Sentiment Analysis using Large Language Models 提出自洽结构化生成(SCSG)方法，利用大语言模型解决维度情感分析问题。 large language model
5 Quantifying Conversational Reliability of Large Language Models under Multi-Turn Interaction 量化大型语言模型在多轮交互中的对话可靠性 large language model
6 OmniRet: Efficient and High-Fidelity Omni Modality Retrieval OmniRet：提出一种高效且高保真的全模态检索模型，支持文本、视觉和音频的复杂组合查询。 large language model multimodal
7 FreeAct: Freeing Activations for LLM Quantization FreeAct：释放激活的LLM量化方法，提升扩散和多模态LLM性能。 large language model multimodal
8 EstLLM: Enhancing Estonian Capabilities in Multilingual LLMs via Continued Pretraining and Post-Training EstLLM：通过持续预训练和后训练增强多语言LLM中的爱沙尼亚语能力 large language model instruction following
9 OpenAutoNLU: Open Source AutoML Library for NLU OpenAutoNLU：开源AutoML库，面向自然语言理解任务，实现数据感知训练。 large language model
10 Building a Strong Instruction Language Model for a Less-Resourced Language 针对低资源斯洛文尼亚语，提出GaMS3-12B指令语言模型，性能媲美GPT-4o large language model
11 Zero- and Few-Shot Named-Entity Recognition: Case Study and Dataset in the Crime Domain (CrimeNER) 提出CrimeNERdb数据集，研究犯罪领域零样本和少样本命名实体识别问题 large language model
12 LLMs as Strategic Actors: Behavioral Alignment, Risk Calibration, and Argumentation Framing in Geopolitical Simulations 评估LLM在复杂地缘政治模拟中的战略行为与决策模式 large language model
13 ALTER: Asymmetric LoRA for Token-Entropy-Guided Unlearning of LLMs 提出ALTER框架，通过非对称LoRA和token熵引导实现LLM高效可控的知识遗忘 large language model
14 Enhancing Persona Following at Decoding Time via Dynamic Importance Estimation for Role-Playing Agents 提出PDD框架，通过动态重要性估计增强角色扮演Agent的Persona一致性 large language model
15 Recursive Think-Answer Process for LLMs and VLMs 提出递归思考-回答过程以解决单次推理错误问题 large language model
16 AdaPonderLM: Gated Pondering Language Models with Token-Wise Adaptive Depth AdaPonderLM：提出一种token自适应深度的门控Pondering语言模型，提升推理效率。 large language model
17 Beyond the Resumé: A Rubric-Aware Automatic Interview System for Information Elicitation 提出基于LLM的自动面试系统，用于更有效地筛选候选人 large language model ✅
18 Bootstrapping Embeddings for Low Resource Languages 提出基于LLM的自举方法，为低资源语言生成高质量词嵌入 large language model
19 Beyond the Grid: Layout-Informed Multi-Vector Retrieval with Parsed Visual Document Representations ColParse：利用布局信息的解析视觉文档表示，实现高效多向量检索 multimodal

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

#题目一句话要点标签🔗⭐
20 KDFlow: A User-Friendly and Efficient Knowledge Distillation Framework for Large Language Models 提出KDFlow框架，通过解耦架构和SGLang加速大语言模型知识蒸馏。 distillation large language model ✅
21 CharacterFlywheel: Scaling Iterative Improvement of Engaging and Steerable LLMs in Production CharacterFlywheel：在生产环境中迭代提升LLM的互动性和可控性 reinforcement learning large language model instruction following
22 LongRLVR: Long-Context Reinforcement Learning Requires Verifiable Context Rewards LongRLVR：通过可验证上下文奖励解决长文本强化学习中的梯度消失问题 reinforcement learning large language model ✅
23 Learning to Draft: Adaptive Speculative Decoding with Reinforcement Learning 提出Learning to Draft (LTD)，通过强化学习自适应优化推测解码吞吐量。 reinforcement learning large language model
24 Surgical Post-Training: Cutting Errors, Keeping Knowledge SPoT：通过外科手术式后训练，提升LLM推理能力并保留先验知识 DPO direct preference optimization large language model ✅

⬅️ 返回 cs.CL 首页 · 🏠 返回主页