cs.CL（2026-01-21）

📊 共 24 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) (17 🔗1) 支柱二：RL算法与架构 (RL & Architecture) (4 🔗2) 支柱一：机器人控制 (Robot Control) (2) 支柱三：空间感知与语义 (Perception & Semantics) (1)

🔬 支柱九：具身大模型 (Embodied Foundation Models) (17 篇)

#题目一句话要点标签🔗⭐
1 Self-Blinding and Counterfactual Self-Simulation Mitigate Biases and Sycophancy in Large Language Models 利用自盲和反事实自模拟缓解大语言模型中的偏见和谄媚 large language model
2 Render-of-Thought: Rendering Textual Chain-of-Thought as Images for Visual Latent Reasoning 提出Render-of-Thought，将文本推理链渲染为图像，用于视觉潜在推理。 large language model chain-of-thought ✅
3 Social Caption: Evaluating Social Understanding in Multimodal Models 提出Social Caption框架，评估多模态模型中的社会理解能力 large language model multimodal
4 RSNA Large Language Model Benchmark Dataset for Chest Radiographs of Cardiothoracic Disease: Radiologist Evaluation and Validation Enhanced by AI Labels (REVEAL-CXR) 提出REVEAL-CXR：一个AI辅助的胸部X光片基准数据集，用于评估心胸疾病大语言模型。 large language model multimodal
5 RECAP: Resistance Capture in Text-based Mental Health Counseling with Large Language Models 提出RECAP框架，用于识别文本心理咨询中的阻抗行为并提供解释。 large language model
6 Metadata Conditioned Large Language Models for Localization 提出元数据条件化大语言模型，提升模型在特定地理区域的性能且不牺牲跨区域泛化能力。 large language model
7 Comparative Study of Large Language Models on Chinese Film Script Continuation: An Empirical Analysis Based on GPT-5.2 and Qwen-Max 构建中文电影剧本续写基准，对比GPT-5.2与Qwen-Max在创意写作中的性能差异。 large language model
8 Multi-Agent Constraint Factorization Reveals Latent Invariant Solution Structure 提出多智能体约束分解框架，揭示大模型系统潜在不变解结构 large language model
9 Knowledge Restoration-driven Prompt Optimization: Unlocking LLM Potential for Open-Domain Relational Triplet Extraction 提出知识重建驱动的提示优化框架，提升LLM在开放域关系三元组抽取中的性能。 large language model
10 Obscuring Data Contamination Through Translation: Evidence from Arabic Corpora 提出翻译感知污染检测方法，解决多语言大模型评估中数据污染的盲区问题 large language model
11 CorpusQA: A 10 Million Token Benchmark for Corpus-Level Analysis and Reasoning 提出CorpusQA，一个千万token级别的语料库分析与推理基准。 large language model
12 Say Anything but This: When Tokenizer Betrays Reasoning in LLMs 揭示Tokenizer缺陷：LLM推理中Token化不一致性导致幻影编辑 large language model
13 The Effect of Scripts and Formats on LLM Numeracy 揭示LLM在不同数字脚本和格式下的计算能力退化问题，并提出改进策略 large language model
14 Supporting Humans in Evaluating AI Summaries of Legal Depositions 提出基于Nugget的方法，辅助法律专家评估和改进法律文书摘要。 large language model
15 CodeDelegator: Mitigating Context Pollution via Role Separation in Code-as-Action Agents CodeDelegator：通过角色分离缓解代码即动作Agent中的上下文污染 large language model
16 PodBench: A Comprehensive Benchmark for Instruction-Aware Audio-Oriented Podcast Script Generation PodBench：一个面向指令感知的播客脚本生成综合评测基准 instruction following
17 AdaTIR: Adaptive Tool-Integrated Reasoning via Difficulty-Aware Policy Optimization 提出AdaTIR以解决工具调用冗余问题 large language model

🔬 支柱二：RL算法与架构 (RL & Architecture) (4 篇)

#题目一句话要点标签🔗⭐
18 Rewarding How Models Think Pedagogically: Integrating Pedagogical Reasoning and Thinking Rewards for LLMs in Education PedagogicalRL-Thinking：融合教学推理和思维奖励，优化LLM在教育中的应用 reinforcement learning large language model
19 The Flexibility Trap: Why Arbitrary Order Limits Reasoning Potential in Diffusion Language Models 揭示扩散语言模型中任意顺序生成导致推理能力受限的“灵活性陷阱” reinforcement learning large language model ✅
20 Language-Coupled Reinforcement Learning for Multilingual Retrieval-Augmented Generation 提出LcRL框架，通过语言耦合强化学习提升多语言检索增强生成效果 reinforcement learning ✅
21 DARL: Encouraging Diverse Answers for General Reasoning without Verifiers DARL：无需验证器，鼓励生成多样化答案的通用推理强化学习框架 reinforcement learning large language model

🔬 支柱一：机器人控制 (Robot Control) (2 篇)

#题目一句话要点标签🔗⭐
22 Robust Fake News Detection using Large Language Models under Adversarial Sentiment Attacks 提出AdSent框架，解决大型语言模型对抗性情感攻击下的假新闻稳健检测问题 manipulation large language model
23 SearchGym: Bootstrapping Real-World Search Agents via Cost-Effective and High-Fidelity Environment Simulation SearchGym：通过高保真、低成本环境模拟引导现实世界搜索Agent sim-to-real reinforcement learning curriculum learning

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

#题目一句话要点标签🔗⭐
24 Taxonomy-Aligned Risk Extraction from 10-K Filings with Autonomous Improvement Using LLMs 提出一种基于LLM的企业10-K文件中风险因素的分类提取与自主改进方法。 semantic mapping semantic map

⬅️ 返回 cs.CL 首页 · 🏠 返回主页