cs.CL（2026-04-29）

📊 共 13 篇论文

🎯 兴趣领域导航

支柱九：具身大模型 (Embodied Foundation Models) (8) 支柱二：RL算法与架构 (RL & Architecture) (4) 支柱一：机器人控制 (Robot Control) (1)

🔬 支柱九：具身大模型 (Embodied Foundation Models) (8 篇)

#题目一句话要点标签🔗⭐
1 Benchmarking Complex Multimodal Document Processing Pipelines: A Unified Evaluation Framework for Enterprise AI EnterpriseDocBench：构建企业级文档AI流水线的统一评测框架 multimodal
2 Multimodal LLMs are not all you need for Pediatric Speech Language Pathology 针对儿童言语病理，提出基于语音表征模型的SSD分类方法，性能优于多模态LLM。 multimodal
3 MoRFI: Monotonic Sparse Autoencoder Feature Identification 提出MoRFI方法，用于识别微调后大语言模型中导致幻觉的单调稀疏自编码器特征。 large language model
4 Tree-of-Text: A Tree-based Prompting Framework for Table-to-Text Generation in the Sports Domain 提出Tree-of-Text框架，解决体育领域表格到文本生成中的幻觉问题 large language model
5 Naamah: A Large Scale Synthetic Sanskrit NER Corpus via DBpedia Seeding and LLM Generation 提出Naamah，一个基于DBpedia和LLM生成的大规模梵语命名实体识别合成数据集。 large language model
6 DSIPA: Detecting LLM-Generated Texts via Sentiment-Invariant Patterns Divergence Analysis 提出DSIPA，通过情感不变模式差异分析检测LLM生成文本，无需训练。 large language model
7 StratMem-Bench: Evaluating Strategic Memory Use in Virtual Character Conversation Beyond Factual Recall 提出StratMem-Bench基准，用于评估虚拟角色对话中策略性记忆使用能力 large language model
8 SafeReview: Defending LLM-based Review Systems Against Adversarial Hidden Prompts SafeReview：提出对抗性框架防御LLM评审系统中恶意隐藏提示 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🔗⭐
9 Turning the TIDE: Cross-Architecture Distillation for Diffusion Large Language Models TIDE：面向扩散大语言模型的跨架构蒸馏框架，提升小模型性能 distillation large language model
10 TLPO: Token-Level Policy Optimization for Mitigating Language Confusion in Large Language Models 提出TLPO，通过Token级策略优化缓解大语言模型中的语言混淆问题 DPO large language model
11 What Kind of Language is Easy to Language-Model Under Curriculum Learning? 研究课程学习对语言模型学习不同类型语言难易程度的影响 curriculum learning
12 Shorthand for Thought: Compressing LLM Reasoning via Entropy-Guided Supertokens 提出基于熵引导Supertokens的LLM推理压缩方法，提升推理效率。 reward shaping large language model

🔬 支柱一：机器人控制 (Robot Control) (1 篇)

#题目一句话要点标签🔗⭐
13 Breaking the Autoregressive Chain: Hyper-Parallel Decoding for Efficient LLM-Based Attribute Value Extraction 提出超并行解码（HPD）算法，加速基于LLM的属性值抽取任务。 manipulation

⬅️ 返回 cs.CL 首页 · 🏠 返回主页