cs.CL（2026-10-09）

📊 共 18 篇论文 | 🔗 2 篇有代码

🎯 兴趣领域导航

支柱九：具身大模型 (Embodied Foundation Models) (11 🔗1) 支柱二：RL算法与架构 (RL & Architecture) (5 🔗1) 支柱三：空间感知与语义 (Perception & Semantics) (1) 支柱六：视频提取与匹配 (Video Extraction) (1)

🔬 支柱九：具身大模型 (Embodied Foundation Models) (11 篇)

#题目一句话要点标签🔗⭐
1 Neurograf v1: An End-to-End Romanian-First Language Model Family 提出Neurograf v1以解决罗马尼亚低资源语言模型训练问题 instruction following
2 What Should SFT Learn? Conditional Information Gain for Generalizable Post-Training 提出CIG-SFT以解决后训练模型泛化能力不足的问题 large language model
3 Task Context Shapes Component Roles in Language Models 提出任务上下文视角以解析语言模型组件角色 large language model
4 Inject a Turn, Not a Token: Learning Where to Redirect LLM Reasoning for Self-Correction 提出UPTURN框架以解决LLM自我纠错问题 large language model
5 Surfing the Activation Wave: Analyzing the Information and Geometry of LLM Hidden States for Speculative Decoding 提出基于信息和几何分析的多层隐藏状态选择方法以优化LLM的推测解码 large language model
6 Tracing and Taming Hallucinated Terms via Subsequence-Association Filtering and Breadth-Aware Selection for LLM Query Expansion 提出SAFE-QE以解决大语言模型的幻觉术语问题 large language model
7 Towards the Roof of the World: NLP Progress in Tibetan Languages 系统性调查西藏语言自然语言处理的研究进展 large language model
8 MoPrO-SD: Modular Prompt Optimization for Stance Detection 提出MoPrO-SD框架以优化立场检测中的提示设计 large language model
9 Test-Time Memory Construction 提出T$^2$Mem以解决长交互历史中的记忆构建问题 large language model ✅
10 Host-Aware Web Crawling for LLM Pretraining Data 提出主机感知网络爬虫以优化LLM预训练数据收集 large language model
11 Evaluating LLM Creativity through Long-Tail Performance 提出长尾性能评估方法以提升大语言模型创造力 large language model

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

#题目一句话要点标签🔗⭐
12 RSI-Anything: Unlocking Recursive Self-Improvement with World Models 提出RSI-Anything以解决递归自我改进中的反馈成本问题 world model world models JEPA
13 Can Language Models Learn When to Abstain? Policy Learning and Calibrated Answer Selection 提出学习选择答案的模型以解决语言模型的回答不确定性问题 reinforcement learning policy learning
14 Proactive Agentic Orchestration 提出POrchestra框架以解决现有代理编排的适应性不足问题 distillation large language model ✅
15 PreSTO: Predictive Subtree Prefetching for Fast LLM Power Sampling 提出PreSTO以加速LLM的PowerMH采样 distillation large language model
16 SOAR: Source-Aware Reinforcement Learning with Precise Credit Assignment for Search Agents 提出SOAR以解决搜索代理中的证据伪造问题 reinforcement learning

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

#题目一句话要点标签🔗⭐
17 Oído: Open-Vocabulary Speech Recognition on a $5 Microcontroller without a Neural Accelerator 提出Oído以实现低成本微控制器上的开放词汇语音识别 open-vocabulary open vocabulary

🔬 支柱六：视频提取与匹配 (Video Extraction) (1 篇)

#题目一句话要点标签🔗⭐
18 Unreliable Narrator Framing as a Cost-effective Mechanism for Satire Explanation of Australian and American Satirical News 提出不可靠叙述者框架以有效解释澳美讽刺新闻 HuMoR

⬅️ 返回 cs.CL 首页 · 🏠 返回主页