cs.CL（2026-05-06）

📊 共 24 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) (17 🔗1) 支柱二：RL算法与架构 (RL & Architecture) (6 🔗1) 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 Anticipating Innovation Using Large Language Models 提出TechToken模型，利用专利语言预测未来技术组合创新。 large language model
2 Gradients with Respect to Semantics Preserving Embeddings Tell the Uncertainty of Large Language Models 提出SemGrad，一种基于语义梯度的无采样LLM不确定性量化方法，提升效率和准确性。 large language model
3 Detecting Hallucinations in Large Language Models via Internal Attention Divergence Signals 提出基于内部注意力发散信号的大语言模型幻觉检测方法 large language model
4 Adapting Large Language Models to a Low-Resource Agglutinative Language: A Comparative Study of LoRA and QLoRA for Bashkir 针对低资源黏着语巴什基尔语，对比LoRA与QLoRA微调大语言模型的效果。 large language model
5 Misaligned by Reward: Socially Undesirable Preferences in LLMs 揭示奖励模型社会偏见：LLM中社会不良偏好的评估框架 large language model instruction following
6 Assessing Cognitive Effort in L2 Idiomatic Processing: An Eye-Tracking Dataset 构建眼动追踪数据集，评估二语习得者成语理解的认知负荷 large language model
7 PSK at SemEval-2026 Task 9: Multilingual Polarization Detection Using Ensemble Gemma Models with Synthetic Data Augmentation 提出基于集成Gemma模型和合成数据增强的多语言极化检测方法，在SemEval-2026任务中取得优异成绩。 large language model
8 UniVer: A Unified Perspective for Multi-step and Multi-draft Speculative Decoding 提出UniVer，通过条件最优传输统一多步多草稿推测解码，提升LLM推理效率。 large language model
9 StoryAlign: Evaluating and Training Reward Models for Story Generation StoryAlign提出StoryReward模型和StoryRMB基准，提升故事生成中人类偏好对齐 large language model ✅
10 GEM: Graph-Enhanced Mixture-of-Experts with ReAct Agents for Dialogue State Tracking GEM：图增强混合专家模型，结合ReAct智能体，提升对话状态追踪性能 large language model
11 Beyond Semantics: An Evidential Reasoning-Aware Multi-View Learning Framework for Trustworthy Mental Health Prediction 提出一种基于证据推理的多视角学习框架，用于可信赖的精神健康预测。 large language model
12 The Pinocchio Dimension: Phenomenality of Experience as the Primary Axis of LLM Psychometric Differences 通过心理测量揭示LLM的“匹诺曹维度”，区分体验丰富度和行为反应 large language model
13 Why Expert Alignment Is Hard: Evidence from Subjective Evaluation 研究专家对齐的困难：主观评估中的证据揭示专家判断的异质性与不稳定性 large language model
14 BenCSSmark: Making the Social Sciences Count in LLM Research BenCSSmark：通过社会科学任务提升LLM的评估与泛化能力 large language model
15 Elicitation Matters: How Prompts and Query Protocols Shape LLM Surrogates under Sparse Observations 研究提示词和查询协议如何影响稀疏观测下LLM代理模型的性能 large language model
16 Paraphrase-Induced Output-Mode Collapse: When LLMs Break Character Under Semantically Equivalent Inputs 揭示大语言模型在语义等价输入下输出模式崩溃问题，并提出PARACONSIST基准进行评估。 large language model
17 Graph-Augmented LLMs for Swiss MP Ideology Prediction 提出PG-RAG框架，利用图增强LLM预测瑞士议员意识形态立场 large language model

🔬 支柱二：RL算法与架构 (RL & Architecture) (6 篇)

#题目一句话要点标签🔗⭐
18 RLearner-LLM: Balancing Logical Grounding and Fluency in Large Language Models via Hybrid Direct Preference Optimization RLearner-LLM：通过混合直接偏好优化平衡大语言模型的逻辑性和流畅性 PPO RLHF DPO
19 Gyan: An Explainable Neuro-Symbolic Language Model Gyan：一种可解释的神经符号语言模型，无需Transformer即可实现SOTA性能。 world model world models large language model
20 TabEmbed: Benchmarking and Learning Generalist Embeddings for Tabular Understanding TabEmbed：提出通用表格嵌入模型，统一表格理解的分类与检索任务。 representation learning contrastive learning foundation model ✅
21 Automatically Finding and Validating Unexpected Side-Effects of Interventions on Language Models 提出一种自动化的对比评估流程，用于审计干预措施对大型语言模型行为的影响。 distillation large language model
22 Distilling Bayesian Belief States into Language Models for Auditable Negotiation 提出BOND框架，通过知识蒸馏提升谈判Agent信念状态的可审计性 distillation large language model
23 Every Step Counts: Step-Level Credit Assignment for Tool-Integrated Text-to-SQL 提出FineStep框架，解决工具集成Text-to-SQL中步骤级信用分配问题。 reinforcement learning reward design

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

#题目一句话要点标签🔗⭐
24 Implicit Representations of Grammaticality in Language Models 提出隐式语法性表示以提升语言模型的判断能力 implicit representation

⬅️ 返回 cs.CL 首页 · 🏠 返回主页