cs.CL（2025-09-28）

📊 共 33 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) (21) 支柱二：RL算法与架构 (RL & Architecture) (12 🔗2)

🔬 支柱九：具身大模型 (Embodied Foundation Models) (21 篇)

#题目一句话要点标签🔗⭐
1 Do LLMs Understand Romanian Driving Laws? A Study on Multimodal and Fine-Tuned Question Answering 评估LLM在罗马尼亚驾驶法规问答任务中的表现，并探索领域微调和多模态输入的影响。 large language model multimodal
2 Compose and Fuse: Revisiting the Foundational Bottlenecks in Multimodal Reasoning 提出逻辑基础评估框架以解决多模态推理瓶颈问题 large language model multimodal
3 A Cross-Lingual Analysis of Bias in Large Language Models Using Romanian History 通过罗马尼亚历史案例，跨语言分析大型语言模型中的偏见 large language model
4 Spiral of Silence in Large Language Model Agents 提出LLM Agent螺旋沉默评估框架，揭示历史和人设信号对群体意见的影响 large language model
5 Assessing Large Language Models in Updating Their Forecasts with New Information EVOLVECAST框架评估LLM在接收新信息后预测更新能力，揭示其保守偏差 large language model
6 Jackal: A Real-World Execution-Based Benchmark Evaluating Large Language Models on Text-to-JQL Tasks 提出Jackal：一个基于真实执行的文本到JQL的大语言模型评测基准 large language model
7 GEAR: A General Evaluation Framework for Abductive Reasoning 提出GEAR：通用且无标签的归纳推理评估框架，并用于提升LLM的推理能力。 large language model instruction following
8 Large-Scale Constraint Generation -- Can LLMs Parse Hundreds of Constraints? 提出大规模约束生成问题LSCG，并设计FoCusNet提升LLM在复杂约束下的解析能力。 large language model chain-of-thought
9 Vision-Grounded Machine Interpreting: Improving the Translation Process through Visual Cues 提出视觉 grounding 的机器同声传译方法，利用视觉信息提升翻译质量 multimodal visual grounding
10 EduVidQA: Generating and Evaluating Long-form Answers to Student Questions based on Lecture Videos 提出EduVidQA数据集，利用多模态大语言模型解决讲座视频问答难题 large language model multimodal
11 Open-DeBias: Toward Mitigating Open-Set Bias in Language Models 提出Open-DeBias，缓解语言模型中开放域偏见问题 large language model zero-shot transfer
12 VIVA+: Human-Centered Situational Decision-Making VIVA+：面向以人为本场景的认知驱动多模态决策基准 large language model multimodal
13 The Role of Logic and Automata in Understanding Transformers 综述Transformer能力：逻辑、自动机与电路复杂性的视角 large language model
14 From Personal to Collective: On the Role of Local and Global Memory in LLM Personalization 提出LoGo框架，融合局部与全局记忆，提升LLM个性化效果 large language model
15 DiffuSpec: Unlocking Diffusion Language Models for Speculative Decoding DiffuSpec：利用扩散语言模型解锁推测解码，显著提升LLM推理速度。 large language model
16 Aligning LLMs for Multilingual Consistency in Enterprise Applications 提出批量对齐微调策略，解决企业应用中LLM多语言一致性问题 large language model
17 Pragmatic Inference for Moral Reasoning Acquisition: Generalization via Distributional Semantics 提出基于语用推理的道德推理获取方法，提升LLM的道德推理泛化能力 large language model
18 Winning the Pruning Gamble: A Unified Approach to Joint Sample and Token Pruning for Efficient Supervised Fine-Tuning 提出Q-Tuning，联合优化样本和Token剪枝，提升大模型SFT效率。 large language model
19 Bridging the Knowledge-Prediction Gap in LLMs on Multiple-Choice Questions 提出KAPPA以解决大语言模型在多项选择题上的知识预测差距问题 large language model
20 Evaluating Program Semantics Reasoning with Type Inference in System F 提出TF-Bench基准测试，评估LLM在System F类型推断中的程序语义推理能力 large language model
21 LLM Hallucination Detection: HSAD 提出HSAD，通过频域分析LLM隐藏层信号以检测幻觉 large language model

🔬 支柱二：RL算法与架构 (RL & Architecture) (12 篇)

#题目一句话要点标签🔗⭐
22 Toward Preference-aligned Large Language Models via Residual-based Model Steering 提出基于残差引导的偏好对齐方法以提升大语言模型性能 reinforcement learning DPO direct preference optimization
23 Fast Thinking for Large Language Models 提出基于隐式码本的快速推理框架，提升大语言模型推理效率。 reinforcement learning large language model chain-of-thought
24 Knowledge-Level Consistency Reinforcement Learning: Dual-Fact Alignment for Long-Form Factuality 提出知识一致性强化学习框架KLCF，解决长文本生成中大语言模型的幻觉问题。 reinforcement learning RLHF reward design
25 Beyond Magic Words: Sharpness-Aware Prompt Evolving for Robust Large Language Models with TARE 提出TARE框架，通过文本锐度感知的Prompt演化，提升大语言模型在语义扰动下的鲁棒性。 reinforcement learning large language model
26 Beyond English-Centric Training: How Reinforcement Learning Improves Cross-Lingual Reasoning in LLMs 强化学习提升LLM跨语言推理能力，优于监督微调 reinforcement learning large language model foundation model
27 The Hidden Costs of Translation Accuracy: Distillation, Quantization, and Environmental Impact 通过蒸馏和量化，在机器翻译中降低模型计算成本和环境影响 distillation large language model
28 Towards Efficient CoT Distillation: Self-Guided Rationale Selector for Better Performance with Fewer Rationales 提出MoRSD，通过自引导的理由选择器实现高效CoT蒸馏，以更少的理由获得更好的性能。 distillation chain-of-thought ✅
29 ByteSized32Refactored: Towards an Extensible Interactive Text Games Corpus for LLM World Modeling and Evaluation ByteSized32Refactored：面向LLM世界建模与评估的可扩展交互式文本游戏语料库 world model large language model
30 HiPO: Hybrid Policy Optimization for Dynamic Reasoning in LLMs HiPO：面向LLM动态推理的混合策略优化框架，提升效率并保持精度。 reinforcement learning large language model chain-of-thought
31 SPELL: Self-Play Reinforcement Learning for evolving Long-Context Language Models 提出SPELL框架，通过自博弈强化学习提升长文本语言模型的推理能力。 reinforcement learning large language model
32 Taming Masked Diffusion Language Models via Consistency Trajectory Reinforcement Learning with Fewer Decoding Step 提出EOSER、ASS与CJ-GRPO，提升Masked Diffusion语言模型推理效率与一致性。 reinforcement learning ✅
33 Dual-Scale World Models for LLM Agents Towards Hard-Exploration Problems 提出GLoW，利用双尺度世界模型解决LLM Agent在复杂探索任务中的难题 world model

⬅️ 返回 cs.CL 首页 · 🏠 返回主页