cs.CL（2026-04-15）

📊 共 29 篇论文 | 🔗 4 篇有代码

🎯 兴趣领域导航

支柱九：具身大模型 (Embodied Foundation Models) (26 🔗3) 支柱二：RL算法与架构 (RL & Architecture) (3 🔗1)

🔬 支柱九：具身大模型 (Embodied Foundation Models) (26 篇)

#题目一句话要点标签🔗⭐
1 MERRIN: A Benchmark for Multimodal Evidence Retrieval and Reasoning in Noisy Web Environments MERRIN：用于评估噪声Web环境中多模态证据检索与推理的基准 multimodal
2 Adaptive Conformal Prediction for Improving Factuality of Generations by Large Language Models 提出自适应共形预测方法，提升大语言模型生成结果的事实性。 large language model
3 From Anchors to Supervision: Memory-Graph Guided Corpus-Free Unlearning for Large Language Models 提出MAGE框架，通过记忆图引导的无语料库卸载，解决大语言模型的信息遗忘问题 large language model
4 Correct Prediction, Wrong Steps? Consensus Reasoning Knowledge Graph for Robust Chain-of-Thought Synthesis CRAFT：基于共识推理知识图谱的链式思考合成方法，提升LLM推理鲁棒性 chain-of-thought
5 Dual-Enhancement Product Bundling: Bridging Interactive Graph and Large Language Model 提出双重增强产品捆绑方法，融合交互图学习与大语言模型，提升电商推荐效果。 large language model
6 Beyond Static Personas: Situational Personality Steering for Large Language Models IRIS：面向大语言模型的情境化人格引导框架，无需训练。 large language model
7 Robust Reward Modeling for Large Language Models via Causal Decomposition 提出基于因果分解的鲁棒奖励模型，提升大语言模型对齐效果 large language model
8 Foresight Optimization for Strategic Reasoning in Large Language Models 提出FoPO，增强大语言模型在多智能体环境下的战略推理能力 large language model
9 Empirical Evidence of Complexity-Induced Limits in Large Language Models on Finite Discrete State-Space Problems with Explicit Validity Constraints 提出受控基准测试框架，揭示大语言模型在复杂性增加时推理能力崩溃现象 large language model
10 TLoRA+: A Low-Rank Parameter-Efficient Fine-Tuning Method for Large Language Models 提出TLoRA+，一种高效低秩参数微调方法，提升大语言模型在特定任务上的性能。 large language model
11 MedRCube: A Multidimensional Framework for Fine-Grained and In-Depth Evaluation of MLLMs in Medical Imaging MedRCube：用于医学影像中多模态大语言模型细粒度和深度评估的多维框架 large language model multimodal ✅
12 Synthesizing Instruction-Tuning Datasets with Contrastive Decoding 提出CoDIT方法，通过对比解码合成指令调优数据集，提升模型指令遵循能力。 large language model instruction following
13 From Relevance to Authority: Authority-aware Generative Retrieval in Web Search Engines 提出AuthGR框架，将权威性融入生成式检索，提升Web搜索的可靠性。 large language model multimodal
14 Rhetorical Questions in LLM Representations: A Linear Probing Study 通过线性探测研究LLM中反问句的表征，揭示其多重编码特性 large language model
15 How Can We Synthesize High-Quality Pretraining Data? A Systematic Study of Prompt Design, Generator Model, and Source Data 系统性研究提示词设计、生成模型与源数据对合成预训练数据质量的影响，并提出FinePhrase数据集。 large language model
16 Calibrated Speculative Decoding: Frequency-Guided Candidate Selection for Efficient Inference 提出校准推测解码CSD，通过频率引导候选选择加速LLM推理。 large language model
17 IndicDB -- Benchmarking Multilingual Text-to-SQL Capabilities in Indian Languages 提出IndicDB，用于评估印度语言多语言Text-to-SQL能力。 large language model
18 BenGER: A Collaborative Web Platform for End-to-End Benchmarking of German Legal Tasks BenGER：一个协同Web平台，用于端到端评测德语法律任务的大语言模型 large language model
19 Interpretable Stylistic Variation in Human and LLM Writing Across Genres, Models, and Decoding Strategies 通过词汇语法特征分析人类与LLM在不同领域和策略下的写作风格差异 large language model
20 From Where Words Come: Efficient Regularization of Code Tokenizers Through Source Attribution 提出SA-BPE，通过源属性正则化代码Tokenizer，减少低训练token。 large language model
21 ToolOmni: Enabling Open-World Tool Use via Agentic learning with Proactive Retrieval and Grounded Execution ToolOmni：通过主动检索和具身执行的 Agentic 学习实现开放世界工具使用 large language model
22 QuantileMark: A Message-Symmetric Multi-bit Watermark for LLMs QuantileMark：一种消息对称的LLM多比特水印方案，提升水印鲁棒性。 large language model ✅
23 Co-FactChecker: A Framework for Human-AI Collaborative Claim Verification Using Large Reasoning Models 提出Co-FactChecker框架，利用人类专家反馈协同增强大型推理模型的事实核查能力 large language model
24 Breaking the Generator Barrier: Disentangled Representation for Generalizable AI-Text Detection 提出解耦表征框架DRGD，提升AI文本检测在未知生成器上的泛化能力 large language model ✅
25 YOCO++: Enhancing YOCO with KV Residual Connections for Efficient LLM Inference YOCO++：利用KV残差连接增强YOCO，实现高效LLM推理 large language model
26 ToolSpec: Accelerating Tool Calling via Schema-Aware and Retrieval-Augmented Speculative Decoding ToolSpec：通过模式感知和检索增强的推测解码加速工具调用 large language model

🔬 支柱二：RL算法与架构 (RL & Architecture) (3 篇)

#题目一句话要点标签🔗⭐
27 Training-Free Test-Time Contrastive Learning for Large Language Models 提出TF-TTCL，一种免训练的测试时对比学习框架，提升大语言模型在分布偏移下的推理能力。 contrastive learning distillation large language model ✅
28 From Prediction to Justification: Aligning Sentiment Reasoning with Human Rationale via Reinforcement Learning 提出ABSA-R1框架，通过强化学习对齐情感推理与人类理性，提升情感分析可解释性与性能。 reinforcement learning large language model
29 Debate to Align: Reliable Entity Alignment through Two-Stage Multi-Agent Debate 提出AgentEA，通过多智能体辩论提升知识图谱实体对齐的可靠性。 representation learning large language model

⬅️ 返回 cs.CL 首页 · 🏠 返回主页