cs.CL（2024-05-30）

📊 共 44 篇论文 | 🔗 11 篇有代码

🎯 兴趣领域导航

支柱九：具身大模型 (Embodied Foundation Models) (32 🔗7) 支柱二：RL算法与架构 (RL & Architecture) (12 🔗4)

🔬 支柱九：具身大模型 (Embodied Foundation Models) (32 篇)

#题目一句话要点标签🔗⭐
1 Jailbreaking Large Language Models Against Moderation Guardrails via Cipher Characters 提出JAMBench和JAM方法，提升大语言模型对抗恶意提示的防御能力 large language model
2 How Multilingual Are Large Language Models Fine-Tuned for Translation? 研究翻译微调对大型语言模型多语言翻译能力的影响 large language model
3 Confidence-Aware Sub-Structure Beam Search (CABS): Mitigating Hallucination in Structured Data Generation with Large Language Models 提出置信度感知子结构束搜索(CABS)，缓解LLM在结构化数据生成中的幻觉问题。 large language model
4 JoPA:Explaining Large Language Model's Generation via Joint Prompt Attribution 提出JoPA框架，通过联合提示归因解释大型语言模型生成结果的影响因素。 large language model
5 GNN-RAG: Graph Neural Retrieval for Large Language Model Reasoning 提出GNN-RAG框架，结合图神经网络和大型语言模型进行知识图谱问答 large language model
6 Is In-Context Learning Sufficient for Instruction Following in LLMs? 研究表明，上下文学习在指令跟随任务中仍逊于指令微调，并揭示了解码参数的关键作用。 instruction following ✅
7 SeamlessExpressiveLM: Speech Language Model for Expressive Speech-to-Speech Translation with Chain-of-Thought 提出SeamlessExpressiveLM，利用思维链提示实现富有表现力的端到端语音翻译。 chain-of-thought
8 ANAH: Analytical Annotation of Hallucinations in Large Language Models ANAH：提出用于大语言模型幻觉分析标注的双语数据集 large language model
9 Quest: Query-centric Data Synthesis Approach for Long-context Scaling of Large Language Model Quest：一种面向查询的数据合成方法，用于扩展大型语言模型的长上下文能力 large language model
10 Significance of Chain of Thought in Gender Bias Mitigation for English-Dravidian Machine Translation 利用思维链（Chain of Thought）缓解英-德拉维机器翻译中的性别偏见 chain-of-thought
11 Detecting Hallucinations in Large Language Model Generation: A Token Probability Approach 提出基于Token概率的轻量级方法，用于检测大语言模型生成内容中的幻觉。 large language model ✅
12 TAIA: Large Language Models are Out-of-Distribution Data Learners TAIA：利用大语言模型进行领域外数据学习，提升下游任务性能 large language model ✅
13 One Token Can Help! Learning Scalable and Pluggable Virtual Tokens for Retrieval-Augmented Large Language Models 提出可扩展和可插拔的虚拟Token，用于增强检索式大语言模型，提升效果并保持通用性。 large language model
14 PATIENT-Ψ: Using Large Language Models to Simulate Patients for Training Mental Health Professionals PATIENT-Ψ：利用大语言模型模拟患者，用于精神健康专业人员的认知行为疗法训练。 large language model ✅
15 Retrieval Augmented Structured Generation: Business Document Information Extraction As Tool Use 提出检索增强结构化生成框架RASG，解决商业文档信息抽取难题 large language model multimodal
16 X-Instruction: Aligning Language Model in Low-resource Languages with Self-curated Cross-lingual Instructions 提出X-Instruction，通过自策跨语言指令对低资源语言大模型进行对齐。 large language model instruction following
17 Jina CLIP: Your CLIP Model Is Also Your Text Retriever 提出多任务对比学习方法，使CLIP模型同时具备优秀的图文和文本检索能力 multimodal
18 SPOT: Text Source Prediction from Originality Score Thresholding 提出SPOT以解决文本来源预测问题 large language model
19 Transfer Q Star: Principled Decoding for LLM Alignment 提出Transfer Q Star，通过迁移学习进行大语言模型对齐的原则性解码 foundation model
20 Student Answer Forecasting: Transformer-Driven Answer Choice Prediction for Language Learning 提出MCQStudentBert，用于语言学习中学生答案选择预测，提升个性化教学。 large language model
21 From Symbolic Tasks to Code Generation: Diversification Yields Better Task Performers 通过指令多样化提升代码生成能力：从符号任务到代码生成 large language model
22 SpecDec++: Boosting Speculative Decoding via Adaptive Candidate Lengths SpecDec++通过自适应候选长度提升推测解码效率，加速大语言模型推理。 large language model ✅
23 Reasoning about concepts with LLMs: Inconsistencies abound 揭示LLM概念理解不一致性，提出基于知识图谱的prompt策略提升模型性能 large language model
24 SLM as Guardian: Pioneering AI Safety with Small Language Models 利用小型语言模型作为安全卫士，探索AI安全新途径 large language model
25 CharacterGPT: A Persona Reconstruction Framework for Role-Playing Agents 提出CharacterGPT框架以解决角色扮演代理中的角色一致性问题 large language model ✅
26 PertEval: Unveiling Real Knowledge Capacity of LLMs with Knowledge-Invariant Perturbations PertEval：通过知识不变扰动揭示大语言模型真实知识能力 large language model ✅
27 Automated Focused Feedback Generation for Scientific Writing Assistance 提出SWIF$^{2}$T，用于自动生成针对科学写作的聚焦反馈，辅助科研新手改进论文。 large language model
28 Cutting Through the Noise: Boosting LLM Performance on Math Word Problems 提出PROBLEMATHIC数据集并微调LLM，提升其在含噪声数学应用题上的鲁棒性 large language model
29 Who Writes the Review, Human or AI? 提出基于迁移学习的方法，用于区分人类撰写和AI生成的书评 large language model
30 Towards Hierarchical Multi-Agent Workflows for Zero-Shot Prompt Optimization 提出层级多智能体工作流HMAW，实现零样本提示优化，提升LLM在开放场景下的性能。 large language model
31 Divide-and-Conquer Meets Consensus: Unleashing the Power of Functions in Code Generation 提出FunCoder以解决复杂代码生成问题 large language model
32 DevEval: A Manually-Annotated Code Generation Benchmark Aligned with Real-World Code Repositories DevEval：一个与真实代码仓库对齐的手动标注代码生成基准 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🔗⭐
33 Beyond Imitation: Learning Key Reasoning Steps from Dual Chain-of-Thoughts in Reasoning Distillation 提出EDIT方法，通过错误驱动的关键推理步骤蒸馏提升小模型推理能力。 distillation large language model chain-of-thought ✅
34 TS-Align: A Teacher-Student Collaborative Framework for Scalable Iterative Finetuning of Large Language Models TS-Align：一种可扩展的迭代微调框架，用于对齐大型语言模型 teacher-student large language model instruction following
35 Towards Ontology-Enhanced Representation Learning for Large Language Models 提出本体增强表示学习方法，提升LLM在特定领域的知识建模能力 representation learning contrastive learning large language model
36 InstructionCP: A fast approach to transfer Large Language Models into target language 提出InstructionCP，加速大语言模型向目标语言迁移，同时保留对话能力。 reinforcement learning RLHF large language model
37 Improve Student's Reasoning Generalizability through Cascading Decomposed CoTs Distillation 提出CasCoD，通过级联分解CoT蒸馏提升学生模型推理泛化性 teacher-student distillation large language model ✅
38 Evaluating Large Language Model Biases in Persona-Steered Generation 评估大型语言模型在角色引导生成中的偏见，揭示其在不协调角色上的可控性不足。 reinforcement learning RLHF large language model
39 Enhancing Reinforcement Learning with Label-Sensitive Reward for Natural Language Understanding 提出RLLR框架，利用标签敏感奖励增强强化学习在自然语言理解任务中的性能 reinforcement learning RLHF large language model ✅
40 Group Robust Preference Optimization in Reward-free RLHF 提出GRPO方法，在无奖励RLHF中实现对不同群体偏好的鲁棒优化。 reinforcement learning RLHF direct preference optimization
41 Would I Lie To You? Inference Time Alignment of Language Models using Direct Preference Heads 提出Direct Preference Heads，在不影响语言模型输出分布的前提下，提升模型对人类偏好的对齐。 reinforcement learning RLHF DPO
42 Safe Multi-agent Reinforcement Learning with Natural Language Constraints 提出SMALL方法，利用自然语言约束实现安全多智能体强化学习。 reinforcement learning policy learning
43 Xwin-LM: Strong and Scalable Alignment Practice for LLMs Xwin-LM：用于大型语言模型对齐的强大且可扩展的实践方法 DPO direct preference optimization large language model ✅
44 GKT: A Novel Guidance-Based Knowledge Transfer Framework For Efficient Cloud-edge Collaboration LLM Deployment 提出GKT框架，通过引导式知识迁移实现高效云边协同LLM部署 distillation large language model

⬅️ 返回 cs.CL 首页 · 🏠 返回主页