cs.CL（2024-12-10）

📊 共 41 篇论文 | 🔗 3 篇有代码

🎯 兴趣领域导航

支柱九：具身大模型 (Embodied Foundation Models) (38 🔗3) 支柱一：机器人控制 (Robot Control) (2) 支柱二：RL算法与架构 (RL & Architecture) (1)

🔬 支柱九：具身大模型 (Embodied Foundation Models) (38 篇)

#题目一句话要点标签🔗⭐
1 How to Choose a Threshold for an Evaluation Metric for Large Language Models 提出一种基于风险管理的LLM评估指标阈值选择方法，保障模型可靠性 large language model
2 Rethinking Emotion Annotations in the Era of Large Language Models 利用大型语言模型辅助情感标注，提升标注质量与效率 large language model
3 Zero-Shot ATC Coding with Large Language Models for Clinical Assessments 提出基于LLM的零样本ATC编码方法，解决临床评估中人工编码瓶颈。 large language model
4 Searching for Structure: Investigating Emergent Communication with Large Language Models 利用大型语言模型探索涌现通信结构，模拟语言进化。 large language model
5 Generating Knowledge Graphs from Large Language Models: A Comparative Study of GPT-4, LLaMA 2, and BERT 利用大型语言模型生成知识图谱，提升GraphRAG性能 large language model
6 CMT: A Memory Compression Method for Continual Knowledge Learning of Large Language Models 提出CMT：一种面向大语言模型持续知识学习的记忆压缩方法 large language model
7 SpecFuse: Ensembling Large Language Models via Next-Segment Prediction SpecFuse：通过下一片段预测集成大型语言模型，提升生成质量。 large language model
8 Towards Predictive Communication with Brain-Computer Interfaces integrating Large Language Models 结合大型语言模型的脑机接口预测通信方法研究综述 large language model
9 Automatic Item Generation for Personality Situational Judgment Tests with Large Language Models 利用GPT-4自动生成人格情境判断测验，提升心理测评效率与质量 large language model
10 The Rise and Down of Babel Tower: Investigating the Evolution Process of Multilingual Code Large Language Model 提出“巴别塔假说”，揭示多语言代码大模型能力演进过程并优化预训练语料。 large language model
11 Multimodal Sentiment Analysis Based on Causal Reasoning 提出基于因果推理的对抗多模态情感分析框架，解决单模态数据偏差问题。 multimodal
12 PrisonBreak: Jailbreaking Large Language Models with at Most Twenty-Five Targeted Bit-flips PrisonBreak：仅需少量目标比特翻转即可破解大型语言模型的安全对齐 large language model
13 Breaking the Stage Barrier: A Novel Single-Stage Approach to Long Context Extension for Large Language Models 提出HARPE：一种单阶段长文本建模方法，突破LLM上下文长度限制 large language model
14 ChocoLlama: Lessons Learned From Teaching Llamas Dutch ChocoLlama：探索将Llama模型适配到低资源荷兰语的策略 large language model foundation model
15 Identifying and Manipulating Personality Traits in LLMs Through Activation Engineering 通过激活工程识别和操纵LLM中的人格特质 large language model
16 HalluCana: Fixing LLM Hallucination with A Canary Lookahead HalluCana：利用Canary Lookahead机制修正大语言模型的事实性幻觉 large language model
17 Forking Paths in Neural Text Generation 提出一种新方法，无需微调或访问模型权重即可评估LLM文本生成中的不确定性，并发现关键分叉token。 large language model
18 Asking Again and Again: Exploring LLM Robustness to Repeated Questions 研究重复提问对LLM阅读理解能力的影响，发现重复提问对性能提升不显著 large language model
19 Granite Guardian Granite Guardian：开源LLM安全保障模型，覆盖多维度风险检测。 large language model ✅
20 TRIM: Token Reduction and Inference Modeling for Cost-Effective Language Generation TRIM：通过Token缩减与推理建模，实现高性价比的语言生成。 large language model
21 FlexLLM: Exploring LLM Customization for Moving Target Defense on Black-Box LLMs Against Jailbreak Attacks FlexLLM：探索LLM定制化，针对黑盒LLM的越狱攻击实施移动目标防御 large language model
22 Active Inference for Self-Organizing Multi-LLM Systems: A Bayesian Thermodynamic Approach to Adaptation 提出主动推理框架以解决大语言模型适应性不足问题 large language model
23 DRUM: Learning Demonstration Retriever for Large MUlti-modal Models DRUM：学习演示检索器，提升大型多模态模型上下文学习能力 large language model
24 LLM-as-an-Interviewer: Beyond Static Testing Through Dynamic LLM Evaluation 提出LLM-as-an-Interviewer框架，用于动态评估大型语言模型 large language model ✅
25 CoPrUS: Consistency Preserving Utterance Synthesis towards more realistic benchmark dialogues CoPrUS：通过保持一致性的语句合成，生成更真实的基准对话数据集 large language model
26 Bilingual BSARD: Extending Statutory Article Retrieval to Dutch 提出bBSARD：扩展法规条文检索至荷兰语，并benchmark多种检索模型。 foundation model
27 Optimizing Alignment with Less: Leveraging Data Augmentation for Personalized Evaluation 利用数据增强优化对齐：提升个性化评估中LLM与人类偏好的一致性 large language model
28 Algorithmic Phase Transitions in Language Models: A Mechanistic Case Study of Arithmetic 揭示语言模型算术能力相变：Gemma-2-2b在不同位数加法中采用不同策略 large language model
29 Na'vi or Knave: Jailbreaking Language Models via Metaphorical Avatars 提出AVATAR框架，利用隐喻化身攻击大型语言模型，实现越狱并暴露安全风险。 large language model
30 My Words Imply Your Opinion: Reader Agent-based Propagation Enhancement for Personalized Implicit Emotion Analysis 提出RAPPIE模型，通过读者代理增强个性化隐式情感分析 large language model
31 Frame Representation Hypothesis: Multi-Token LLM Interpretability and Concept-Guided Text Generation 提出Frame Representation Hypothesis，用于多token LLM的可解释性和概念引导文本生成。 large language model ✅
32 Enhancing Relation Extraction via Supervised Rationale Verification and Feedback 提出基于监督式理由验证与反馈的关系抽取增强框架，提升LLM在RE任务上的性能。 large language model
33 HARP: Hesitation-Aware Reframing in Transformer Inference Pass HARP：Transformer推理中基于犹豫感知的重构方法，提升模型性能。 large language model
34 Label-Confidence-Aware Uncertainty Estimation in Natural Language Generation 提出标签置信度感知的不确定性估计方法，提升自然语言生成模型的可靠性。 large language model
35 KULTURE Bench: A Benchmark for Assessing Language Model in Korean Cultural Context 提出KULTURE Bench：一个评估语言模型在韩国文化背景下理解能力的基准 large language model
36 Multi-Response Preference Optimization with Augmented Ranking Dataset 提出基于增强排序数据集的多响应偏好优化方法，提升LLM性能 large language model
37 Exploring Coding Spot: Understanding Parametric Contributions to LLM Coding Performance 探索LLM代码能力：揭示参数对代码性能的贡献，发现“代码区域” large language model
38 Predictable Emergent Abilities of LLMs: Proxy Tasks Are All You Need 提出基于代理任务的大语言模型涌现能力预测方法 large language model

🔬 支柱一：机器人控制 (Robot Control) (2 篇)

#题目一句话要点标签🔗⭐
39 Defensive Dual Masking for Robust Adversarial Defense 提出防御性双重掩码（DDM）算法，增强NLP模型对抗文本对抗攻击的鲁棒性。 manipulation large language model
40 RAZOR: Sharpening Knowledge by Cutting Bias with Unsupervised Text Rewriting RAZOR：通过无监督文本重写消除偏差，提升知识泛化能力 manipulation

🔬 支柱二：RL算法与架构 (RL & Architecture) (1 篇)

#题目一句话要点标签🔗⭐
41 Filling Memory Gaps: Enhancing Continual Semantic Parsing via SQL Syntax Variance-Guided LLMs without Real Data Replay 提出LECSP方法，利用SQL语法变异引导LLM增强持续语义解析，无需真实数据回放。 distillation large language model

⬅️ 返回 cs.CL 首页 · 🏠 返回主页