cs.AI（2025-09-20）

📊 共 9 篇论文 | 🔗 1 篇有代码

🎯 兴趣领域导航

#	题目	一句话要点	标签	🔗
1	FESTA: Functionally Equivalent Sampling for Trust Assessment of Multimodal LLMs	提出FESTA，通过功能等效采样评估多模态LLM的置信度	large language model multimodal
2	SalaMAnder: Shapley-based Mathematical Expression Attribution and Metric for Chain-of-Thought Reasoning	提出SalaMAnder，基于Shapley值评估CoT推理中数学表达式的贡献度，并优化提示构建。	large language model chain-of-thought
3	Zero-Shot Human Mobility Forecasting via Large Language Model with Hierarchical Reasoning	提出ZHMF框架，利用层级推理大语言模型实现零样本人类移动预测	large language model
4	Roundtable Policy: Improving Scientific Reasoning and Narratives through Confidence-Weighted Consensus of LLMs	提出Roundtable Policy，通过LLM置信度加权共识提升科学推理和叙事能力	large language model chain-of-thought
5	NUMINA: A Natural Understanding Benchmark for Multi-dimensional Intelligence and Numerical Reasoning Abilities	提出NUMINA基准，用于评估多模态LLM在3D室内场景中的数值推理能力	large language model multimodal	✅
6	ACCeLLiuM: Supervised Fine-Tuning for Automated OpenACC Pragma Generation	ACCeLLiuM：用于自动生成OpenACC编译指导语句的监督式微调方法	large language model
7	Design and Development of an Intelligent LLM-based LDAP Honeypot	提出基于LLM的智能LDAP蜜罐，提升网络安全防御的适应性和易用性	large language model

#	题目	一句话要点	标签	🔗	⭐
8	Sycophancy Mitigation Through Reinforcement Learning with Uncertainty-Aware Adaptive Reasoning Trajectories	提出SMART框架，通过不确定性自适应推理缓解大语言模型的谄媚问题	reinforcement learning large language model
9	AISTAT lab system for DCASE2025 Task6: Language-based audio retrieval	AISTAT实验室提出基于双编码器和对比学习的语音检索系统，用于DCASE2025 Task6。	contrastive learning distillation large language model