约束生成策略优化（CGPO）框架解决基于人类反馈强化学习（RLHF）在多任务学习（MTL）中的局限性
论文The Perfect Blend: Redefining RLHF with Mixture of Ju...
Read MoreCoT（Chain of Thought）在数学和符号推理任务中表现突出
论文To CoT or not to CoT? Chain-of-thought helps mainly o...
Read More
识别和评估大型语言模型（LLM）在医疗健康领域的潜在偏见和健康不公平
论文A toolbox for surfacing health equity harms and biase...
Read More
SciAgents：通过多代理系统（Multi-Agent System）和知识图谱（Knowledge Graphs），实现科学发现的自动化
论文SciAgents: Automating scientific discovery through mu...
Read More大模型（LLMs）扮演宏观经济学家（Macroeconomists）
论文Large Language Models as Macroeconomists详细探讨了如何利用大型语言...
Read More物理信息神经算子（Physics-Informed Neural Operator，PINO）有效提升混沌系统模拟的效率和精度
论文Beyond Closure Models: Learning Chaotic Systems via P...
Read More
DemoStart：新强化学习方法，通过少量模拟演示和稀疏奖励，让三指机械手的器人学习复杂操作行为
论文DemoStart: Demonstration-led auto-curriculum applied...
Read MoreAgentTorch：扩展基于代理的模型的新型框架，通过高效的计算方式可实现对百万级别代理的模拟
论文On the Limits of Agency in Agent-Based Models（《论基于代理的...
Read More
MaskSR2：基于MaskSR，结合语义知识蒸馏和声学语言建模的全频带语音恢复生成框架
论文Joint Semantic Knowledge Distillation and Masked Acou...
Read More
机器人效用模型（Robot Utility Models, RUM）实现零样本部署（Zero-Shot Deployment）
论文《Robot Utility Models: General Policies for Zero-Shot...
Read More