gradient conflicts

5 papers

• Bigger, Regularized, Categorical: High-Capacity Value Functions are Efficient Multi-Task Learners
• Learning and Planning Multi-Agent Tasks via an MoE-based World Model
• Mixture-of-Experts Meets In-Context Reinforcement Learning
• PiKE: Adaptive Data Mixing for Large-Scale Multi-Task Learning Under Low Gradient Conflicts
• Searching Efficient Semantic Segmentation Architectures via Dynamic Path Selection