The Trinity：面向多模态智能体的能力驱动递归自我改进
原标题：Paper page - The Trinity:Capability-Driven Recursive Self-Improvement for Multimodal Agents
AI 摘要
该论文提出 The Trinity 框架，通过将任务组织为能力树、生成可验证的模拟训练世界，并路由专用 Harness 模块，实现多模态智能体的递归自我改进。框架包含 Creator、Explorer、Evolver 三个角色，将任务组织从表面类别转为共享能力需求。实验显示，进化后的 Harness 在 28 个训练基准上平均提升 34 分，在 18 个未见基准上平均提升 26 分，包括两个真实世界具身任务。
正文节选
The Trinity:Capability-Driven Recursive Self-Improvement for Multimodal Agents Abstract The Trinity framework recursively improves multimodal agents by organizing tasks into a capability tree, generating simulated training worlds, and routing specialized harness modules to boost performance on diverse benchmarks. Multimodal intelligence can be viewed as human-guided co-evolution between worlds and agents through interaction. However, scaling this process faces two bottlenecks: constructing heter