The Trinity:Capability-Driven Recursive Self-Improvement for Multimodal Agents
On this page
arXiv:2610.36077v1 Announce Type: new Abstract: Multimodal intelligence can be viewed as human-guided co-evolution between worlds and agents through interaction. However, scaling this process faces two bottlenecks: constructing heterogeneous, verifiable environments is costly, and manual Harness iteration cannot keep pace with growing task diversity and complexity. A key insight is that tasks with different modalities and surface forms can share the same underlying capability requirements. We therefore introduce The Trinity, a recursive self-improvement framework coupling world generation with Harness evolution through three roles: Creator, Explorer, and Evolver. Drawing on verified execution trajectories, Evolver shifts the basis of task organization from surface categories to shared requirements, reorganizing the original task tree into a hierarchical capability tree. We introduce Capability-Driven Real2Sim2Real: for heterogeneous tasks at the same node, Creator abstracts their common requirements into simplified, verifiable simulated worlds for capability learning, while evaluation on original tasks measures transfer. Along this capability tree, Explorer progressively evolves reusable capability modules through selective inheritance, forming a Harness tree whose validated updates accumulate across iterations within a fixed topology. The resulting Harnesses specialize in different capabilities, motivating a Capability-Aware Harness Router that uses public task information to select compatible module configurations from the tree. Our evolved Harnesses improve mean performance by 34 points across 28 training benchmarks and 26 points across 18 unseen benchmarks, including two real-world embodied tasks.
Why It Matters
The Trinity is a recursive self-improvement framework for multimodal agents that couples world generation with harness evolution, reorganizing task trees into a hierarchy based on shared capability requirements rather than surface categories. It addresses two scaling bottlenecks in agentic training—the cost of constructing heterogeneous, verifiable environments and the difficulty of manually iterating harnesses as task diversity grows. This matters because capability-driven, verifiable world generation could make agent training more sample-efficient and generalizable. It connects to agents and autonomy, multimodal generative capabilities, world models, and reinforcement learning and training.
Sources
Back to Recent News · Sub-article index.