Learning Self-Orchestration in Decentralized Agent Swarms
arXiv cs.CL · 16 Oct 2026, 03:51 AEST
A new framework called market-based post-training (MPT) enables decentralized agent swarms to learn both action policies and self-orchestration. This approach allows individual capabilities and participation patterns to adapt together without a central coordinating agent.
Key facts
- The capabilities of an agent swarm depend on both its individual members and how they coordinate.
- We introduce market-based post-training (MPT), a framework for jointly learning action policies and self-orchestration in decentralized agent swarms.
- Learning both action and bidding behavior allows individual capabilities and patterns of participation to adapt together without a central coordinating agent.
- Results show gains over monolithic post-training and confirm that agent participation indeed evolves with learning.
- Larger populations also suffer smaller performance losses under agent dropout, demonstrating that learned self-orchestration supports reuse, scaling, and robustness in decentralized agent swarms.
Read aloud
A new framework called market-based post-training MPT enables decentralized agent swarms to learn both action policies and self-orchestration. This approach allows individual capabilities and participation patterns to adapt together without a central coordinating agent. The capabilities of an agent swarm depend on both its individual members and how they coordinate. We introduce market-based post-training MPT, a framework for jointly learning action policies and self-orchestration in decentralized agent swarms. Learning both action and bidding behavior allows individual capabilities and patterns of participation to adapt together without a central coordinating agent. Results show gains over monolithic post-training and confirm that agent participation indeed evolves with learning. Larger populations also suffer smaller performance losses under agent dropout, demonstrating that learned self-orchestration supports reuse, scaling, and robustness in decentralized agent swarms.
Sources in this cluster
- Learning Self-Orchestration in Decentralized Agent Swarms
arXiv cs.CL · 16 Oct 2026, 03:51 AEST