Learning Self-Orchestration in Decentralized Agent Swarms
Market-based post-training lets decentralized agent swarms learn both actions and self-orchestration.
Market-based post-training (MPT) jointly learns action policies and bidding behavior so decentralized swarms decide who participates without a central coordinator. Bids allocate action and supply local training credit from collective outcomes. On embodied crafting, robot navigation, and deep research, MPT beat monolithic post-training, and learned orchestration outperformed other coordination strategies using the same action policies. Trained swarms adapted faster to related tasks, scaled better with larger populations, and lost less performance under agent dropout.
- Agents bid to decide participation and training credit
- No central coordinator is required
- Learned bids outperform fixed coordination with the same policies
- Larger swarms scale better and resist agent dropout
Full article208 words · extracted from huggingface.co · click to collapse
The capabilities of an agent swarm depend on both its individual members and how they coordinate. Yet existing multi-agent systems often train agents within fixed orchestration structures, separating the learning of individual capabilities from the organization of collective behavior. We introduce market-based post-training (MPT), a framework for jointly learning action policies and self-orchestration in decentralized agent swarms. Agents express their willingness to participate through bids, which determine who acts and how collective outcomes provide local training credit. Learning both action and bidding behavior allows individual capabilities and patterns of participation to adapt together without a central coordinating agent. We evaluate MPT on embodied crafting, robot navigation, and deep research. Results show gains over monolithic post-training and confirm that agent participation indeed evolves with learning. With the same post-trained action policies, learned self-orchestration outperforms alternative coordination strategies. Ablations show that both action and bidding benefit from learning, and that learned bids provide more effective action-training signals. Beyond the original training objectives, trained swarms adapt faster to related tasks and achieve better performance with larger populations at inference time. Larger populations also suffer smaller performance losses under agent dropout, demonstrating that learned self-orchestration supports reuse, scaling, and robustness in decentralized agent swarms. Code and trained models are available at https://github.com/Embodied-Minds-Lab/MPT.
Text extracted automatically; images, tables and formatting may be missing. Original: https://huggingface.co/papers/2610.22322