POrchestra
POrchestra is a proactive agent orchestration framework that allows sub-agents to actively communicate with an orchestrator during execution to request toolset adjustments. It employs the OPOD method to provide process supervision signals for optimizing orchestration strategies, and is designed to address execution bottlenecks in long-horizon, complex tasks.
Timeline (last 90 days)
POrchestra stands for on-the-fly proactive agentic orchestration framework, which is an instantaneous proactive agent orchestration framework.
POrchestra allows sub-agents to proactively communicate with the orchestrator at any logical node during execution to request tool set adjustments, while maintaining continuity of local execution.
The researchers proposed the optimization method On-Policy Orchestration Distillation (OPOD).
POrchestra is evaluated on three long-horizon agent benchmarks covering three scenario types: deep research, dynamic digital assistant, and software engineering.
Compared to the strongest baseline, POrchestra improves average task success rate by 9.1%.
With OPOD, the success rate of small orchestrators improves by an average of 10.0% compared to existing training baselines.
POrchestra achieves a better cost-performance trade-off.
The research team has open-sourced the POrchestra code in the GitHub repository XueyangFeng/POrchestra.
All Facts (8)
POrchestra stands for on-the-fly proactive agentic orchestration framework, which is an instantaneous proactive agent orchestration framework.
50%UnverifiedPOrchestra allows sub-agents to proactively communicate with the orchestrator at any logical node during execution to request tool set adjustments, while maintaining continuity of local execution.
50%UnverifiedThe researchers proposed the optimization method On-Policy Orchestration Distillation (OPOD).
50%UnverifiedPOrchestra is evaluated on three long-horizon agent benchmarks covering three scenario types: deep research, dynamic digital assistant, and software engineering.
50%UnverifiedCompared to the strongest baseline, POrchestra improves average task success rate by 9.1%.
50%UnverifiedWith OPOD, the success rate of small orchestrators improves by an average of 10.0% compared to existing training baselines.
50%UnverifiedPOrchestra achieves a better cost-performance trade-off.
50%UnverifiedThe research team has open-sourced the POrchestra code in the GitHub repository XueyangFeng/POrchestra.
50%