Proactive orchestration improves large language model agent success rates
Proactive Agentic Orchestration
Computation and Language
Summary
Sometimes, AI agents use smaller helper agents to handle big or long tasks, but these helpers usually have fixed goals that don't change even when the situation does. The authors propose a new way called POrchestra that lets these smaller agents ask to adjust their goals while still working, making them better at handling changing situations. They also introduce a learning method to improve how the main agent manages helpers by using feedback from these adjustment requests. This approach boosts performance on various complex tasks by about 9 to 10 percent compared to older methods.
What this means in practice
- •For digital assistant developers: Improve the adaptability and task success of AI assistants handling complex, changing user requests by allowing on-the-fly goal adjustment among sub-agents.
- •For software engineering teams: Enhance autonomous coding or testing systems by enabling sub-components to request real-time changes in their tasks, improving long-horizon project completions.
Authors
Xueyang Feng, Chaozhuo Li, Xiaohe Bo, Jiakai Tang, Chenxi Li, Qiwei Ye, Xu Chen
Abstract
Agentic orchestration delegates tasks to sub-agents with independent contexts, providing a key path to extending the effective execution horizon of large language model agents. However, existing orchestration methods face two limitations: in execution, current methods fix sub-agents' goals and harnesses at creation time, limiting their adaptability to dynamic real-world environments; in optimization, sparse and delayed feedback constrains effective learning of the orchestration policy. To address these limitations, we introduce POrchestra, an on-the-fly proactive agentic orchestration framework. It allows sub-agents to proactively communicate with the orchestrator at any logical point to request harness adjustments while maintaining continuity of local execution. This mechanism not only enables sub-agents to adapt seamlessly to dynamic environments but also naturally inspires On-Policy Orchestration Distillation (OPOD) based on hindsight feedback. OPOD attributes sub-agents' proactive revision requests back to preceding orchestration decisions to derive targeted improvement signals, and leverages these signals as process supervision alongside the final reward to optimize the orchestration policy. We evaluate POrchestra on three long-horizon agent benchmarks spanning deep research, dynamic digital assistance, and software engineering. POrchestra improves task success by an average of 9.1% relative to the strongest baseline; combined with OPOD, it improves small-orchestrator success by an average of 10.0% relative to existing training baselines. POrchestra also achieves a more favorable cost-performance trade-off. Our code is publicly available at https://github.com/XueyangFeng/POrchestra.