ProgRouter: Online Progress-Guided Orchestration for Multi-Agent LLM Workflows
What it is
ProgRouter is a routing framework for multi-agent LLM workflows that adaptively selects the best LLM at each step based on dynamic task progress. It uses a multi-view progress scorer and dual-path predictor to balance task quality, time budgets, and operating costs in real-time.
The problem it solves
Existing cascade routing methods make static, one-shot decisions that fail to adapt to the dynamic, state-dependent nature of multi-step workflows, leading to unnecessary costs or quality degradation.
What you could use it for
- Reducing API costs in agentic code generation workflows
- Optimizing LLM usage in mathematical reasoning chains
- Managing latency and cost in retrieval-augmented long-form QA systems
Read the source ↗Back to the library →
Summary written by an AI model from the source; check the source before you rely on it.