---
title: "ProgRouter: Online Progress-Guided Orchestration for Multi-Agent LLM Workflows under Quality-Cost Tradeoffs"
source: "arXiv AI/CL/LG（论文）"
source_tier: "t2"
url: "https://arxiv.org/abs/2608.25992v1"
permalink: "https://feed.astralor.com/items/itm_e258b40d9736b6fc7af3877e"
published: "2026-08-26T16:42:02.000Z"
score: 71
channel: "ai"
tags: ["ProgRouter","Task Progress Scoring","Multi-Agent LLM Workflows","Cost Optimization","Cascade Routing"]
---

# ProgRouter: Online Progress-Guided Orchestration for Multi-Agent LLM Workflows under Quality-Cost Tradeoffs

## AI commentary

> What makes this paper worth attention is that it not only identifies the static limitations of cascade routing but also provides an actionable online step-wise routing solution. For engineering teams deploying multi-agent AI into production, ProgRouter offers a fresh path to control token costs without sacrificing task quality. I look forward to seeing its performance in real-world workflows once the code is open-sourced.

## At a glance

ProgRouter presents an online progress-guided routing framework that addresses the dynamic routing challenge in multi-agent LLM workflows under quality-cost tradeoffs. Existing cascade routing methods make one-shot, query-level decisions and cannot adapt to the state-dependent nature of multi-step workflows. ProgRouter introduces a multi-view task progress scorer that combines coarse workflow outcome regimes with fine-grained signals on subtask completion, progress trends, and workflow state quality. A dual-path task progress predictor and an adaptive meta-gating mechanism estimate the progress gain for each candidate routed LLM. Experiments on HumanEval Plus, MBPP, MATH-500, and ASQA—covering agentic code generation, mathematical reasoning, and retrieval-augmented long-form question answering—demonstrate that ProgRouter reduces operating costs relative to key baselines while maintaining strong task-solving performance.

## Key points

- ProgRouter introduces online progress-guided routing for quality-cost tradeoffs, overcoming the static decision paradigm of cascade routing.
- A multi-view progress scorer fuses subtask completion, progress trends, and workflow state quality for fine-grained progress awareness.
- A dual-path progress predictor with an adaptive meta-gating mechanism enables online progress-gain estimation for each candidate LLM.
- Experiments on HumanEval Plus, MBPP, MATH-500, and ASQA show significant cost reduction with maintained performance.

## Full text

Multi-agent large language model (LLM) workflows have emerged as a powerful paradigm for solving complex, open-ended tasks through collaborative reasoning among specialized LLM agents, but they incur substantial operating costs due to repeated LLM invocations and long-horizon context accumulation. Existing cascade routing methods make one-shot, query-level decisions and cannot adapt to the dynamic, state-dependent nature of multi-step workflows, in which the right LLM at each step depends on evolving task progress, remaining task difficulty, and cost-efficiency requirements. We present ProgRouter, an online progress-guided routing framework that adaptively selects LLM agents across workflow steps to preserve task-solving quality while adhering to time and cost budgets. ProgRouter introduces a multi-view task progress scorer that combines coarse workflow outcome regimes with fine-grained signals on subtask completion, progress trends, and workflow state quality. Then, a dual-path task progress predictor and an adaptive meta-gating mechanism estimate the progress gain for each candidate routed LLM. ProgRouter makes online step-wise routing decisions that balance progress gain, task time budgets, and long-term operating cost efficiency. Experiments on HumanEval Plus, MBPP, MATH-500, and ASQA, spanning agentic code generation, mathematical reasoning, and retrieval-augmented long-form question answering, demonstrate that ProgRouter reduces the operating cost relative to key baselines while maintaining strong task-solving performance.

---

Verify against the original before citing; curated by Astral Feed: https://feed.astralor.com/items/itm_e258b40d9736b6fc7af3877e