---
title: "PyroDash：成本高效的Token级小-大语言模型协同推理框架"
source: "arXiv AI/CL/LG（论文）"
source_tier: "t2"
url: "https://arxiv.org/abs/2607.20327v1"
permalink: "https://feed.astralor.com/items/itm_95c5a09999cbcfc43335d2fc"
published: "2026-07-22T16:14:26.000Z"
score: 78
channel: "ai"
tags: ["PyroDash","SLM-LLM协同推理"]
---

# PyroDash：成本高效的Token级小-大语言模型协同推理框架

## AI 点评

> 这篇论文提出的PyroDash让我眼前一亮，它通过巧妙的控制令牌机制让SLM学会何时求助，从而大幅降低LLM调用成本。对于从事大模型推理部署和成本控制的工程师来说，这是一个值得深入研究的方案。尤其是它不需要修改LLM也不需要路由器，非常实用。

## 速览

PyroDash是一种成本感知的token级小模型（SLM）与大模型（LLM）协同推理框架。其核心思路是在SLM生成过程中，通过控制令牌决定是否向冻结的LLM发送部分推理轨迹以辅助完成。该框架无需独立路由器、LLM微调或logits访问，通过三阶段训练（控制令牌嵌入学习、卸载导向的有监督微调、基于GRPO的成本感知对齐）实现了准确性与推理成本的灵活平衡。在五个数学推理基准上，PyroDash能以极低的LLM调用次数（最低1.90% token比率）实现接近或超过纯LLM的准确率，同时大幅降低成本（如从49.36美元降至1.78美元）。

## 关键点

- 提出token级SLM-LLM协同推理，SLM通过控制令牌决定何时向LLM请求协助，无需独立路由器或LLM微调。
- 三阶段训练策略：控制令牌嵌入学习、卸载导向SFT、基于GRPO的成本感知对齐，平衡准确率与推理成本。
- 在五个数学推理基准上，PyroDash以1.90%的LLM token比率实现54.55%准确率，成本降低96.4%。

## 全文

Large language models (LLMs) provide strong reasoning capabilities but are expensive to serve at scale, whereas small language models (SLMs) are cheaper but less reliable on difficult problems. We introduce PyroDash, a cost-aware framework for token-level SLM-LLM collaborative inference. During generation, the SLM decides whether to request assistance by emitting a control token. A Collaborate Engine then sends the query and partial reasoning trace to a frozen LLM for completion through a single handoff. The policy is internalized in the SLM, requiring neither a separate router, LLM retraining, nor access to LLM logits. PyroDash trains the SLM in three stages: control-token embedding learning, offloading-oriented supervised fine-tuning, and cost-aware alignment with Group Relative Policy Optimization. Its reward balances answer accuracy against inference cost normalized by LLM-only inference. Across five mathematical reasoning benchmarks, PyroDash supports different accuracy-cost operating points. With $λ=0.05$, it achieves 64.04 percent average accuracy, 6.36 percentage points above the LLM-only baseline, while reducing cost by 20.4 percent. With $λ=0.6$, it achieves 54.55 percent accuracy with a 1.90 percent LLM token ratio and 0.012 LLM calls per example, reducing total cost from USD 49.36 to USD 1.78. These results show that learned token-level handoffs can reduce LLM use while preserving strong reasoning performance.

---

引用前请核对原文;本文由星际流动整理:https://feed.astralor.com/items/itm_95c5a09999cbcfc43335d2fc