---
title: "AsymSpec：面向智能体LLM的上下文非对称推测解码"
source: "arXiv AI/CL/LG（论文）"
source_tier: "t2"
url: "https://arxiv.org/abs/2608.26004v1"
permalink: "https://feed.astralor.com/items/itm_be18dedbe13e4be53fe4f353"
published: "2026-08-26T16:50:02.000Z"
score: 72
channel: "ai"
tags: ["AsymSpec","δ融合","发散感知接受门","推测解码","智能体LLM"]
---

# AsymSpec：面向智能体LLM的上下文非对称推测解码

## AI 点评

> 这篇论文值得关注，因为它巧妙地将推测解码的上下文一致性假设打破，用非对称设计同时保住精度和速度。对做Agent推理优化和模型部署的团队来说，这个思路很可能成为压缩场景下的标准选项。

## 速览

论文提出AsymSpec，一种上下文非对称推测解码框架，用于解决智能体LLM推理中压缩上下文导致精度下降的问题。传统推测解码假设草稿模型和验证模型共享相同上下文，而AsymSpec让轻量草稿模型读取完整输入，大型验证模型使用压缩视图，通过对比δ融合和发散感知接受门保持稳定性。实验显示在四个能力基准和两个端到端基准上达到约90%全上下文精度，吞吐量提升1.3-1.7倍，计算成本降低至0.2-0.3倍。

## 关键点

- 打破传统推测解码的上下文一致性假设，草稿模型读取完整上下文，验证模型使用压缩上下文。
- 采用对比δ融合和发散感知接受门，在保持验证稳定性的同时提高草稿接受率。
- 实验平均达到约90%全上下文精度，吞吐量提升1.3-1.7倍，计算成本大幅降低。
- 在孤立文本能力和端到端Agent基准上均验证了方法有效性。

## 全文

Agentic LLM pipelines face escalating inference costs as context accumulates across retrieval, tool use, and multi-turn interactions. To control latency, deployments routinely compress inputs, but this degrades task accuracy. Speculative decoding (SD) accelerates generation losslessly, yet it assumes the drafter and verifier share an identical context, preventing SD from resolving the accuracy-overhead trade-off. We propose AsymSpec, an asymmetric speculative decoding framework that breaks this symmetry: a lightweight drafter reads the full input while the large verifier operates on the compressed view. The drafter steers the verifier via a contrastive $δ$-fusion of logits, modulated by a divergence-aware acceptance gate that preserves verification stability and high draft acceptance rates. Evaluated across four agentic capabilities and two end-to-end agent benchmarks, AsymSpec reaches $\approx 90\%$ of full-context accuracy on average, delivering $1.3$--$1.7\times$ throughput speedups at $0.2$--$0.3\times$ the compute cost on isolated text capabilities. These results show that asymmetric context access yields substantial gains precisely when compression discards critical reasoning signals.

---

引用前请核对原文;本文由星际流动整理:https://feed.astralor.com/items/itm_be18dedbe13e4be53fe4f353