# 微软：用蒸馏技能替代昂贵测试时推理，一次付清推理成本

- 来源：Rohan Paul (@rohanpaul_ai)
- 发布时间：2026-08-15 23:58
- AIHOT 分数：36
- AIHOT 链接：https://aihot.virxact.com/items/cmsul4tpt02f4rouhx2xklhux
- 原文链接：https://x.com/rohanpaul_ai/status/2088656461274517888

## AI 摘要

微软新论文提出，可将昂贵的测试时推理替换为从过往智能体运行中提取的小型规则集。用 GPT-5.4-mini 提取 35–50 条轨迹中的失败模式并转为 markdown 技能，在 4 个智能体基准上恢复了非推理与推理模式间 55%–100%+ 的差距，且输出 token 减少 2.9–4.5 倍。

## 正文

What if you could pay the reasoning cost once, then reuse what the model learned across future tasks?

New Microsoft paper finds that some expensive test-time reasoning can be replaced with a small set of rules learned from previous agent runs.

The paper tests a cheaper alternative: collect 35-50 past trajectories, have a coding agent extract recurring failure patterns, then turn those patterns into a small markdown skill added to the non-reasoning model's system prompt.

For GPT-5.4-mini, those skills recovered 55%-100%+ of the gap between non-reasoning and reasoning modes across 4 agent benchmarks, while using 2.9-4.5× fewer output tokens than reasoning.

On ALFWorld and τ2-retail, the skilled non-reasoning model actually beat the reasoning mode.

The useful part is that the distiller did not need reasoning traces: skills built only from cheap non-reasoning rollouts were competitive across all 4 domains.

The limit is equally useful: reasoning still won on telecom and SpreadsheetBench, where each task contains more instance-specific dependencies that a fixed skill cannot capture.

So the practical split is: distill repeated procedures once, then reserve expensive test-time reasoning for the tasks that genuinely need fresh search.

- arxiv. org/abs/2608.07885

Title: "Reason Wide, Not Deep: Amortizing the Reasoning Premium into Distilled Skills"