---
title: "深度交互：用于大型推理模型的高效人机交互方法"
source: "arXiv AI/CL/LG（论文）"
source_tier: "t2"
url: "https://arxiv.org/abs/2607.14049v1"
permalink: "https://feed.astralor.com/items/itm_d4d6982a4bb84bd1dc0af415"
published: "2026-07-15T17:16:43.000Z"
score: 75
channel: "ai"
tags: ["Chain-of-Thought","Deep Interaction","人机交互","大型推理模型","错误修正"]
---

# 深度交互：用于大型推理模型的高效人机交互方法

## AI 点评

> 这篇工作精准切中了当前LLM交互中的一个痛点——错误修正的低效。对于需要高准确率推理的开发者或研究者而言，Deep Interaction提供了一种直接、节省token的纠错方案，尤其值得关注。虽然还在预印本阶段，但方法简洁有效，有潜力被集成到后续工具中。

## 速览

本文提出Deep Interaction方法，允许用户直接编辑大语言模型的CoT推理过程中的错误步骤，同时保留正确部分。通过将编辑后的推理路径精炼为蒸馏提示，引导模型沿修正路径生成。在STEM任务上，该方法比基线方法纠正成功率提高25%以上，token使用减少约40%。

## 关键点

- 提出直接编辑模型原始输出的交互方式，避免重新生成或繁琐指错。
- 将编辑后的CoT精炼为蒸馏提示，引导模型正确推理。
- STEM任务纠正成功率提升25%以上，token使用减少40%。

## 全文

The emergence of Chain-of-Thought (CoT) reasoning has significantly enhanced the ability of large language models (LLMs) to tackle complex, multi-step tasks. However, when errors occur, current interaction approaches typically involve re-generating another response that may make mistakes again, or users laboriously flag the faulty step in follow-up turns that may get responses <You are right, I made a mistake here> followed by similar errors recurring. To address this issue, we propose an efficient human intervention mechanism for precisely correcting reasoning errors in LLMs, termed Deep Interaction. Our approach enables direct editing of the original response, allowing erroneous parts to be corrected while preserving accurate reasoning steps. We refine the edited CoT into a distilled prompt, which then steers the LLM along the corrected reasoning path. Experimental results show that our method achieves over a 25% improvement in correction success rate and reduces token usage by approximately 40% on STEM tasks reasoning compared to baseline approaches.

---

引用前请核对原文;本文由星际流动整理:https://feed.astralor.com/items/itm_d4d6982a4bb84bd1dc0af415