---
title: "From Watching to Steering: Enabling Human-Coding Agent Collaboration through Execution Traces"
source: "arXiv AI/CL/LG（论文）"
source_tier: "t2"
url: "https://arxiv.org/abs/2611.04466v1"
permalink: "https://feed.astralor.com/items/itm_100457ae16aedb800428a278"
published: "2026-11-03T07:56:06.000Z"
score: 57
channel: "tech"
tags: ["LLM","AI Agent","From Watching","Steering","Enabling Human-Coding Agent","Collaboration","Execution Traces Recent","LLMs"]
---

# From Watching to Steering: Enabling Human-Coding Agent Collaboration through Execution Traces

## At a glance

Recent advances in large language models (LLMs) have enabled coding agents to autonomously perform complex software development tasks, yet their long-running execution remains difficult for users to inspect, understand,…

## Key points

- Recent advances in large language models (LLMs) have enabled coding agents to autonomously perform complex software development tasks, yet their long-running execution remains difficult for users to inspect, understand,…
- We present Vibetrace, an interactive trace visualization and intervention system for agentic programming built upon OpenCode.
- Vibetrace introduces a human-agent collaboration paradigm of Observe, Understand, Intervene, and Evolve.

## Full text

Recent advances in large language models (LLMs) have enabled coding agents to autonomously perform complex software development tasks, yet their long-running execution remains difficult for users to inspect, understand, and steer. We present Vibetrace, an interactive trace visualization and intervention system for agentic programming built upon OpenCode. Vibetrace introduces a human-agent collaboration paradigm of Observe, Understand, Intervene, and Evolve. It (1) visualizes agent workflows as interactive execution traces for inspecting actions, intermediate states, and failures; (2) enables users to fork execution from prior steps and provide targeted corrections without restarting validated progress; and (3) distills traces and human feedback into reusable skills for continual agent evolution. On SWE-bench Verified, distilled skills reduce input tokens by 23.0% and cost by 15.5%, while improving pass rate from 65.0% to 68.0%. User studies further demonstrate improved accuracy and efficiency over conventional log-based interaction.

---

Verify against the original before citing; curated by Astral Feed: https://feed.astralor.com/items/itm_100457ae16aedb800428a278