Skip to content
NewA calmer, faster ZeroHour

Deployed multi-agent system finds and repairs documentation drift

Original title: Mind the Gap: A Deployed Multi-Agent System for Documentation Maintenance

arXiv cs.AI / cs.LG / cs.CLby Lorenzo Masoero ()1 min read1
AI summary · x-ai/grok-4.7Summaries by Vani AI

In 43 shadow-mode days, TriAgent triaged 499 tickets and judges called 81.1% of verifiable assertions correct.

Authors deployed a multi-agent documentation system built around a triage agent named TriAgent. It extracts signals from resolved support tickets and code-documentation mismatches, then turns reviewed insights into documentation fixes. LLM judges from independent model families scored 245 of 302 assertions correct. After a repair, blinded judges preferred the agent's re-answer 71.4% of the time on 35 tickets.

Unconfirmed: Correctness and preference rates come from the authors' judge panel, not an independent replication.

Key facts
  • The panel preferred TriAgent on 70.5% of length-comparable pairs and humans on 23.2%.
  • Human and agent replies were often complementary rather than duplicates.
  • A blinded human audit reported strong alignment on self-improvement.
  • Each reviewed human insight becomes a documented gap and then a fix.
Explain this AI news with Vani AIWhat was announced, what is new and why it matters, in a new chat

From the original article

1 min read at arxiv.org

The "AI revolution" has enabled teams across the technology industry to deliver software features at unprecedented velocity. While code evolves faster than ever, customer-facing documentation often struggles to stay up to speed.

arXiv cs.AI / cs.LG / cs.CL

By Lorenzo Masoero · arxiv.org

Read the full article at arXiv cs.AI / cs.LG / cs.CL

A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.