Saved in:
Bibliographic Details
Main Authors: Xavier, Daull, Bellot, Patrice, Bruno, Emmanuel, Martin, Vincent, Murisasco, Elisabeth
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2512.01434
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866915646965547008
author Xavier, Daull
Bellot, Patrice
Bruno, Emmanuel
Martin, Vincent
Murisasco, Elisabeth
author_facet Xavier, Daull
Bellot, Patrice
Bruno, Emmanuel
Martin, Vincent
Murisasco, Elisabeth
contents We introduce CollabToolBuilder, a flexible multiagent LLM framework with expert-in-the-loop (HITL) guidance that iteratively learns to create tools for a target goal, aligning with human intent and process, while minimizing time for task/domain adaptation effort and human feedback capture. The architecture generates and validates tools via four specialized agents (Coach, Coder, Critic, Capitalizer) using a reinforced dynamic prompt and systematic human feedback integration to reinforce each agent's role toward goals and constraints. This work is best viewed as a system-level integration and methodology combining multi-agent in-context learning, HITL controls, and reusable tool capitalization for complex iterative problems such as scientific document generation. We illustrate it with preliminary experiments (e.g., generating state-of-the-art research papers or patents given an abstract) and discuss its applicability to other iterative problem-solving.
format Preprint
id arxiv_https___arxiv_org_abs_2512_01434
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle A Flexible Multi-Agent LLM-Human Framework for Fast Human Validated Tool Building
Xavier, Daull
Bellot, Patrice
Bruno, Emmanuel
Martin, Vincent
Murisasco, Elisabeth
Artificial Intelligence
We introduce CollabToolBuilder, a flexible multiagent LLM framework with expert-in-the-loop (HITL) guidance that iteratively learns to create tools for a target goal, aligning with human intent and process, while minimizing time for task/domain adaptation effort and human feedback capture. The architecture generates and validates tools via four specialized agents (Coach, Coder, Critic, Capitalizer) using a reinforced dynamic prompt and systematic human feedback integration to reinforce each agent's role toward goals and constraints. This work is best viewed as a system-level integration and methodology combining multi-agent in-context learning, HITL controls, and reusable tool capitalization for complex iterative problems such as scientific document generation. We illustrate it with preliminary experiments (e.g., generating state-of-the-art research papers or patents given an abstract) and discuss its applicability to other iterative problem-solving.
title A Flexible Multi-Agent LLM-Human Framework for Fast Human Validated Tool Building
topic Artificial Intelligence
url https://arxiv.org/abs/2512.01434