Skip to content

MLToolBench: Learning Tool-Augmented Agents for Machine Learning Development

Sep 2026 · 0 citations · 37 references
Computer Science

TL;DR

This work introduces ToolMLBench, a suite of executable tools for data inspection, code verification, and experiment diagnosis, together with an SFT and RL pipeline for learning their use, and measures how privileged context changes the likelihood of a sampled tool action and uses this difference as a turn-level reward alongside the final outcome.

Abstract

Machine learning engineering (MLE) agents have made substantial progress, but learning through ML experimentation remains costly in time and computation. Synthetic environments reduce these costs while introducing variations in data and experimental settings that require task-specific diagnosis. Access to diagnostic tools alone does not ensure that agents learn when to use them or how to act on their findings. We introduce ToolMLBench, a suite of executable tools for data inspection, code verification, and experiment diagnosis, together with an SFT and RL pipeline for learning their use. Diagnostic calls acquire evidence whose value depends on subsequent decisions, so final outcomes provide limited guidance on which calls to reinforce. We address this challenge with SPICE, which measures how privileged context changes the likelihood of a sampled tool action and uses this difference as a turn-level reward alongside the final outcome. We train on 80 synthetic tasks and evaluate on 25 in-domain and 10 out-of-domain tasks. Providing tool interfaces and descriptions alone yields inconsistent gains across unadapted models. With the same diagnostic interface, our training pipeline raises in-domain success from 24.8% to 52.4% for Qwen3-8B and from 35.6% to 69.2% for Qwen3.5-35B-A3B. The latter also improves from 31% to 48% out-of-domain, supporting learned diagnostic tool use on held-out sources and targets.

View source

Similar papers

#artificial intelligence Preprint Oct 2026

Assembling Insights for Agentic Machine Learning Engineering Systems

This work systematically study how insights from related MLE tasks affect agentic MLE systems, and how their impact depends on insight type and injection method, and introduces MLE-InsightBench, a benchmark of 160 Kaggle competitions across 12 domains.

Bi-Hui Jin, Yin-Xi Li, Kai-Yuan Wang et al. · 0 citations
#machine learning Preprint Oct 2026

AgentTune: A Unified Training Lifecycle for Tool-Using LLM Agents

Tool-using agents increasingly require a common substrate for building, training, evaluating, distilling, and maintaining agentic systems, yet existing tooling typically addresses these stages in isolation. We present AgentTune, a source-available Python library that provides a unified framework for developing and opti...

R E Zera Marveen Lyngkhoi, Abhivansh Gupta, Vidushee Vats et al. · 0 citations
Preprint Aug 2026

DeltaML-Bench: Evaluating Machine Learning Agents on Real-World Research Repositories

DeltaML-Bench is introduced, a benchmark comprising 48 tasks sourced from research papers that require agents to improve published baselines within imperfect, open-source repositories that indicates that scaffolding design and integrity checks are important considerations when deploying agents for autonomous ML experim...

Josias Moukpe, Priyanka Aryal, M. Kenney · 3 citations · ⚡1
#artificial intelligence Preprint Sep 2026

TuiML: Machine Learning for AI Agents

TuiML is a self-contained machine-learning library built for AI agents, with native algorithms across supervised, unsupervised, time-series, data handling, tuning, and evaluation tasks, and Benchmarks show TuiML remains predictively competitive with scikit-learn and Weka.

Nilesh Verma, N. Lim, Albert Bifet et al. · 0 citations
Preprint Aug 2026

TraceML: What Auto-Research Agents Miss in Long-Horizon ML Development

Auto-research agents now run machine-learning development unattended for hours, revising data pipelines, models, and validation from their own feedback, yet on most competitions they still finish below strong human competitors. Outcome-based benchmarks record this gap but not its cause, because they grade the final sub...

J. Yan, Weiwei Sun, Si-Jie Li et al. · 0 citations

How Much of a Harness Does a Strong Agent Need for Autonomous ML Engineering?

Recent autonomous machine learning engineering (MLE) agents have made significant progress on public leaderboards. Often motivated by progress stagnation over long-horizon cycles and limited Large Language Model (LLM) primitives, modern MLE agents are deployed on top of increasingly elaborate machinery: multi-agent orc...

Kirill Brilliantov, Alejandro Hernández-Cano, Emmanuel Abbé · 0 citations

Related blog posts

We use cookies to run the site and, with your consent, for analytics and to show ads. See our Cookie Policy.