Search papers, labs, and topics across Lattice.
This study introduces TraceML, a comprehensive analysis of human-agent planning in machine learning development, revealing that while large language models can generate correct code for isolated tasks, they struggle significantly in autonomous development, often trailing behind human competitors. By analyzing 4,465 human trajectories across 134 Kaggle competitions and comparing them with agent performance, the research highlights that agents tend to operate within a limited feedback loop, failing to adapt and pivot as effectively as humans do. A brief planning prompt inspired by human practices improves agent performance but does not fully replicate the adaptive behaviors of human developers, indicating a persistent gap in autonomous learning capabilities.
Agents fall short in machine learning development, often locked in narrow loops while humans adapt and innovate across tasks.
Large language models write correct code for isolated problems but remain far weaker at autonomous machine-learning development, where an agent must revise data pipelines, models, and validation over hours of feedback, and on most competitions still finishes below strong human competitors. Outcome-based benchmarks record this gap but not its cause, because they grade the final submission and discard the development process behind it. We introduce TraceML, which pairs human and agent work on the same competitions under one version-level schema: 4,465 human Kaggle trajectories across 134 competitions, seven of which are also worked by two agent scaffolds, giving 430 paired human and 207 agent trajectories. Every code version carries its score, its timestamp, and labels for the action taken, its intent, the edit size, and the score effect. Read this way, the gap becomes concrete. Experts alternate data work, validation, model changes, and ensembling, and return to approaches they had set aside. Each agent scaffold instead collapses into a narrow loop: Codex spends its steps re-weighting ensembles and tuning submissions, MLEvolve mutates its model in place, and neither pivots at the human rate nor reopens abandoned work. A short planning prompt distilled from human practice moves the behaviors it names toward the human profile and lifts scores, but the effort profile stays agent-shaped: instruction closes only the part of the gap that reduces to instructions. We release the corpus, the schema, the labelers, and the extraction pipeline at https://huggingface.co/datasets/jerryyan/TraceML.
CMU ML