autoresearch - Autonomous ML Research Experiments
Run autonomous ML experiments that modify training code, train models, evaluate results, and iterate
Tags
Updated: 2026-06-30Capabilities
Typical Inputs
Typical Outputs
What this skill does
- read training code
- plan experiments
- modify training code
- commit code changes
- train models
- evaluate validation metrics
- keep or discard changes
- log experiment results
Inputs
- ML training code
- training repository
- evaluation harness
- experiment tag
Outputs
- modified training code
- trained model
- evaluation results
- results.tsv log
Requirements
- Python 3.10+
- uv package manager
- NVIDIA RTX GPU (8GB+ VRAM)
- CUDA via PyTorch