autoresearch

|

What it does

|Scan complete

An autonomous AI research agent that iteratively modifies LLM training code, runs experiments, evaluates validation loss (val_bpb), and keeps only improvements — fully automating the experiment loop without human intervention.