Data explorer

Every raw generation with its verdict, straight from the Parquet files on Hugging Face. Queries run in your browser with DuckDB-WASM over HTTP range requests: nothing is downloaded beyond the rows and columns your query touches, and there is no server. Pick a rollouts dataset, then a cell, filter, search inside the generated text, click a row for the full completion with its prompt.

Failure categories (DS-1000 t = 0 and a t = 0.8 subsample; LiveCodeBench t = 0 and t = 0.8) and bug themes (DS-1000 t = 0/0.8; LiveCodeBench t = 0.2) appear where that classification exists. Using these filters narrows to classified completions.
Custom SQL (advanced)

This box is prefilled with the SQL for your current dropdown selection (each Run refreshes it) — tweak it to go beyond the dropdowns. Write any DuckDB SQL; the short aliases cell (the selected rollouts partition), failures/executions, themes, and prompts are rewritten to the matching Parquet files (join on instance_id, sample; quote "primary" in themes). The prefill uses the full read_parquet(...) paths directly, which run as-is.

pick a cell and hit Run