llamaindex/ParseBench-results

ParseBench results

Per-document inference outputs of every pipeline on the ParseBench leaderboard, so anyone can re-run the evaluation locally without calling any provider, inspect what a pipeline returned for a given page, and re-score when the ground truth changes.

Scores come from the ParseBench package. Pipeline results were produced by the same package or by the equivalent provider API calls; only the normalized output and the cost, usage and timing counters are published.

Layout

pipelines.jsonl                  one line per pipeline: name, display name, category, documents, data revision
results/<pipeline>/table.jsonl   one line per document, see schema/record.v1.schema.json
results/<pipeline>/chart.jsonl
results/<pipeline>/text.jsonl    scored as both text_content and text_formatting
results/<pipeline>/layout.jsonl
data_manifest.json               document ids of the data revision these results were scored on
tools/                           explode.py, pack.py, validate.py (Python 3.9+, standard library)

Each line is a minimal InferenceResult as the evaluator loads it: request.example_id is <category>/<document stem>, output is the normalized parse output (markdown, pages, layout), and raw_output holds cost, page and token counters. For layout records of providers whose layout adapter reads the provider payload, raw_output also carries that payload with harness fields removed.

Two tiers: eval records and full results

• results/<pipeline>/<category>.jsonl is the small tier: only the fields the evaluator needs. Download this to re-score a pipeline quickly.
• raw/<pipeline>/<category>.jsonl is the full tier: one line per document with everything a user of that pipeline receives, meaning the complete provider response (raw_output), the normalized output with all citation metadata, timing and cost. Fields that exist only because of our harness (internal paths, run ids, job ids, signed download links) are removed.

Download one tier at a time with --include "results/<pipeline>/*", --include "raw/<pipeline>/*" or --include "traces/<pipeline>/*".

Replay a pipeline locally

pip install -U huggingface_hub
hf download llamaindex/ParseBench-results --repo-type dataset --local-dir pb-results \
  --include "results/datalab_accurate/*" "tools/*" "data_manifest.json" "pipelines.jsonl"

git clone https://github.com/run-llama/ParseBench && cd ParseBench && uv sync
uv run parse-bench data download                       # ground truth into ./data
python ../pb-results/tools/explode.py --benchmark parsebench ../pb-results/results/datalab_accurate output/datalab_accurate
uv run parse-bench run datalab_accurate --skip_inference

The run reads output/datalab_accurate/<category>/<stem>.result.json, scores every document, and writes the reports next to them. Pass --group table to score one category.

Submit your pipeline

1. Register the pipeline in the ParseBench package and run it: uv run parse-bench run <pipeline>.
2. Pack and check:

   python tools/pack.py --benchmark parsebench --pipeline <pipeline> output/<pipeline> results/<pipeline>
   python tools/validate.py --benchmark parsebench results/<pipeline>
   

   validate.py fails on anything that is not a public document id, on local paths or hostnames, on credentials, and on non-canonical JSON. Identical model output always packs to identical files; only timestamps, latency and cost differ between runs.
3. Open a pull request on this dataset:

   hf upload llamaindex/ParseBench-results results/<pipeline> results/<pipeline> --repo-type dataset --create-pr
   

   and link it from your leaderboard pull request on GitHub. The leaderboard row is re-scored from the uploaded results before merge.

Citation

See the ParseBench dataset card.