saumia-bot's picture
Remove stale no-scores disclaimer from dataset card
218c745 verified
|
Raw History Blame Contribute Delete
4.93 kB
---
license: apache-2.0
viewer: false
pretty_name: ParseBench-2 results — URL extract pipelines
tags:
- document-parsing
- benchmark
- evaluation
- ocr
---
# ParseBench-2 results — URL extract pipelines
Per-document outputs of five URL-extract pipelines on the [ParseBench-2](https://huggingface.co/datasets/llamaindex/ParseBench-2) benchmark, published in the record format of [llamaindex/ParseBench-2-results](https://huggingface.co/datasets/llamaindex/ParseBench-2-results). Use these outputs to re-run the evaluation locally without calling any provider, inspect what a pipeline returned for a given page, or re-score when the ground truth or the scorers change.
The outputs were produced on 2026-11-12 with a You.com working copy of the [ParseBench](https://github.com/run-llama/ParseBench) harness (`parse_bench` 2.0.1, base commit `ea57ec6`).
## Layout
```
pipelines.jsonl one line per pipeline: name, display name, category, documents, data revision
results/<pipeline>/<subset>.jsonl one line per document, see schema/record.v1.schema.json
results/<pipeline>/failed.json documents that produced no output: {"<subset>": [{"id": ..., "error": ...}]}
data_manifest.json document ids of the data revision these outputs were produced against
schema/record.v1.schema.json record schema v1, verbatim copy from upstream
tools/ explode.py, pack.py, validate.py, verbatim copy from upstream
```
Each JSONL line is a minimal `InferenceResult` as the public evaluator loads it: `request.example_id` is `<subset>/<document name>`, the dataset's own id, `output` is the normalized parse output (markdown, pages, layout), and `raw_output` holds page and usage counters. Provider cost fields are intentionally not published. Twelve subsets are covered: `text`, `headings`, `strikeout`, `underline`, `tables`, `complex_table_headers`, `financial_tables`, `insurance_filing_tables`, `dense_tables`, `charts`, `forms`, `fillable_forms` (the `layout` subset was not run).
## Pipelines
| pipeline | display name | provider |
|---|---|---|
| `youcom_extract` | You.com Extract | youcom |
| `youcom_extract_lite` | You.com Extract (parse_effort: lite) | youcom |
| `tavily_extract` | Tavily Extract | tavily |
| `exa_contents` | Exa Contents | exa |
| `parallel_extract` | Parallel Extract | parallel |
Documents that failed provider-side produce no JSONL record; they are listed per subset in `results/<pipeline>/failed.json`.
## Re-score
From a checkout of [run-llama/ParseBench](https://github.com/run-llama/ParseBench) `main` (it has `--board v2`), with `HF_TOKEN` set while this repo is private. These pipelines are not `leaderboard-v2.csv` rows, so `scripts/leaderboard/rescore.py` gets a results map that points at this repo (`results_repo` is a top-level map key; there is no CLI flag for it):
```bash
cat > /tmp/youdotcom-results-v2.yaml <<'EOF'
results_repo: youdotcom/parsebench-2-results
data_repo: llamaindex/ParseBench-2
data_revision: a12c9785cd8cd015b5284e0ad0d74722091762f2
scored_with: null
rows:
- {provider: You.com Extract, status: verified, pipeline: youcom_extract, results_revision: e593510105fe44281564fadea48e626cb0afbd14}
- {provider: You.com Extract Lite, status: verified, pipeline: youcom_extract_lite, results_revision: e593510105fe44281564fadea48e626cb0afbd14}
- {provider: Tavily Extract, status: verified, pipeline: tavily_extract, results_revision: e593510105fe44281564fadea48e626cb0afbd14}
- {provider: Exa Contents, status: verified, pipeline: exa_contents, results_revision: e593510105fe44281564fadea48e626cb0afbd14}
- {provider: Parallel Extract, status: verified, pipeline: parallel_extract, results_revision: e593510105fe44281564fadea48e626cb0afbd14}
EOF
uv run python scripts/leaderboard/rescore.py run --board v2 --map /tmp/youdotcom-results-v2.yaml \
--pipelines youcom_extract_lite --out rescore-v2.json
```
Pass any comma-separated subset of the five pipelines to `--pipelines`. The first run downloads the ground truth into `.cache/leaderboard-gt/`. Scores land in `rescore-v2.json` under `results.<pipeline>`: `scores` (Text, Formatting, Tables, Charts, Visual_Grounding, Forms), `overall`, and `details.<column>` (published vs. ground-truth page counts per subset and any error). `layout` was not run, so `Visual_Grounding` is reported as an error and `overall` is null. Documents listed in `failed.json` have no record and score 0. `apply.py` is not needed, since it only updates leaderboard csv rows.
`tools/` is a verbatim copy of the same directory from [llamaindex/ParseBench-2-results](https://huggingface.co/datasets/llamaindex/ParseBench-2-results) (Apache-2.0). The re-score flow downloads `tools/explode.py` from the results repo at the pinned revision, so it must live in this repo.
## Citation
See the [ParseBench-2 dataset card](https://huggingface.co/datasets/llamaindex/ParseBench-2).