Open each sample to compare original.pdf, parsed_text.md, cleaned_body.md, raw_docling.md, and low-overlap blocks.
| sample | sections | abstract_chars | clean_body_chars | candidate_pdf_blocks | avg_pdf_block_coverage | min_pdf_block_coverage | low_coverage_blocks | high_risk_body_missing_blocks | front_summary_missing_blocks | back_matter_low_overlap_blocks | clean_noise_hits | sample |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 01-jechem-2023-defect-engineering | 9 | 1726 | 58558 | 53 | 0.831 | 0.0 | 2 | 0 | 0 | 0 | 0 | open |
| 02-jechem-2021-review | 14 | 988 | 50336 | 25 | 0.801 | 0.0 | 3 | 0 | 0 | 1 | 0 | open |
| 03-jpowsour-2015-severe-truncation | 8 | 818 | 21981 | 17 | 0.923 | 0.823 | 0 | 0 | 0 | 0 | 0 | open |
| 04-jpowsour-2016-multi-issue | 3 | 1182 | 4140 | 17 | 0.259 | 0.0 | 13 | 12 | 0 | 1 | 0 | open |
| 05-acsenergylett-2019-low-ratio | 2 | 897 | 2105 | 40 | 0.018 | 0.0 | 39 | 36 | 2 | 1 | 0 | open |
| 06-nature-2022-early-stop | 14 | 1427 | 33838 | 56 | 0.561 | 0.0 | 22 | 17 | 0 | 2 | 0 | open |