Repository navigation
v1.0.6
What's Changed
- Add opt-in v2.1 text scoring while preserving legacy benchmarks by @hexapode in #146
- Add DeepSeek-V4.1-Flash parse-with-layout pipelines by @boyang-zhang1 in #147
- Categorize DeepSeek as open-weight and document it by @boyang-zhang1 in #152
- Fix layout element rules dropped when a doc also has parse rules by @AmirF194 in #133
- text-v2.1: fold page sections only when the test case opts in by @hexapode in #158
- fix(evaluation): guard rule alarms outside the main thread by @DivyamTalwar in #151
- Fix layout adapter default fallback shadowing the shape-matcher by @AmirF194 in #156
- Add anyformat parse provider, layout adapter and leaderboard row by @marta-borrajo in #157
- Add Jina-OCR-v1 parse provider, pipeline and leaderboard row by @boyang-zhang1 in #159
- fix: resolve some broken imports for diagrams by @logan-markewich in #161
- Add WeVisDoc parse providers and leaderboard results by @boyang-zhang1 in #163
- Add OvisOCR2 parse provider and leaderboard result by @boyang-zhang1 in #164
- Add TeleOCR parse provider and grounded leaderboard result by @boyang-zhang1 in #165
- Add HPD-Parsing provider and grounded leaderboard result by @boyang-zhang1 in #166
- Add HunyuanOCR 1.5 provider and complete benchmark result by @boyang-zhang1 in #167
- Add Cohere Parse v5 provider, pipeline by @violetd1 in #111
- Add Anthropic Opus 5.5 pipeline and leaderboard result by @boyang-zhang1 in #171
- Add OpenAI GPT-6 Sol and Luna pipelines and leaderboard results by @boyang-zhang1 in #172
- Add Nutrient DWS parse provider, pipelines and leaderboard row by @olihou in #162
- Preserve rotated geometry in layout and attribution evaluation by @SebasGarcia08 in #160
- Add GPT and Claude reasoning-effort sweeps with leaderboard rows by @boyang-zhang1 in #177
- Add GPT-5.6 Sol, GPT-5.6 Terra and GPT-6 Sol reasoning-effort rows by @boyang-zhang1 in #178
- Reprice GPT-5.6 Terra and Sol reasoning-none leaderboard rows by @boyang-zhang1 in #179
- Fix false LaTeX failures from superscript/subscript ordering by @hexapode in #180
- Read checkbox options written marker-first by @ajjimeno in #181
- fix(metrics): ignore arrow length when matching a LaTeX formula by @ajjimeno in #182
- LaTeX normalization: equate pure syntax, keep font wrappers as transcription fidelity by @ajjimeno in #168
- Use listed Anthropic prices for Claude parse cost by @boyang-zhang1 in #183
- Report cache read tokens in evaluation stats by @boyang-zhang1 in #184
- Add Claude Sonnet 5.5 parse pipeline and leaderboard row by @boyang-zhang1 in #185
- Add DocAI (ProvidusAI) parse provider by @SamiHK in #176
- Evaluate Databricks chart JSON returned in figure descriptions by @amruthasrivatsav in #196
- fix: properly handle HTML tables during layout evals by @logan-markewich in #199
- Add Haiku 5.5 and GPT-6.1 Sol Low parse pipelines by @boyang-zhang1 in #202
- Use direct REST for Databricks document parsing by @amruthasrivatsav in #198
- Add aeddix-alpine-ocr (KDL pipeline, open-weight VLM): provider, pipeline, docs, leaderboard entry by @harisyammnv in #201
- Add LightOnOCR-3 provider, pipeline, and update KDL-Frontier-Parser-nano scores by @baptiste-aubertin in #206
- Fix Databricks signature and chart attribution by @amruthasrivatsav in #205
- Add SOTA badge for ParseBench by @boyang-zhang1 in #208
- Fix layout attribution crash on rules without ro_index by @SebasGarcia08 in #212
- Add Nemotron 3.5 Parse pipeline and leaderboard row by @boyang-zhang1 in #214
- Fix canonical labels during stored layout extraction by @SebasGarcia08 in #190
- Add Upstage Document Parse 2 (Enhanced) and ParseBench results by @visionNoob in #203
- Correct Visual Grounding for rows scored without the attribution gate by @logan-markewich in #217
- Add GPT-6.1 Terra parse pipelines and leaderboard rows by @boyang-zhang1 in #221
- Add MiniCPM-OCR 1.0 pipeline and leaderboard row by @boyang-zhang1 in #223
- Add repetition_overshoot rule, --dataset , and 'parse-bench repetition make' by @hexapode in #224
- Add Unlimited-OCR 2 pipeline and leaderboard row by @boyang-zhang1 in #222
- Refresh LlamaParse leaderboard scores by @boyang-zhang1 in #226
- Add OvisOCR3 pipeline and leaderboard row by @boyang-zhang1 in #227
- Refresh Datalab leaderboard scores by @boyang-zhang1 in #229
- fix: isolate truncated table fragments during scoring by @hexapode in #189
New Contributors
- @hexapode made their first contribution in #146
- @AmirF194 made their first contribution in #133
- @DivyamTalwar made their first contribution in #151
- @marta-borrajo made their first contribution in #157
- @violetd1 made their first contribution in #111
- @olihou made their first contribution in #162
- @SamiHK made their first contribution in #176
- @amruthasrivatsav made their first contribution in #196
- @harisyammnv made their first contribution in #201
- @baptiste-aubertin made their first contribution in #206
- @visionNoob made their first contribution in #203
Full Changelog: v1.0.4...v1.0.6