Sep 14 15:34 update github's link 15:18 Add database_clear as a denoised control environment Provide aligned, noise-removed counterparts of assets/database so evaluations can isolate how data friction affects measured model capability. May 16 15:34 Rename dataset branding from DataClaw to DataClawBench in READMEs May 26 07:05 Update README.zh-CN.md 07:04 Update README.md May 17 14:39 Update README.zh-CN.md 14:36 Update README.md May 6 06:19 Add Parquet preview exports and point HF YAML configs to themGenerate schema-stable Parquet copies for Hub Dataset Viewer and update README data_files to use the preview parquet files. 04:54 Use per-category JSON globs under assets/qa_gold/, default subset comprehensive_decision, drop database CSV configs from the card YAML. May 4 18:53 split HF dataset configs per CSV; default to internal_metrics 18:35 update dataset viewer configs 16:27 update dataset viewer configs 15:49 update DataClaw v0.1.0 image link Apr 30 07:14 update github link 04:14 Delete .ms_upload_cache Apr 29 12:34 fix dataset card metadata and add explicit data_files config 12:14 update README.md 11:56 track dataset CSV files with Git LFS 10:46 update README.md 09:31 Release v1.0: add dataset card and unify MIT license for Hugging Face 03:50 initial commit 
