{"updated":"2026-11-02","kind":"AI実務ベンチ","conductor":"AInformation","license":"CC BY 4.0","runs":[{"bench_id":"b013-gazou","slug":"gazou-2026-11-03","title":"Reading numbers from images of receipts, charts, and tables (4 AIs, 6 tasks, 3 runs each)","field":"Reading images and photos","run_date":"2026-11-03","winner":"Claude","version":"v1","trials_valid":72,"url":"/research/bench/gazou-2026-11-03","data_url":"/data/research/bench/gazou-2026-11-03/latest.json"},{"bench_id":"b012-gijiroku","slug":"gijiroku-2026-10-29","title":"Extracting decisions and action items from meeting minutes (4 AIs, 6 tasks, 3 runs each)","field":"Meeting minutes and summaries","run_date":"2026-10-29","winner":"ChatGPT","version":"v1","trials_valid":72,"url":"/research/bench/gijiroku-2026-10-29","data_url":"/data/research/bench/gijiroku-2026-10-29/latest.json"},{"bench_id":"b011-gazou","slug":"gazou-2026-10-24","title":"Reading numbers from images of receipts, charts, and tables (4 AIs, 6 tasks, 3 runs each)","field":"Reading images and photos","run_date":"2026-10-24","winner":"ChatGPT","version":"v1","trials_valid":72,"url":"/research/bench/gazou-2026-10-24","data_url":"/data/research/bench/gazou-2026-10-24/latest.json"},{"bench_id":"b010-meibo","slug":"meibo-2026-10-19","title":"Standardize inconsistent spellings of client names (4 AIs, 6 tasks, 3 runs each)","field":"Data cleanup and name matching","run_date":"2026-10-19","winner":"ChatGPT","version":"v1","trials_valid":72,"url":"/research/bench/meibo-2026-10-19","data_url":"/data/research/bench/meibo-2026-10-19/latest.json"},{"bench_id":"b009-mail","slug":"mail-2026-10-14","title":"Writing an apology email for a delayed delivery (4 AIs, 6 tasks, 3 runs each)","field":"Email and business documents","run_date":"2026-10-14","winner":null,"version":"v1","trials_valid":72,"url":"/research/bench/mail-2026-10-14","data_url":"/data/research/bench/mail-2026-10-14/latest.json"},{"bench_id":"b008-suushiki","slug":"suushiki-2026-10-08","title":"Writing working spreadsheet formulas (4 AIs, 6 tasks, 3 runs each)","field":"Excel and spreadsheet formulas","run_date":"2026-10-08","winner":null,"version":"v1","trials_valid":72,"url":"/research/bench/suushiki-2026-10-08","data_url":"/data/research/bench/suushiki-2026-10-08/latest.json"},{"bench_id":"b007-kossi","slug":"kossi-2026-10-03","title":"Building a document outline with a set slide count and character count (4 AIs, 6 tasks, 3 runs each)","field":"Document outlining","run_date":"2026-10-03","winner":"ChatGPT","version":"v1","trials_valid":72,"url":"/research/bench/kossi-2026-10-03","data_url":"/data/research/bench/kossi-2026-10-03/latest.json"},{"bench_id":"b006-hyokeisan","slug":"hyokeisan-2026-10-02","title":"Calculating totals and growth rates from a sales table (4 AIs, 6 tasks, 3 runs each)","field":"Spreadsheets and aggregation","run_date":"2026-10-02","winner":null,"version":"v1","trials_valid":72,"url":"/research/bench/hyokeisan-2026-10-02","data_url":"/data/research/bench/hyokeisan-2026-10-02/latest.json"},{"bench_id":"b005-eibun","slug":"eibun-2026-09-30","title":"Translating an English press release (4 AIs, 6 tasks, 3 runs each)","field":"English text and translation","run_date":"2026-09-30","winner":"ChatGPT","version":"v3","trials_valid":72,"url":"/research/bench/eibun-2026-09-30","data_url":"/data/research/bench/eibun-2026-09-30/latest.json"},{"bench_id":"b004-shirabemono","slug":"shirabemono-2026-09-29","title":"Researching facts with a single correct answer (4 AIs, 6 tasks, 3 runs each)","field":"Research and fact-checking","run_date":"2026-09-29","winner":"ChatGPT","version":"v1","trials_valid":54,"url":"/research/bench/shirabemono-2026-09-29","data_url":"/data/research/bench/shirabemono-2026-09-29/latest.json"},{"bench_id":"b003-kessan","slug":"kessan-2026-09-28","title":"Pulling figures from an earnings report PDF (3 AIs, 6 tasks, 3 runs each)","field":"Reading PDFs and financial statements","run_date":"2026-09-28","winner":null,"version":"v1","trials_valid":45,"url":"/research/bench/kessan-2026-09-28","data_url":"/data/research/bench/kessan-2026-09-28/latest.json"},{"bench_id":"b002-gazou","slug":"gazou-2026-09-27","title":"Reading numbers from images of receipts, charts, and tables (4 AIs, 6 tasks, 3 runs each)","field":"Reading images and photos","run_date":"2026-09-27","winner":null,"version":"v1","trials_valid":37,"url":"/research/bench/gazou-2026-09-27","data_url":"/data/research/bench/gazou-2026-09-27/latest.json"},{"bench_id":"b001-gijiroku","slug":"gijiroku-2026-09-24","title":"Extracting decisions and action items from meeting minutes (4 AIs, 6 tasks, 3 runs each)","field":"Meeting minutes and summaries","run_date":"2026-09-24","winner":null,"version":"v2","trials_valid":72,"url":"/research/bench/gijiroku-2026-09-24","data_url":"/data/research/bench/gijiroku-2026-09-24/latest.json"}]}