{"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":"Leer cifras en imágenes de recibos, gráficos y tablas (4 IA, 6 tareas, 3 intentos cada una)","field":"Lectura de imágenes y fotografías","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":"Extraer decisiones y tareas pendientes de un acta de reunión (4 IA, 6 tareas, 3 intentos cada una)","field":"Actas de reunión y resúmenes","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":"Leer cifras en imágenes de recibos, gráficos y tablas (4 IA, 6 tareas, 3 intentos cada una)","field":"Lectura de imágenes y fotografías","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":"Unificar las variantes de escritura de los nombres de socios comerciales (4 IA, 6 tareas, 3 rondas cada una)","field":"Limpieza de datos y unificación de nombres","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":"Escribir un correo de disculpa por un retraso en la entrega (4 IA, 6 tareas, 3 intentos cada una)","field":"Correo electrónico y documentos de negocios","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":"Escribir fórmulas funcionales de hoja de cálculo (4 IA, 6 tareas, 3 intentos cada una)","field":"Fórmulas de Excel y hojas de cálculo","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":"Crear el esquema de un documento con un número fijo de páginas y caracteres (4 IA, 6 tareas, 3 intentos cada una)","field":"Creación de esquemas de documentos","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":"Calcular el total y la tasa de crecimiento a partir de una tabla de ventas (4 IA, 6 tareas, 3 intentos cada una)","field":"Hojas de cálculo y agregación de datos","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":"Traducir un comunicado de prensa en inglés (4 IA, 6 tareas, 3 intentos cada una)","field":"Textos en inglés y traducción","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":"Investigar hechos que tienen una única respuesta correcta (4 IA, 6 tareas, 3 intentos cada una)","field":"Investigación y verificación de hechos","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":"Extraer cifras de un PDF de resultados financieros (3 IA, 6 tareas, 3 intentos cada una)","field":"Lectura de PDF e informes financieros","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":"Leer cifras en imágenes de recibos, gráficos y tablas (4 IA, 6 tareas, 3 intentos cada una)","field":"Lectura de imágenes y fotografías","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":"Extraer decisiones y tareas pendientes de un acta de reunión (4 IA, 6 tareas, 3 intentos cada una)","field":"Actas de reunión y resúmenes","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"}]}