BIG-Bench Hard (BBH) Corrected Dataset v2.0
Abstract
BBH Corrected Dataset - 99 verified errors fixed across 9 categories. KYVER Framework (Kyung Yeosun Verification Framework): - KYVER-Math: Mathematical SVG parsing (geometric_shapes: 49) - KYVER-AD: Adversarial Debate System (other 8 categories: 50) Corrections: - geometric_shapes: 49 - salient_translation_error_detection: 21 - movie_recommendation: 10 - disambiguation_qa: 5 - ruin_names: 5 - causal_judgement: 4 - date_understanding: 2 - snarks: 2 - sports_understanding: 1
Publication details
- Journal / source
- Zenodo (CERN European Organization for Nuclear Research)
- Publication date
- 2 January 2026
- Document type
- Dataset