Back to resources
Research
Pix2Fact (2026)
An expert-level visual reasoning benchmark project designed to identify factual reasoning flaws in complex Visual Question Answering (VQA) models. The core findings are currently being submitted to ICML 2026 with GADE members as lead authors
An expert-level visual reasoning benchmark project designed to identify factual reasoning flaws in complex Visual Question Answering (VQA) models. The core findings are currently being submitted to ICML 2026 as lead authors.