VIABLE - Dual-Failure Diagnosis: leaderboard
VIABLE (Zhao et al., 2026) evaluates vision-language models as judges of answers to video questions from visually impaired users, drawn from outdoor walking, indoor scene and hand-object manipulation corpora. Here each answer carries two failures from different dimensions of a 12-type taxonomy, injected by Claude Sonnet 4.6, and a judgment counts only when the judge names exactly both. Full accuracy in percent over 122,023 answers; higher is better.
Metric: Full accuracy (%; 122,023 dual-injected answers, both injected failure codes predicted exactly, pooled over the WAD, VisAssist and VIA-EgoDex corpora). Source: arxiv.org. Saturation forecast: Around April 2028. 7 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5.4 | 20.6 | #121 |
| 2 | Claude Sonnet 4.6 | 15 | #158 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=viable-dual-failure-diagnosis · How It Works · Data refreshed daily, snapshot 2026-11-12.