VIGOR Study - Vanilla - POPE A-OKVQA Adversarial - F1: leaderboard
VIGOR Study - Vanilla - POPE A-OKVQA Adversarial - F1: F1 (%). VIGOR v1 author-run vanilla baselines; original LLaDA-V and Lumina-DiMOO weights, greedy decoding at temperature zero, model-default block schedule. VCD, ICT and VIGOR intervention rows are not model baseline results. No inference speed metrics.
Metric: F1 (%). Source: arxiv.org. Saturation forecast: Rough model projection: around 2026. 2 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | LLaDA-V (VIGOR Run: vanilla; temperature=0; default-block-schedule) | 71.4 | |
| 2 | Lumina-DiMOO (VIGOR Run: vanilla; temperature=0; default-block-schedule) | 65.6 |
Interactive version: theaggregate.ai/benchmark?slug=vigor-study-vanilla-pope-a-okvqa-adversarial-f1 · How It Works · Data refreshed daily, snapshot 2026-11-06.