VIGOR Study - Vanilla - CHAIR Length 64 - CHAIRs: leaderboard

VIGOR Study - Vanilla - CHAIR Length 64 - CHAIRs: CHAIRs (%). 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. CHAIR uses a 500-image random COCO 2014 validation sample; the paper does not publish sampled image IDs.

Metric: CHAIRs (%). Source: arxiv.org. Saturation forecast: Rough model projection: around 2026. 2 models tracked.

Top models

#ModelScoreOverall rank
1LLaDA-V (VIGOR Run: vanilla; temperature=0; default-block-schedule)32.7
2Lumina-DiMOO (VIGOR Run: vanilla; temperature=0; default-block-schedule)33.3

Interactive version: theaggregate.ai/benchmark?slug=vigor-study-vanilla-chair-length-64-chairs · How It Works · Data refreshed daily, snapshot 2026-11-06.