build1 distinct publisher
A blind model scores on your vision benchmark, which means the benchmark grades priors
MMEvalPro reports the best text-only LLM trailing the best multimodal model by only 14.64% on three popular benchmarks. That is a smaller gap than the spread among the vision models themselves.
Publishers:arxiv.org
Reality
- Evidence52
- Adoption
- Insufficient
- Hype gap+18
- Incentives68
- Confidence47