Benchmark

VLMsAreBlind

A vision-language benchmark that probes blind spots and brittle reasoning in multimodal models.

Modality
multimodal
Categories
multimodal, reasoning, vision
Openness
unknown
Source
llm_stats
Reported scores
4

Reported scores

llm_stats

ModelOrganizationReported valueReportedEvidence
Qwen3.5-122B-A10BQwen0.9672026-02-24evidence
Qwen3.5-27BQwen0.9692026-02-24evidence
Qwen3.5-35B-A3BQwen0.972026-02-24evidence
Qwen3.6-27BQwen0.972026-04-21evidence

Scores are partitioned by the source that reported them and are never merged into a single cross-source ranking, because the sources measure different things and say so.