Wikibench

RefSpatialBench

vision

Multilingual: no

RefSpatialBench evaluates spatial reference understanding and grounding.

Methodology

Imported from llm-stats public benchmark metadata. Modality: image. Max score: 100. Categories: grounding, spatial_reasoning, vision. Language: en. Verified by llm-stats: no.

Leaderboard

Showing 5 of 5 results

  1. Qwen3.6-27B self-reported llm-stats
    70.0%
  2. Qwen3 VL 235B A22B Thinking self-reported llm-stats
    69.9%
  3. Qwen3.5-122B-A10B self-reported llm-stats
    69.3%
  4. Qwen3.5-27B self-reported llm-stats
    67.7%
  5. Qwen3.5-35B-A3B self-reported llm-stats
    63.5%