Wikibench

CharadesSTA

multimodal

Charades-STA is a benchmark dataset for temporal activity localization via language queries, extending the Charades dataset with sentence temporal annotations. It contains 12,408 training and 3,720 testing segment-sentence pairs from videos with natural language descriptions and precise temporal boundaries for localizing activities based on language queries.

Leaderboard

Showing 12 of 12 results

  1. Qwen3 VL 235B A22B Instruct
    64.8%
    i
  2. Qwen3 VL 235B A22B Thinking
    63.5%
    i
  3. Qwen3 VL 30B A3B Instruct
    63.5%
    i
  4. Qwen3 VL 32B Thinking
    62.8%
    i
  5. Qwen3 VL 30B A3B Thinking
    62.7%
    i
  6. Qwen3 VL 32B Instruct
    61.2%
    i
  7. Qwen3 VL 8B Thinking
    59.9%
    i
  8. Qwen3 VL 4B Thinking
    59.0%
    i
  9. Qwen3 VL 8B Instruct
    56.0%
    i
  10. Qwen3 VL 4B Instruct
    55.5%
    i
  11. Qwen2.5 VL 32B Instruct
    54.2%
    i
  12. Qwen2.5 VL 7B Instruct
    43.6%
    i