List your company

Knowledge-Intensive Visual Question Answering on Museum Artifact Dataset V+K 1.0 (test)

69Answer AccuracyQwen3-VL-8B (Direct)
52.3656.686165.32Jul 7, 2026

Evaluation results

MethodLinks
Qwen3-VL-8B (Direct)search=disabled, reliability control=none2026.076929703.46
Qwen3-VL-8B + Searchsearch=enabled, reliability control=none2026.075944552.91
VaseAgent (Full)search=enabled, reliability control=complete system, GRPO selection=enabled2026.075422733.75
VaseAgent (w/o GRPO)search=enabled, reliability control=source/response control, GRPO selection=disabled2026.075338603.04