List your company

Procedural Temporal Understanding on LEGO-Puzzles (test)

42.82Overall AccuracyTPRU-7B
26.64830.846535.04539.2435Feb 21, 2026

Evaluation results

MethodLinks
TPRU-7BBackbone=7B2026.0242.823,200
LLaVA-Video-Qwen2-72BBackbone=Qwen2-72B2026.0241.64500
Qwen2.5-Omni-7BBackbone=Qwen2.5-7B2026.0236.451,200
Long-VITA-16KContext Window=16K2026.0234.45200
LLaVA-Video-Qwen2-7BBackbone=Qwen2-7B2026.0231.91100
SmolVLM2-256M-VideoParameters=256M2026.0228.180
SmolVLM2-500M-VideoParameters=500M2026.0227.270