List your company
ResearchFormerly WizWand

Spatiotemporal-sensitive video understanding on TimeBlind

77.2AccuracyGemini-3.1-Pro
55.67261.26166.8572.439May 21, 2026
Updated 5mo ago

Evaluation results

MethodLinks
Gemini-3.1-ProModel Type=Proprietary2026.0577.245.5
Qwen3-VL-8B + CRPOBackbone=Qwen3-VL-8B, Post-training Method=CRPO2026.0571.736
Qwen3-VL-8B + T-GRPOBackbone=Qwen3-VL-8B, Post-training Method=T-GRPO2026.0570.332.5
Qwen3-VL-4B + CRPOBackbone=Qwen3-VL-4B, Post-training Method=CRPO2026.0569.831.7
Qwen3-VL-8B + GRPOBackbone=Qwen3-VL-8B, Post-training Method=GRPO2026.0569.830.5
Qwen3-VL-8B + ArrowRLBackbone=Qwen3-VL-8B, Post-training Method=ArrowRL2026.0569.329.3
Qwen3-VL-4B + T-GRPOBackbone=Qwen3-VL-4B, Post-training Method=T-GRPO2026.0567.929
Qwen3-VL-4B + GRPOBackbone=Qwen3-VL-4B, Post-training Method=GRPO2026.0567.828
Qwen3-VL-8BBackbone=Qwen3-VL-8B, Post-training Method=base2026.0567.827.8
Qwen3-VL-4B + ArrowRLBackbone=Qwen3-VL-4B, Post-training Method=ArrowRL2026.0567.526.8
GPT-5.1Model Type=Proprietary2026.0567.327
Qwen3-VL-4BBackbone=Qwen3-VL-4B, Post-training Method=base2026.0566.226.5
Qwen2.5-VL-7BModel Type=Open-Source2026.056522.5
ArrowRL*Model Type=Open-Source2026.056520.2
InternVL3-8BModel Type=Open-Source2026.056318.3
LLaVA-OV-7BModel Type=Open-Source2026.0556.57.8