List your company

Frame-level glitch detection on RefGlitch (test)

77AccuracyGPT-5
48.9256.2163.570.79Apr 13, 2026

Evaluation results

MethodLinks
GPT-5Setting=Ref2026.0477748764
Qwen3-VL-8BSetting=Ref2026.0473747276
GPT-5Setting=LastCleanFrame2026.0470628548
Qwen3-VL-8BSetting=LastCleanFrame2026.0470697069
Gemini 3 Flash PreviewSetting=IrRef2026.0468657260
Ministral-3-8BSetting=Ref2026.0468588345
Gemini 3 Flash PreviewSetting=NoRef2026.0467657061
GPT-5Setting=IrRef2026.0466518836
GPT-5Setting=NoRef2026.0465498834
Gemini 3 Flash PreviewSetting=Ref2026.0463735898
Ministral-3-8BSetting=LastCleanFrame2026.0463458630
Gemma-3-12BSetting=Ref2026.0463636363
Qwen3-VL-8BSetting=NoRef2026.0462428828
Gemma-3-12BSetting=LastCleanFrame2026.0461606258
Gemini 3 Flash PreviewSetting=LastCleanFrame2026.0459685686
Gemma-3-12BSetting=NoRef2026.0458486438
Ministral-3-8BSetting=NoRef2026.0457357323
Qwen3-VL-8BSetting=IrRef2026.0454198311
Gemma-3-12BSetting=IrRef2026.045114608
Ministral-3-8BSetting=IrRef2026.04503442