Open-Ended Video Game Glitch Detection on VideoGlitchBench
34.55Description Generation PrecisionQwen2.5-VL-7B-Instruct + GliDe
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| Qwen2.5-VL-7B-Instruct + GliDeModel Type=Open-source, GliDe Enhancement=true2026.04 | 34.55 | 45.09 | 39.12 | 53 | 19.46 | |
| UI-TARS-1.5-7B + GliDeModel Type=Open-source, GliDe Enhancement=true2026.04 | 34.31 | 49.28 | 40.45 | 51 | 17.02 | |
| InternVL2.5-8B + GliDeModel Type=Open-source, GliDe Enhancement=true2026.04 | 32.64 | 40.97 | 36.33 | 50 | 17.04 | |
| Qwen2.5-VL-3B-Instruct + GliDeModel Type=Open-source, GliDe Enhancement=true2026.04 | 32.36 | 39.38 | 35.52 | 48 | 17.29 | |
| LLaVA-OneVision-7B + GliDeModel Type=Open-source, GliDe Enhancement=true2026.04 | 30.13 | 34.85 | 32.32 | 50 | 16.24 | |
| InternVL2.5-4B + GliDeModel Type=Open-source, GliDe Enhancement=true2026.04 | 29.15 | 36.87 | 32.56 | 52 | 15.27 | |
| claude-3.5-haikuModel Type=Proprietary2026.04 | 21.66 | 32.54 | 26.01 | 47 | 12.91 | |
| UI-TARS-1.5-7BModel Type=Open-source, GliDe Enhancement=false2026.04 | 19.08 | 24.93 | 21.62 | 40 | 9.11 | |
| gemini-2.0-flashModel Type=Proprietary2026.04 | 18.36 | 25.35 | 21.29 | 44 | 10.55 | |
| gpt-4o-miniModel Type=Proprietary2026.04 | 12.33 | 32.36 | 17.86 | 35 | 6.2 | |
| InternVL2.5-8BModel Type=Open-source, GliDe Enhancement=false2026.04 | 11.9 | 23.36 | 15.77 | 25 | 4.06 | |
| Qwen2.5-VL-3B-InstructModel Type=Open-source, GliDe Enhancement=false2026.04 | 11.08 | 26.53 | 15.63 | 31 | 4.81 | |
| Qwen2.5-VL-7B-InstructModel Type=Open-source, GliDe Enhancement=false2026.04 | 10.41 | 14.74 | 12.2 | 30 | 4.11 | |
| InternVL2.5-4BModel Type=Open-source, GliDe Enhancement=false2026.04 | 8.62 | 22.93 | 12.53 | 18 | 1.92 | |
| nova-lite-v1Model Type=Proprietary2026.04 | 6.66 | 23.77 | 10.41 | 28 | 2.98 | |
| LLaVA-OneVision-7BModel Type=Open-source, GliDe Enhancement=false2026.04 | 5.9 | 19.5 | 9.06 | 26 | 2.18 |