Video-to-Text Retrieval on MSR-VTT (zero-shot)
50.9R@1InternVideo2-1B
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| InternVideo2-1BSetting=ZS2026.05 | 50.9 | — | — | |
| X-CLIPSetting=FT2026.05 | 48.9 | — | — | |
| Gemini Embedding 2Setting=ZS2026.05 | 48.3 | — | — | |
| PE-core familySetting=ZS2026.05 | 47.3 | — | — | |
| OmniRetriever-7BSetting=ZS2026.05 | 43.7 | — | — | |
| CLIP4ClipSetting=FT2026.05 | 42.7 | — | — | |
| LanguageBind (CLIP-H/14)Setting=ZS2026.05 | 40.9 | — | — | |
| UncompressedPruning Mode=/, Pruning Ratio=/, Backbone=BLIP, Zero-shot=true2026.04 | 35.8 | 59.7 | 733.4 | |
| CoMPPruning Mode=C, Pruning Ratio=0.7, Backbone=BLIP, Zero-shot=true2026.04 | 35.4 | 60.3 | 261.1 | |
| MADTPPruning Mode=T, Pruning Ratio=0.7, Backbone=BLIP, Zero-shot=true2026.04 | 34.8 | 59.8 | 265.3 | |
| Omni-Embed-NemotronSetting=ZS2026.05 | 30.6 | — | — | |
| SigLIP-2 familySetting=ZS2026.05 | 30.1 | — | — | |
| UPopPruning Mode=P, Pruning Ratio=0.7, Backbone=BLIP, Zero-shot=true2026.04 | 19.1 | 37.3 | 256.2 |