Intent Understanding on IntentQA
48.6LLM-Judge AccuracyOMNISAPIENS-7B RL
Evaluation results
| Method | Links | |
|---|---|---|
| OMNISAPIENS-7B RLMethod variant=RL2025.10 | 48.6 | |
| HumanOmniV2-7BModel size=7B2025.10 | 26.3 | |
| OMNISAPIENS-7B SFTMethod variant=SFT2025.10 | 25.6 | |
| Qwen 2.5-Omni-7BModel size=7B2025.10 | 25.4 | |
| Qwen-2.5-VL-7BModel size=7B2025.10 | 24.9 | |
| Gemma-3-4BModel size=4B2025.10 | 22.7 | |
| OMNISAPIENS-7B BAMMethod variant=BAM2025.10 | 17.7 |