Multimodal Question Answering on WebQA Text-grounded n=2,455 (test)
48.6F1 ScoreNemo-Emb-1B
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| Nemo-Emb-1Bk=5, Generator VLM=LLaVA-1.5-13B, #Tok=6292026.06 | 48.6 | 57.9 | 87.1 | |
| Qwen3-VL-Emb-8Bk=5, Generator VLM=LLaVA-1.5-13B, #Tok=6122026.06 | 47.8 | 57.2 | 87.2 | |
| Nemo-Emb-1Bk=2, Generator VLM=LLaVA-1.5-13B, #Tok=2332026.06 | 46.8 | 50.9 | 66.6 | |
| BM25 Retrievalk=5, Generator VLM=LLaVA-1.5-13B, #Tok=5522026.06 | 46.7 | 54.3 | 73 | |
| Qwen3-VL-Emb-8Bk=2, Generator VLM=LLaVA-1.5-13B, #Tok=2352026.06 | 46.1 | 50.2 | 66.4 | |
| BM25 Retrievalk=2, Generator VLM=LLaVA-1.5-13B, #Tok=2342026.06 | 43.9 | 48.4 | 51 | |
| Nemo-Emb-1Bk=1, Generator VLM=LLaVA-1.5-13B, #Tok=1302026.06 | 41.1 | 44 | 40.4 | |
| Qwen3-VL-Emb-8Bk=1, Generator VLM=LLaVA-1.5-13B, #Tok=1312026.06 | 40.8 | 43.6 | 40.7 | |
| BM25 Retrievalk=1, Generator VLM=LLaVA-1.5-13B, #Tok=1332026.06 | 39.9 | 42.7 | 31.8 | |
| Dense Retrievalk=2, Generator VLM=LLaVA-1.5-13B, #Tok=5382026.06 | 38.5 | 49.3 | 40.1 | |
| Dense Retrievalk=5, Generator VLM=LLaVA-1.5-13B, #Tok=13962026.06 | 37.6 | 59.5 | 62 | |
| Dense Retrievalk=1, Generator VLM=LLaVA-1.5-13B, #Tok=2642026.06 | 34.9 | 40.4 | 25.5 | |
| Latent Memoryk=5, Generator VLM=LLaVA-1.5-13B, #Tok=842026.06 | 30.7 | 34.3 | 70.5 | |
| Latent Memoryk=2, Generator VLM=LLaVA-1.5-13B, #Tok=542026.06 | 30 | 32.8 | 41.1 | |
| Latent Memoryk=1, Generator VLM=LLaVA-1.5-13B, #Tok=442026.06 | 28.8 | 30.8 | 24 | |
| Full ContextGenerator VLM=LLaVA-1.5-13B, #Tok=83712026.06 | 6 | 10 | — |