Model card
LLaVA-1.5.
UW-Madison / Microsoftopen-sourceUnknown paramsCLIP ViT-L + MLP projector + Vicuna-13B
Improved LLaVA with MLP connector and VQA data. 13B params. Strong open-source VLM baseline 2023-2024. Source: arxiv:2310.03744.
§ 01 · Benchmarks
Every benchmark LLaVA-1.5 has a recorded score for.
| # | Benchmark | Area · Task | Metric | Value | Rank | Date | Source |
|---|---|---|---|---|---|---|---|
| 01 | VQA v2.0 | Multimodal · Visual Question Answering | accuracy | 80.0% | #5 | 2023-10-05 | source ↗ |
| 02 | MMBench | Multimodal · Visual Question Answering | accuracy | 67.7% | #8 | 2023-10-05 | source ↗ |
| 03 | TextVQA | Multimodal · Visual Question Answering | accuracy | 61.3% | #8 | 2023-10-05 | source ↗ |
Rank column shows this model’s position vs all other models scored on the same benchmark + metric (competitors after the slash). #1 in red means current SOTA. Sorted by rank, then newest result.
§ 03 · Papers
1 paper with results for LLaVA-1.5.
- 2023-10-05· Multimodal· 3 results
Improved Baselines with Visual Instruction Tuning (LLaVA-1.5)
§ 05 · Sources & freshness
Where these numbers come from.
arxiv
3
results
3 of 3 rows marked verified.