Research map / Vision-language and multimodal / Multimodal LLMs
BRAVE: Broadening the visual encoding of vision-language models
Oğuzhan Fatih Kar, Alessio Tonioni, Petra Poklukar, Achin Kulshrestha, Amir Zamir, Federico Tombari
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| combine five frozen vision encoders with MEQ-Former instead of one encoder | BRAVE · MMVP test | average pair accuracy: 27.3 (best single, EVA-CLIP-g) -> 42.0 | Improved |
| replace ensemble of Q-Formers with a single MEQ-Former (ablation) | BRAVE · COCO captioning (224 res) | CIDEr: 140.9 -> 145.2 | Improved |
| fine-tune the LM for VQA instead of frozen LM (ablation) | BRAVE · VQAv2 Karpathy val | accuracy: 78.6 -> 81.8 | Improved |
| LoRA rank 128 instead of full LM fine-tuning (ablation) | BRAVE · VQAv2 Karpathy val | accuracy: 81.8 -> 81.0 | Hurt |
| add synthetic VQA data (ablation) | BRAVE · OKVQA val | accuracy: 64.0 -> 65.7 | Improved |
| add encoder dropout during pre-training (ablation) | BRAVE · COCO captioning Karpathy val | CIDEr: 145.3 -> 147.0 | Improved |
| add encoder dropout during pre-training (ablation) | BRAVE · OKVQA val | accuracy: 66.0 -> 65.7 | Hurt |
| add high-resolution fine-tuning stage (ablation) | BRAVE · VQAv2 Karpathy val | accuracy: 79.6 -> 81.8 | Improved |
| use FlanT5-XL instead of FlanT5-L (ablation) | BRAVE · COCO captioning Karpathy val | CIDEr: 142.5 -> 147.0 | Improved |
| pre-train on WebLI instead of CC3M (ablation) | BRAVE · COCO captioning (224 res) | CIDEr: 138.3 -> 145.2 | Improved |
| scale single vision encoder from ViT-G (1.8B) to ViT-e (3.8B) | Q-Former VLM · VQAv2 Karpathy val | accuracy: 74.2 -> 75.6 | Improved |
| use OpenCLIP-G (larger, more data) instead of CLIP-L/14 | Q-Former VLM · COCO captioning Karpathy val | CIDEr: 133.0 -> 128.3 | Hurt |
Later papers that compared with this method
2 later accepted papers report BRAVE as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- Language-Instructed Vision Embeddings for Controllable and Generalizable Perception (ICLR 2026 · better on 1 of 1 shared result)
- EquiAV: Leveraging Equivariance for Audio-Visual Contrastive Learning (ICML 2024 · better on 1 of 2 shared results)
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 2 on most shared results.
- PaLI-17B — PaLI: A Jointly-Scaled Multilingual Language-Image Model (ICLR 2023) · better on 2 of 4 shared results
- SPHINX-2k — "SPHINX: A Mixer of Weights, Visual Embeddings and Image Scales for Multi-modal Large Language Models" (ECCV 2024) · better on 1 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Unveiling Encoder-Free Vision-Language Models (NeurIPS 2024 · similarity 0.91)
- Beyond Encoder Accumulation: Measuring Encoder Roles in Multi-Encoder VLMs (NeurIPS 2026 · similarity 0.90)
- Vision‑Language‑Vision Auto‑Encoder: Scalable Knowledge Distillation from Diffusion Models (NeurIPS 2025 · similarity 0.90)
- Synthesize Diagnose and Optimize: Towards Fine-Grained Vision-Language Understanding (CVPR 2024 · similarity 0.90)
- From Pixels to Words -- Towards Native Vision-Language Primitives at Scale (ICLR 2026 · similarity 0.90)
- VladVA: Discriminative Fine-tuning of LVLMs (CVPR 2025 · similarity 0.90)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
