Research map / Vision-language and multimodal / Visual question answering
Flamingo: a Visual Language Model for Few-Shot Learning
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| increase model size and provide 32 in-context examples | Flamingo · OKVQA | top1: 44.7 -> 51.0 | Helped |
| remove M3W interleaved image-text training data (ablation) | Flamingo base · Ablation validation benchmarks | Overall score: 70.7 -> 53.4 | Hurt |
| remove paired image-text training data (ablation) | Flamingo base · Ablation validation benchmarks | Overall score: 70.7 -> 60.9 | Hurt |
| remove video-text pairs (ablation) | Flamingo base · Video tasks | task scores: 70.7 -> 67.3 | Hurt |
| replace image-text pairs with LAION (ablation) | Flamingo base · Ablation validation benchmarks | Overall score: 70.7 -> 66.4 | Hurt |
| replace gradient accumulation with round-robin updates (ablation) | Flamingo base · Ablation validation benchmarks | Overall score: 70.7 -> 62.9 | Hurt |
| remove zero-initialized tanh gating (ablation) | Flamingo base · Ablation validation benchmarks | Overall score: 70.7 -> 66.5 | Hurt |
| replace gated xattn-dense with vanilla cross-attention (ablation) | Flamingo base · Ablation validation benchmarks | Overall score: 70.7 -> 66.9 | Hurt |
| replace gated xattn-dense with grafting architecture (ablation) | Flamingo base · Ablation validation benchmarks | Overall score: 70.7 -> 63.1 | Hurt |
| insert cross-attention every fourth layer rather than every layer (ablation) | Flamingo base · Ablation validation benchmarks | Overall score: 70.7 -> 68.8 | Hurt |
| replace Perceiver Resampler with an MLP (ablation) | Flamingo base · Ablation validation benchmarks | Overall score: 70.7 -> 66.6 | Hurt |
| replace Perceiver Resampler with a vanilla Transformer (ablation) | Flamingo base · Ablation validation benchmarks | Overall score: 70.7 -> 66.7 | Hurt |
Closest papers by meaning
- Prismatic VLMs: Investigating the Design Space of Visually-Conditioned Language Models (ICML 2024 · similarity 0.88)
- VLM2Vec: Training Vision-Language Models for Massive Multimodal Embedding Tasks (ICLR 2025 · similarity 0.88)
- VL-ICL Bench: The Devil in the Details of Multimodal In-Context Learning (ICLR 2025 · similarity 0.88)
- Audio Flamingo: A Novel Audio Language Model with Few-Shot Learning and Dialogue Abilities (ICML 2024 · similarity 0.87)
- Language Models as Black-Box Optimizers for Vision-Language Models (CVPR 2024 · similarity 0.87)
- Emu: Generative Pretraining in Multimodality (ICLR 2024 · similarity 0.87)
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.