---
title: "Flamingo: a Visual Language Model for Few-Shot Learning"
description: "Flamingo connects a frozen vision encoder and a frozen language model with a fixed visual bottleneck and gated cross-attention layers. Read it to see how interleaved images and text share context, why the gates begin closed, and which celebrated few-shot claims remain hard to reproduce."
canonical_url: "https://fanout.sh/daily/2026-10-07-flamingo-visual-language-model"
md_url: "https://fanout.sh/daily/2026-10-07-flamingo-visual-language-model.md"
last_updated: "2026-10-07"
access: "public"
---

# Flamingo: a Visual Language Model for Few-Shot Learning

Flamingo connects a frozen vision encoder and a frozen language model with a fixed visual bottleneck and gated cross-attention layers. Read it to see how interleaved images and text share context, why the gates begin closed, and which celebrated few-shot claims remain hard to reproduce.

## Paper details

- Authors: Jean-Baptiste Alayrac et al.

- Venue: NeurIPS 2022

- Track: AI Research

- Difficulty: Medium

- Reading time: 42 min

- Original paper: https://arxiv.org/abs/2204.14198v2

## What you will learn

- How a Perceiver Resampler turns variable image and video features into sixty-four visual tokens.

- Why zero-initialized gated cross-attention preserves the language model before the visual bridge learns to open.

- How the single-image attention mask, broad task results, proprietary data, and unreleased model bound the evidence.

## Continue exploring

[Start with CLIP's contrastive pairing](https://fanout.sh/daily/2026-08-23-clip-natural-language-supervision): See how image-text matching produces a useful visual encoder before Flamingo turns visual evidence into generated language.

---
This representation contains public Fanout content only. Protected Pro lessons, account data, billing, checkout, and pricing are not included.

Browse the public content map: https://fanout.sh/sitemap.md