# Beyond Synthetic Replays: Turning Diffusion Features into Few-Shot Class-Incremental Learning Knowledge

arXiv: https://arxiv.org/abs/2503.23402 · HTML: https://mlsanity.com/paper/2503.23402/

## Key facts

- Cited by 1 papers in the map (data through 14 Oct 2026).
- Builds on 35 of its 47 references that are in the map.
- First posted on arXiv 30 Mar 2025.

## Builds on (most cited first, 30 of 35)

- [Deep Residual Learning for Image Recognition](https://mlsanity.com/paper/1512.03385/) (2015)
- [Learning Transferable Visual Models From Natural Language Supervision](https://mlsanity.com/paper/2103.00020/) (2021)
- [An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale](https://mlsanity.com/paper/2010.11929/) (2020)
- [U-Net: Convolutional Networks for Biomedical Image Segmentation](https://mlsanity.com/paper/1505.04597/) (2015)
- [High-Resolution Image Synthesis with Latent Diffusion Models](https://mlsanity.com/paper/2112.10752/) (2021)
- [Denoising Diffusion Implicit Models](https://mlsanity.com/paper/2010.02502/) (2020)
- [ImageNet Large Scale Visual Recognition Challenge](https://mlsanity.com/paper/1409.0575/) (2014)
- [DINOv2: Learning Robust Visual Features without Supervision](https://mlsanity.com/paper/2304.07193/) (2023)
- [Classifier-Free Diffusion Guidance](https://mlsanity.com/paper/2207.12598/) (2022)
- [Taming Transformers for High-Resolution Image Synthesis](https://mlsanity.com/paper/2012.09841/) (2020)
- [CutMix: Regularization Strategy to Train Strong Classifiers with Localizable Features](https://mlsanity.com/paper/1905.04899/) (2019)
- [An Image is Worth One Word: Personalizing Text-to-Image Generation using Textual Inversion](https://mlsanity.com/paper/2208.01618/) (2022)
- [Learning to Prompt for Continual Learning](https://mlsanity.com/paper/2112.08654/) (2021)
- [GLIGEN: Open-Set Grounded Text-to-Image Generation](https://mlsanity.com/paper/2301.07093/) (2023)
- [Few-Shot Class-Incremental Learning](https://mlsanity.com/paper/2004.10956/) (2020)
- [Diffusion Hyperfeatures: Searching Through Time and Space for Semantic Correspondence](https://mlsanity.com/paper/2305.14334/) (2023)
- [Few-Shot Incremental Learning with Continually Evolved Classifiers](https://mlsanity.com/paper/2104.03047/) (2021)
- [Forward Compatible Few-Shot Class-Incremental Learning](https://mlsanity.com/paper/2203.06953/) (2022)
- [Neural Collapse Inspired Feature-Classifier Alignment for Few-Shot Class Incremental Learning](https://mlsanity.com/paper/2302.03004/) (2023)
- [Few-Shot Class-Incremental Learning from an Open-Set Perspective](https://mlsanity.com/paper/2208.00147/) (2022)
- [Few-Shot Class-Incremental Learning via Entropy-Regularized Data-Free Replay](https://mlsanity.com/paper/2207.11213/) (2022)
- [Subspace Regularizers for Few-Shot Class Incremental Learning](https://mlsanity.com/paper/2110.07059/) (2021)
- [Learning with Fantasy: Semantic-Aware Virtual Contrastive Constraint for Few-Shot Class-Incremental Learning](https://mlsanity.com/paper/2304.00426/) (2023)
- [SDDGR: Stable Diffusion-based Deep Generative Replay for Class Incremental Object Detection](https://mlsanity.com/paper/2402.17323/) (2024)
- [Text-image Alignment for Diffusion-based Perception](https://mlsanity.com/paper/2310.00031/) (2023)
- [Pre-trained Vision and Language Transformers Are Few-Shot Incremental Learners](https://mlsanity.com/paper/2404.02117/) (2024)
- [Class-Incremental Learning using Diffusion Model for Distillation and Replay](https://mlsanity.com/paper/2306.17560/) (2023)
- [Open-Vocabulary Attention Maps with Token Optimization for Semantic Segmentation in Diffusion Models](https://mlsanity.com/paper/2403.14291/) (2024)
- [OrCo: Towards Better Generalization via Orthogonality and Contrast for Few-Shot Class-Incremental Learning](https://mlsanity.com/paper/2403.18550/) (2024)
- [DiffClass: Diffusion-Based Class Incremental Learning](https://mlsanity.com/paper/2403.05016/) (2024)

## Built on by (most cited first, 1 of 1)

- [Where Is the Bee? Detecting Tiny Pollinators with a Single Collaborative-Head Transformer](https://mlsanity.com/paper/2608.08580/) (2026)

## Abstract (arXiv)

Few-shot class-incremental learning (FSCIL) is challenging due to extremely limited training data while requiring models to acquire new knowledge without catastrophic forgetting. Recent works have explored generative models, particularly Stable Diffusion (SD), to address these challenges. However, existing approaches use SD mainly as a replay generator, whereas we demonstrate that SD's rich multi-scale representations can serve as a unified backbone. Motivated by this observation, we introduce Diffusion-FSCIL, which extracts four synergistic feature types from SD by capturing real image characteristics through inversion, providing semantic diversity via class-conditioned synthesis, enhancing generalization through controlled noise injection, and enabling replay without image storage through generative features. Unlike conventional approaches requiring synthetic buffers and separate classification backbones, our unified framework operates entirely in the latent space with only lightweight networks (≈6M parameters). Extensive experiments on CUB-200, miniImageNet, and CIFAR-100 demonstrate state-of-the-art performance, with comprehensive ablations confirming the necessity of each feature type. Furthermore, we confirm that our streamlined variant maintains competitive accuracy while substantially improving efficiency, establishing the viability of generative models as practical and effective backbones for FSCIL.