# BiGain: Unified Token Compression for Joint Generation and Classification

arXiv: https://arxiv.org/abs/2603.12240 · HTML: https://mlsanity.com/paper/2603.12240/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Placed in Diffusion foundations and sampling, one of 14 directions in Diffusion models.
- Builds on 24 of its 30 references that are in the map.
- First posted on arXiv 12 Mar 2026.

## Placed in

- Diffusion models / [Diffusion foundations and sampling](https://mlsanity.com/diffusion/diffusion-foundations-and-sampling/)
- Efficient LLM inference and serving / [Visual token pruning in multimodal models](https://mlsanity.com/inference/visual-token-pruning-in-multimodal-models/)

## Builds on (most cited first, 24 of 24)

- [Denoising Diffusion Probabilistic Models](https://mlsanity.com/paper/2006.11239/) (2020)
- [PyTorch: An Imperative Style, High-Performance Deep Learning Library](https://mlsanity.com/paper/1912.01703/) (2019)
- [High-Resolution Image Synthesis with Latent Diffusion Models](https://mlsanity.com/paper/2112.10752/) (2021)
- [Microsoft COCO: Common Objects in Context](https://mlsanity.com/paper/1405.0312/) (2014)
- [Denoising Diffusion Implicit Models](https://mlsanity.com/paper/2010.02502/) (2020)
- [ImageNet Large Scale Visual Recognition Challenge](https://mlsanity.com/paper/1409.0575/) (2014)
- [Scalable Diffusion Models with Transformers](https://mlsanity.com/paper/2212.09748/) (2022)
- [DPM-Solver: A Fast ODE Solver for Diffusion Probabilistic Model Sampling in Around 10 Steps](https://mlsanity.com/paper/2206.00927/) (2022)
- [Token Merging: Your ViT But Faster](https://mlsanity.com/paper/2210.09461/) (2022)
- [FLUX.1 Kontext: Flow Matching for In-Context Image Generation and Editing in Latent Space](https://mlsanity.com/paper/2506.15742/) (2025)
- [Emergent Correspondence from Image Diffusion](https://mlsanity.com/paper/2306.03881/) (2023)
- [Your Diffusion Model is Secretly a Zero-Shot Classifier](https://mlsanity.com/paper/2303.16203/) (2023)
- [Token Merging for Fast Stable Diffusion](https://mlsanity.com/paper/2303.17604/) (2023)
- [Structural Pruning for Diffusion Models](https://mlsanity.com/paper/2305.10924/) (2023)
- [Text-to-Image Diffusion Models are Zero-Shot Classifiers](https://mlsanity.com/paper/2303.15233/) (2023)
- [Robust Classification via a Single Diffusion Model](https://mlsanity.com/paper/2305.15241/) (2023)
- [LD-Pruner: Efficient Pruning of Latent Diffusion Models using Task-Agnostic Insights](https://mlsanity.com/paper/2404.11936/) (2024)
- [DiP-GO: A Diffusion Pruner via Few-step Gradient Optimization](https://mlsanity.com/paper/2410.16942/) (2024)
- [Not All Diffusion Model Activations Have Been Evaluated as Discriminative Features](https://mlsanity.com/paper/2410.03558/) (2024)
- [ToDo: Token Downsampling for Efficient Generation of High-Resolution Images](https://mlsanity.com/paper/2402.13573/) (2024)
- [Correcting Deviations from Normality: A Reformulated Diffusion Model for Multi-Class Unsupervised Anomaly Detection](https://mlsanity.com/paper/2503.19357/) (2025)
- [MosaicDiff: Training-free Structural Pruning for Diffusion Model Acceleration Reflecting Pretraining Dynamics](https://mlsanity.com/paper/2510.11962/) (2025)
- [Conditional Diffusion Models are Medical Image Classifiers that Provide Explainability and Uncertainty for Free](https://mlsanity.com/paper/2502.03687/) (2025)
- [Data Augmentation in Earth Observation: A Diffusion Model Approach](https://mlsanity.com/paper/2406.06218/) (2024)

## Abstract (arXiv)

Acceleration methods for diffusion models (e.g., token merging or downsampling) typically optimize synthesis quality under reduced compute, yet often ignore discriminative capacity. We revisit token compression with a joint objective and present BiGain, a training-free, plug-and-play framework that preserves generation quality while improving classification in accelerated diffusion models. Our key insight is frequency separation: mapping feature-space signals into a frequency-aware representation disentangles fine detail from global semantics, enabling compression that respects both generative fidelity and discriminative utility. BiGain reflects this principle with two frequency-aware operators: (1) Laplacian-gated token merging, which encourages merges among spectrally smooth tokens while discouraging merges of high-contrast tokens, thereby retaining edges and textures; and (2) Interpolate-Extrapolate KV Downsampling, which downsamples keys/values via a controllable interextrapolation between nearest and average pooling while keeping queries intact, thereby conserving attention precision. Across DiT- and U-Net-based backbones and ImageNet-1K, ImageNet-100, Oxford-IIIT Pets, and COCO-2017, our operators consistently improve the speed-accuracy trade-off for diffusion-based classification, while maintaining or enhancing generation quality under comparable acceleration. For instance, on ImageNet-1K, with 70% token merging on Stable Diffusion 2.0, BiGain increases classification accuracy by 7.15% while improving FID by 0.34 (1.85%). Our analyses indicate that balanced spectral retention, preserving high-frequency detail and low/mid-frequency semantics, is a reliable design rule for token compression in diffusion models. To our knowledge, BiGain is the first framework to jointly study and advance both generation and classification under accelerated diffusion, supporting lower-cost deployment.