# Sakana AI — DiffusionBlocks: Training Neural Networks One Block at a Time

- Company: Sakana AI (sakana.ai)
- Announced: 2026-05-27T15:00:00+00:00
- Category: research-paper
- Subject: Sakana models
- Source: https://sakana.ai/diffusion-blocks
- Record: https://forck.live/items/4683-diffusionblocks-training-neural-networks-one-block-at-a-time

Sakana AI introduces DiffusionBlocks, a method to train neural networks one block at a time by reinterpreting the forward pass as a diffusion process, reducing memory requirements while matching end-to-end performance across ViT, DiT, masked diffusion, autoregressive transformers, and recurrent-depth transformers. The paper was accepted at ICLR 2026.

## Evidence

Verbatim from https://sakana.ai/diffusion-blocks:

> We found a new way to break the network into blocks and train them independently. The trick? Treating the network’s forward pass like a diffusion model denoising a signal.

---

Record: https://forck.live/items/4683-diffusionblocks-training-neural-networks-one-block-at-a-time
Catalogue: https://forck.live/llms.txt
Feed: https://forck.live/feed.md