SAGE finds unsupervised salient factors in images, hitting 0.950 probe accuracy on FFHQ glasses
Shuang Liang · hf · 2026-10-12
Researchers from HKU, Boston College and others introduce SAGE, a contrastive analysis method that separates target-specific salient factors from shared content given target and background datasets (e.g. faces with vs. without eyeglasses).
- Learns both factors directly in the high-dimensional spatial latent of a frozen representation autoencoder, and conditions a diffusion transformer on a reference image's salient representation for generation.
- On Digits-ImageNet and FFHQ eyeglasses, it combines high-fidelity reconstruction (rFID below 2) with unsupervised subtype discovery: probe accuracy of 0.950 for digit recovery vs. at most 0.281 for baselines, plus discovery of finer eyewear styles and mislabeled images.
- Salient-conditioned generation lifts Digits-ImageNet subtype accuracy from 27.7% to 90.5% and increases diversity on both datasets.
- On retinal OCT, SAGE's salient space separates three diseases using only normal/disease labels.
More from Research
- Critique says OpenAI's Navier-Stokes Lean formalization checked a different argument — thione · 2026-10-12
- OpenAI drops 700+ math manuscripts; human mathematicians group urges boycott — thione · 2026-10-12
- Biohub expands AI-biology data effort to $1.8B with US agencies, Meta and Google — thione · 2026-10-12
- Paper: DMD turns diffusion models into looped models iterating in data space — CSProfKGD · 2026-10-12
- Jina AI to host 4th BoF on embeddings, rerankers & small LMs at EMNLP 2026 — JinaAI_ · 2026-10-12
- Yoav Goldberg defends OpenAI's FFT result: critique TCS culture, not the paper — yoavgo · 2026-10-12