Equilibrium Matching Outperforms Flow Matching with FID 1.96 on ImageNet
du_yilun · x · 2026-08-14
duyilun announces Equilibrium Matching (EqM), which simplifies and outperforms flow matching, achieving strong generative performance with FID 1.96 on ImageNet 256x256. EqM learns a single static EBM landscape for generation, enabling simple gradient-based generation.
Related event: Equilibrium Matching Surpasses Flow Matching with FID 1.96(2 posts)→
More from Research
- HiFi-UMI: Robots Learn Skills from Human Videos with Zero Real-World Practice — jiqizhixin · 2026-08-14
- Universal RCE Deserialization Chain for Ruby 4.0 Released; AI Agents Exploited Similar Flaw — jedisct1 · 2026-08-14
- IOL-AI 2026 Winners Announced: 46 Teams, 731 Submissions, Tier 1 Champions Crowned — Cohere_Labs · 2026-08-14
- Automated alignment research (AAR) runs are hard to study: Arcadia Impact tracks runs to map failure modes — morgymcg · 2026-08-14
- Open-source BabelDOC translates PDF papers and generates bilingual comparisons, 9.3k stars on GitHub — tom_doerr · 2026-08-14
- Developer shares private notes on distilling Flow/Diffusion models, including teacher model link — ariG23498 · 2026-08-14