TEM: first scalable contrastive energy-based model beats diffusion on OOD and scaling
Total-Resort-3120 · reddit · 2026-11-01
TEM (Transformer Energy-based Models) claims to be the first scalable contrastive energy-based model.
Key takeaways: it outscales diffusion without guidance, and shows better generalization, test-time scaling, and out-of-distribution detection than diffusion models. Sample images suggest ImageNet-class generation quality — a notable attempt to scale a generative route beyond diffusion.
More from Research
- New skeletal muscle single-cell eQTL resource from 287 individuals — anshulkundaje · 2026-11-01
- Hackathon study finds agent swarm language shifts measurably under deadline pressure — Miles_Brundage · 2026-11-01
- TensorAgent Adds Local Multimodal Indexing for Documents, Images, Audio and Video — fuzhongkai · 2026-11-01
- FDSML, accepted at WACV 2027, lifts formula-driven pretraining Recall@1 to 74% — HirokatuKataoka · 2026-11-01
- PNAS study: Glioblastoma organoids grown 40 days on ISS mimic aggressive disease better than on Earth — iScienceLuvr · 2026-11-01
- ETH Zurich releases browser-playable Unitree G1 parkour policy with attention-based contacts — ChongZzZhang · 2026-11-01