NeurIPS paper: Joint Energy Models show human vision aligns between generative and discriminative
ShahabBakht · x · 2026-11-18
A NeurIPS 2026 paper argues vision is explained by neither purely generative nor discriminative learning. Training a single Joint Energy-based Model (JEM) across the generative↔discriminative continuum via an interpolation parameter α, the authors evaluate human alignment on six tasks (gloss perception, similarity, response distributions, diagnostic features, OOD generalization, shape-texture bias) using ImageNet, synthetic and CIFAR-10 data. Alignment peaks in the middle of the continuum, suggesting human-like vision needs both objectives.
More from Research
- BitDecoding-FP4: native FP4 KV cache decoding via algorithm-system co-design — UoE-Mlsys · 2026-11-18
- HT-Policy: multimodality isn't why diffusion policies win at imitation learning — chris_j_paxton · 2026-11-18
- A repo of small vision models built for hacking, not chasing SOTA — mervenoyann · 2026-11-18
- Hypothesis: Higher Adam β1 May Help More in Small-Batch, Low-LR Training Regimes — torchcompiled · 2026-11-18
- SWE-CoWork benchmark: best model fully passes only 8 of 101 handoff coding tasks — omarsar0 · 2026-11-18
- Pre-training trend watch: dropping LR decay for constant LR plus checkpoint averaging — rasbt · 2026-11-18