MOJO preprint mixes supervised and self-supervised losses for neural foundation models
hugo_larochelle · x · 2026-07-25
A new preprint explores hybrid losses for pretraining neurofoundation models on mixed labelled and unlabelled neural data.
- The authors describe MOJO, a joint self-supervised + supervised framework inspired by POYO- and POSSM-style models.
- They report improved decoding and few-shot transfer.
- The method also produces more interpretable unit embeddings across neural datasets and tasks, with several practical advantages emerging along the way.
More from Research
- Manifold Muon offers a loss-free path for training MoE routers — tokenbender · 2026-07-25
- Practical multi-agent orchestration for Codex splits work into scout, worker, and coordinator roles — pvncher · 2026-07-25
- AI can scan more code than humans, but engineers still own quality — ingliguori · 2026-07-25
- NVIDIA reposts a GPT-like motion model that reproduces clips with 99.98% success — Syntetisaattori · 2026-07-25
- A new AGI essay argues the field is climbing the same mountain from two slopes — op7418 · 2026-07-25
- Paper shows models implicitly optimize for hidden graders in audit-style tasks — Sauers_ · 2026-07-25