New Paper Teaser: Pre-training with Zero Order
ycombinator · x · 2026-07-17
This post shares a teaser for an upcoming paper with a radical core claim: you don't need transformers or backprop to do pre-training; Zero Order works too.
The poster notes that the results are "very impressive" and could inspire new architectural directions, making the inference phase more CPU-bound rather than IO-bound. Overall, this is a cutting-edge research preview, with full details awaiting the official paper release.
More from Research
- NeurIPS 2026 workshop calls papers on on-device intelligence — YiMaTweets · 2026-07-21
- AI Security Institute says every tested model tried to cheat in cyber evaluations — connoraxiotes · 2026-07-21
- Sakana says multiple diffusion models plus MCTS beat test-time scaling on coding and math — SakanaAILabs · 2026-07-21
- Soofi S 30B-A3B releases a full pretraining report and claims open-model leads in English and German — abursuc · 2026-07-21
- AI Companies Are Buying Tons of Old Books Because They're Free of AI Slop — 404 Media · 2026-07-21
- Shared agent workspaces fail in a fixed order, from stale reads to zombie writes — mrvladp · 2026-07-21