SoftServe: A quasi-Newton method for non-convex objectives that scales to very large neural networks
dianarycai · x · 2026-10-02
A new preprint introduces SoftServe, a quasi-Newton optimization method designed for non-convex objectives that also scales to very large neural networks. Co-authored with optimization researchers including Robert Gower, the work aims to make quasi-Newton methods practical for modern deep learning.
Related event: SoftServe: Quasi-Newton Optimization Scales to Massive Neural Networks(3 posts)→
More from Research
- Company Knowledge Bench: real-world retrieval benchmark tests 7 retrievers — CShorten30 · 2026-10-03
- Gacha Decoding diversifies LLM outputs with 11x better sample efficiency — natolambert · 2026-10-03
- New paper sparks debate: intelligence vs parroting is a computational distinction — ctjlewis · 2026-10-03
- RLE-Bench grades coding agents as robot learning engineers across four workflows — RLE-Bench · 2026-10-03
- Distillation study: KL direction and learning rate matter more than on-policy rollouts — CambUni · 2026-10-03
- Stanford revisits 2018 claim that self-driving was 90% done — the last 10% was everything — StanfordHAI · 2026-10-03