RandOpt: Repeated Runs Continuously Boost Performance Without Test-Time Ensembling
phillip_isola · x · 2026-07-08
Phillip Isola shared an update on his RandOpt work: running RandOpt multiple times continuously improves model performance. Furthermore, this mechanism achieves impressive results without the need for test-time ensembling.
Related event: RandOpt: Achieving High Performance via Iterative Distillation(2 posts)→
More from Research
- Open-source LangGraph coding agent only submits patches after tests pass — wusuiling-if · 2026-07-21
- A Markov-chain note explores using eigenvectors to speed up mixing — michaelchchoi · 2026-07-21
- Xiaomi-Robotics-1 shows robot motion improves more from data than bigger models — The Decoder · 2026-07-21
- FloC 2026 AIMACS workshop on AI for math and CS set for July 25 — swarat · 2026-07-21
- Knowledgeless Language Models cut closed-book recall by anonymizing entities during pretraining — gdm3000 · 2026-07-21
- CPU-native LLM pilot passes 4 of 5 gates, but cross-tokenizer distillation still loses — WildPino25 · 2026-07-21