Google's 53.9M-param diffusion model speeds search query expansion 12-20x
imjustnewatai · x · 2026-09-16
A Google write-up (Sep 15) describes retrieve-for-train: train an LM with reinforcement learning to generate training examples, then distill a tiny diffusion model — just 53.9M parameters — to produce multiple retrieval expansion directions in parallel, yielding a 12-20x speedup over the autoregressive baseline.
- The goal is more diverse result sets: a camping query should surface a tent, sleeping bag, and stove instead of ten versions of the same tent.
- Tested on fashion and music retrieval; the speedup applies only to the query-expansion stage, not all of Google Search.
- The underlying paper first appeared in March.
Key transferable pattern: spend expensive reasoning during training, then deploy a small specialized model for the repeated task at inference time.
More from Research
- Xiaomi MiMo's CodeSpecWeave trains coding agents on interdependent specs for long-horizon coding — XiaomiMiMo · 2026-11-02
- Mistral Large 4 writes a quantum chemistry SCF solver from scratch with one short prompt — MistralAI · 2026-11-02
- Hunting for the most concrete "curing cancer with AI" roadmap of 2025 — iskander · 2026-11-02
- New theory paper proves linear decay is the unique minimax-optimal stepsize schedule for SGD — konstmish · 2026-11-02
- Karpathy's nanochat: the best ChatGPT $100 can buy, near 60k GitHub stars — karpathy · 2026-11-02
- After reading 46 resources on agent evals, one author explains why her 95% score was misleading — blaizedsouza · 2026-11-02