Algorithms & Theory × Developer Tools

Google shifts search cost into training

Google shifts search cost into training

✎ Story body

Google has moved the cost of "going out to search" from serving time to training time. Other work, meanwhile, deliberately keeps that search on the inference side.

What happened

Google Research published Retrieve-for-Train. Rather than returning the single best match, it returns a coherent slate of results that complement one another. Ask for camping gear and it comes back with a tent, a sleeping bag and a stove together. The older approach paid its cost on every query, stacking up autoregressive reasoning each time. This one trains a lightweight diffusion model once, via reinforcement learning, and then generates the whole slate instantly.

Why it matters

Making a model reason on every query is slow and expensive. Paying up front at training time makes serving fast and cheap, but the behaviour is fixed until you retrain. Other work runs the other way. HypoEvolve uses genetic algorithms to let multiple LLM agents search for scientific hypotheses, which effectively pushes more search onto inference. Which choice is right depends on the job. Search that must answer instantly and research that can afford to take its time do not want the cost in the same place.

What to watch

The weakness of paying up front is what happens when the world changes after training ends. How Retrieve-for-Train reports the frequency and cost of retraining will decide whether this becomes the default shape of search or stays a fit for particular uses.

▲ Official & Press
Official

Bypassing inference bottlenecks: Accelerating complex AI search with Retrieve-for-Train

Google Research Blog ・ 2026-09-15 ・ 📌

Google unveils Retrieve-for-Train, skipping inference-time reasoning

Academic (arxiv etc.) 6 ▾
Academic

HypoEvolve: Genetic Algorithms Enable Multi-Agent LLMs to Discover Scientific Hypotheses

arXiv cs.CL (Computation and Language) ・ 2026-09-14

Academic

Quenched Ensemble Sampling

arXiv cs.LG (Machine Learning) ・ 2026-09-14

Academic

Solving Finite-sum Coupled Compositional Optimization via Multi-block-Single-probe Estimator

arXiv cs.LG (Machine Learning) ・ 2026-09-14

Academic

Projection-Free Multi-level Algorithms for Stochastic Constrained Compositional Optimization

arXiv cs.LG (Machine Learning) ・ 2026-09-14

Academic

Potential of Artificial Intelligence Algorithms for Identification of Relevant Diagnostic and Prognostic Biomarkers of Early-Stage Liver Cancer

arXiv cs.AI (Artificial Intelligence) ・ 2026-09-14

Academic

FedLTLib: A Comprehensive Benchmark for Federated Long-Tail Learning

arXiv cs.AI (Artificial Intelligence) ・ 2026-09-14

← Story Archive