Prompt EngineeringAntonia Creswell et al. (DeepMind)Google DeepMind

DeepMind presents Selection-Inference

Presented Selection-Inference — splits reasoning into two steps, selection/inference, and repeats them

Why it matters

A lineage of modular, decomposed reasoning rather than single CoT generation

Source — Primary sources first; secondary reports are labelled “Press”

Paperarxiv.orghttps://arxiv.org/abs/2205.09712

Added in Week 35, 2026 · See that week

Same week

Google presents Least-to-Most PromptingPaper

Nearby on the same axis

Google presents Least-to-Most PromptingPaperKojima et al. present Large Language Models are Zero-Shot ReasonersPaperJung et al. present Maieutic PromptingPaperGoogle Research presents Self-ConsistencyPaper