Prompt EngineeringXuezhi Wang et al. (Google DeepMind)Google DeepMind

Chain-of-Thought Reasoning Without Prompting published

Chain-of-Thought Reasoning Without Prompting published — CoT paths emerge simply by branching on the top-k tokens during decoding instead of greedy decoding

Why it matters

Suggested that CoT may be a decoding problem rather than a prompting one

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

Paperarxiv.orghttps://arxiv.org/abs/2402.10200

Added in Week 35, 2026 · See that week

Same week

Gemini 1.5 ProOfficialOpenAI announces Sora with a technical reportOfficialMicrosoft Research publishes GraphRAG blog postOfficial

Nearby on the same axis

The Unreasonable Effectiveness of Eccentric Automatic Prompts publishedPaperSelf-Discover publishedPaperIEEE Spectrum publishes AI Prompt Engineering Is DeadPressLet's Think Dot by Dot publishedPaper