reasoning

Definition

Reaching an answer through intermediate steps — decomposition, partial results, self-checking — rather than in a single pass. It is the headline capability for math, code, and planning benchmarks, and it motivates spending extra compute at inference time on longer deliberation. Do not confuse it with "inference" in the statistical sense, which merely means running the model forward.

Explainers using this term

• Paper Explained: WearableQA — Testing Whether LLMs Can Reason Over 500 Days of a Real User's Wearable DataWearableQA: A Benchmark for Health Reasoning over Real-World Wearable Data
• Paper Explained: Reason Through the Latent! — Visual information being in a latent state doesn't mean the answer uses itReason Through the Latent! Making Latent Visual Reasoning Necessary
• Paper walkthrough: BeaconKV — evicting KV cache by predicting when a reasoning model will re-read its own planBeaconKV: Key-Value Cache Compression Guided by Beacon Queries for Efficient Large Reasoning Model Inference