Records found: 4
SwiReasoning alternates between silent latent reasoning and explicit chain-of-thought based on next-token entropy, improving accuracy and dramatically cutting token usage on math and STEM benchmarks.
'The Gemini 2.5 Flash and Flash‑Lite preview updates sharpen tool use and multi‑pass reasoning while Flash‑Lite tests as the fastest proprietary model with ~50% fewer output tokens; validate on your workloads before switching to rolling aliases.'
'Meta AI compresses recurring CoT reasoning into short named behaviors that models can retrieve or internalize, cutting reasoning tokens by up to 46% and improving accuracy in some settings.'
Context engineering enhances AI performance by optimizing the input data fed to large language models, enabling more accurate and context-aware outputs across various applications.