Microsoft Research Boosts LLM Performance at Zero Cost via Hidden State Reuse

Microsoft's AI Frontiers team proposed a novel method to improve LLM performance by injecting previous hidden states during the decoding phase, enhancing Transformer capabilities at zero additional inference cost.

2026-08-12 ~ 2026-08-13 · 2 related posts

• Microsoft Research: Injecting Recursive Hidden States Boosts Transformer Performance for Free — furongh · 2026-08-12
• Microsoft Intern Project: Reusing Hidden States at Decoding Boosts LLM Performance for Free — burny_tech · 2026-08-13