Polymathic AI and Princeton Researchers Release FlashMap, a Flow Map Drafter for 6.14x Faster LLM Decoding

Researchers from Polymathic AI, Princeton, and USC ISI have released FlashMap, a speculative decoding drafter designed to accelerate large language model (LLM) inference.

Key Capabilities

• Joint drafting: FlashMap drafts 16 tokens jointly in one pass.
• Speedup: It reaches up to 6.14x lossless speedup on Qwen3-8B.

How It Works

FlashMap is a flow map drafter built for speculative decoding, a technique that uses a smaller drafter model to propose multiple tokens ahead, which are then verified by the target model to preserve output quality.

Research Team

The work comes from researchers at Polymathic AI, Princeton, and USC ISI.