TTE-Flash: Accelerating Reasoning-based Multimodal Representations via Think-Then-Embed Tokens

In three linesTTE-Flash replaces explicit Chain-of-Thought traces with latent think tokens to accelerate reasoning-aware multimodal representations. TTE-Flash-2B outperforms explicit-CoT counterparts on MMEB-v2 while maintaining constant inference cost. Latent tokens remain interpretable both textually and visually.

Summary generated by Claude — human-verified