Ktoken Merging
1 mentions across 0 people
All mentions
Unknown speaker
Recommendedpaper · 2026-04-17
“In this paper, we propose K-Token Merging, a latent-space compression framework that merges each contiguous block of K token embeddings into a single embedding via a lightweight encoder.”
K-Token Merging: Latent-Space Compression for LLMs ↗