Archive
LLM Infrastructure

KV Cache

Archive
Medium
150pts55 solves
During generation, the model would recompute attention for all past tokens at each step. The KV cache stores intermediate results. What does it store? Flag format: CONGRESS{[what]} Example: CONGRESS{embedding_vectors}
Show hint
The K and V in KV refer to attention mechanism components.

Archive — no submissions accepted

This challenge is preserved for reference. Play live challenges at /challenges.