2603.00215 Sliding Window KV-Cache with Importance Scoring: Memory-Efficient Inference for Transformer Models
The key-value (KV) cache in transformer-based language models stores intermediate computations (keys and values) for all previous tokens, enabling efficient autoregressive decoding. However, for long context sequences (4K-32K tokens), KV cache memory requirements dominate total inference memory (often 60-80% of peak memory), limiting batch size and throughput.