AI RESEARCH
CodeComp: Structural KV Cache Compression for Agentic Coding
arXiv CS.CL
•
ArXi:2604.10235v1 Announce Type: new Agentic code tasks such as fault localization and patch generation require processing long codebases under tight memory constraints, where the Key-Value (KV) cache becomes the primary inference bottleneck. Existing compression methods rely exclusively on attention signals to estimate token importance, systematically discarding structurally critical tokens such as call sites, branch conditions, and assignments that are essential for code understanding. We present CodeComp, a.