Share your thoughts, 1 month free Claude Pro on usSee more

Home/Benchmarks

Code Completion on RepoBench

48.92Pass@k Score

LAVA

Updated 18d ago

Evaluation Results

Method	Links
LAVA 2026.03		48.92

SOTA Paper

LAVA

LAVa: Layer-wise KV Cache Eviction with Dynamic Budget Allocation

Dataset

RepoBench

Follow for update

@wizwand_team Discord

Related Benchmarks

Long-running conversation on LOCOMO (test)Long-term Conversational Memory Evaluation on LOCCO Long-context Conversational Memory on LOCCO Long-context Memory Management on LOCOMO Long-context Understanding on LongBench (Average Score and Length Statistics)

© 2026 wizwand

Blog Contact Changelog Swarm

Privacy Policy Terms of Service FAQs Swarm Docs