REAL: REtrieval-reAsoning and Logic-constructed Attention Behaviors for Long-Context KV Cache Compression
-
Updated
Apr 15, 2026 - Python
REAL: REtrieval-reAsoning and Logic-constructed Attention Behaviors for Long-Context KV Cache Compression
Diagnostic evaluation of long-context reasoning in LLMs using LongBench v2 (context scaling, retrieval vs reasoning, prompting strategies)
Add a description, image, and links to the longbench topic page so that developers can more easily learn about it.
To associate your repository with the longbench topic, visit your repo's landing page and select "manage topics."