LongBench

LongBench is a comprehensive benchmark for multilingual and multi-task purposes, with the goal to fully measure and evaluate the ability of pre-trained language models to understand long text. This dataset consists of twenty different tasks, covering key long-text application scenarios such as multi-document QA, single

Papers using LongBench (6)

LongCrafter: Towards Diverse Long-Context Understanding via Evidence-Graph-Guided Instruction Synthesis2026DepthWeave-KV: Token-Adaptive Cross-Layer Residual Factorization for Long-Context KV Cache Compression2026VarRate: Training-Free Variable-Rate KV Cache Compression for Long-Context LLMs2026GRASP: Graph Agentic Search over Propositions for Multi-hop Question Answering2026Learning to Evict from Key-Value Cache2026LightTransfer: Your Long-Context LLM is Secretly a Hybrid Model with Effortless Adaptation2024