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 (38)

LongCrafter: Towards Diverse Long-Context Understanding via Evidence-Graph-Guided Instruction Synthesis2026GSRQ: Gain-Shape Residual Quantization for Sub-1-bit KV Cache2026DepthWeave-KV: Token-Adaptive Cross-Layer Residual Factorization for Long-Context KV Cache Compression2026Forget Without Compromise: Nexus Sampling for Streaming KV-Cache Eviction Under Fixed Budgets2026CompressKV: Semantic-Retrieval-Guided KV-Cache Compression for Resource-Efficient Long-Context LLM Inference2026Dustin: Draft-Augmented Sparse Verification for Efficient Long-Context Generation with Speculative Decoding2026SSM Adapters via Hankel Reduced-order Modeling: Injection Site Determines Task Suitability in Long-Context Fine-Tuning2026IndexMem: Learned KV-Cache Eviction with Latent Memory for Long-Context LLM Inference2026AttentionRAG: Attention-Guided Context Pruning in Retrieval-Augmented Generation2025 · 1 citesActivation-aware Probe-Query: Effective Key-Value Retrieval for Long-Context LLMs Inference2025 · 1 citesART: Attention Run-time Termination for Efficient Large Language Model Decoding2026LycheeDecode: Accelerating Long-Context LLM Inference via Hybrid-Head Sparse Decoding2026Reinforced Fast Weights with Next-Sequence Prediction2026IceCache: Memory-efficient KV-cache Management for Long-Sequence LLMs2026EndPrompt: Efficient Long-Context Extension via Terminal Anchoring2026HISA: Efficient Hierarchical Indexing for Fine-Grained Sparse Attention2026ProxyKV: Cross-Model Proxy Pruning for Efficient Long-Context LLM Inference2026M-RAG: Making RAG Faster, Stronger, and More Efficient2026Developing Adaptive Context Compression Techniques for Large Language Models (LLMs) in Long-Running Interactions2026AllMem: A Memory-centric Recipe for Efficient Long-context Modeling2026Towards robust long-context understanding of large language model via active recap learning2026PagedEviction: Structured Block-wise KV Cache Pruning for Efficient Large Language Model Inference2025DSPC: Dual-Stage Progressive Compression Framework for Efficient Long-Context Reasoning2025Overflow Prevention Enhances Long-Context Recurrent LLMs2025Loong: Synthesize Long Chain-of-Thoughts at Scale through Verifiers2025ChunkKV: Semantic-Preserving KV Cache Compression for Efficient Long-Context LLM Inference2025MDocAgent: A Multi-Modal Multi-Agent Framework for Document Understanding2025Beyond Homogeneous Attention: Memory-Efficient LLMs via Fourier-Approximated KV Cache2025MacRAG: Compress, Slice, and Scale-up for Multi-Scale Adaptive Context RAG2025MiniLongBench: The Low-cost Long Context Understanding Benchmark for Large Language Models2025An Empirical Study on Prompt Compression for Large Language Models2025PromptDistill: Query-based Selective Token Retention in Intermediate Layers for Efficient Large Language Model Inference2025CriticalKV: Optimizing KV Cache Eviction from an Output Perturbation Perspective2025Does RAG Really Perform Bad For Long-Context Processing?2025Task-agnostic Prompt Compression with Context-aware Sentence Embedding and Reward-guided Task Descriptor2025Extending Context Window of Large Language Models from a Distributional Perspective2024Distance between Relevant Information Pieces Causes Bias in Long-Context LLMs2024LIFT: Improving Long Context Understanding Through Long Input Fine-Tuning2024