Saved in:
| Main Authors: | , , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2026
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2603.19635 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866917354232872960 |
|---|---|
| author | Hu, Zhengpei Li, Kai Fu, Dapeng Zeng, Chang Li, Yue Tang, Yuanhao Huang, Jianqiang |
| author_facet | Hu, Zhengpei Li, Kai Fu, Dapeng Zeng, Chang Li, Yue Tang, Yuanhao Huang, Jianqiang |
| contents | The exponential expansion of context windows in LLMs has unlocked capabilities for long-document understanding but introduced severe bottlenecks in inference latency and information utilization. Existing compression methods often suffer from high training costs or semantic fragmentation due to aggressive token pruning. In this paper, we propose BEAVER, a novel training-free framework that shifts compression from linear token removal to structure-aware hierarchical selection. BEAVER maximizes hardware parallelism by mapping variable-length contexts into dense page-level tensors via dual-path pooling, and preserves discourse integrity through a hybrid planner combining semantic and lexical dual-branch selection with sentence smoothing. Extensive evaluations on four long-context benchmarks demonstrate that BEAVER achieves comparable performance to state-of-the-art (SOTA) methods like LongLLMLingua. Notably, on the RULER benchmark, BEAVER maintains high fidelity in multi-needle retrieval where baselines deteriorate. Regarding efficiency, BEAVER reduces latency by 26.4x on 128k contexts, offering a scalable solution for high-throughput applications. Our code is available at https://cslikai.cn/BEAVER/. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2603_19635 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | BEAVER: A Training-Free Hierarchical Prompt Compression Method via Structure-Aware Page Selection Hu, Zhengpei Li, Kai Fu, Dapeng Zeng, Chang Li, Yue Tang, Yuanhao Huang, Jianqiang Computation and Language The exponential expansion of context windows in LLMs has unlocked capabilities for long-document understanding but introduced severe bottlenecks in inference latency and information utilization. Existing compression methods often suffer from high training costs or semantic fragmentation due to aggressive token pruning. In this paper, we propose BEAVER, a novel training-free framework that shifts compression from linear token removal to structure-aware hierarchical selection. BEAVER maximizes hardware parallelism by mapping variable-length contexts into dense page-level tensors via dual-path pooling, and preserves discourse integrity through a hybrid planner combining semantic and lexical dual-branch selection with sentence smoothing. Extensive evaluations on four long-context benchmarks demonstrate that BEAVER achieves comparable performance to state-of-the-art (SOTA) methods like LongLLMLingua. Notably, on the RULER benchmark, BEAVER maintains high fidelity in multi-needle retrieval where baselines deteriorate. Regarding efficiency, BEAVER reduces latency by 26.4x on 128k contexts, offering a scalable solution for high-throughput applications. Our code is available at https://cslikai.cn/BEAVER/. |
| title | BEAVER: A Training-Free Hierarchical Prompt Compression Method via Structure-Aware Page Selection |
| topic | Computation and Language |
| url | https://arxiv.org/abs/2603.19635 |