arXiv cs.CL· Romtham Tungsununtham, Piyawat Lertvittayakumjorn, Atiwong Suchato, Peerapon Vateekul·· 11 小时前AI 评分31
EffComp：用混合强化学习与监督学习实现高效提示词压缩
EffComp: Efficient Prompt Compression via Hybrid Reinforcement & Supervised Learning
AI 导读
EffComp 是一个面向 RAG 问答的高效提示词压缩框架，采用 BERT 风格的文档重排器与句子选择器在句子级别做快速抽取式压缩，抽取式特性可避免压缩后提示词出现模型幻觉。其训练结合强化学习与监督学习，在保持 LLM 准确率的同时优化压缩比。在多个 QA 数据集上，EffComp 在压缩比与准确率之间取得优于 SOTA 提示词压缩方法的权衡，且延迟极低。
正文 · 原文
Abstract:Prompt compression is aimed at reducing input prompt lengths to enable cheaper, faster, and potentially higher-quality LLM predictions. However, existing prompt compression methods are often limited by modest compression gains, a risk of hallucination, and/or high compression latency. We propose EffComp, an Efficient prompt Compression framework for RAG-based question answering (QA). EffComp employs BERT-style document reranker and sentence selector models to allow fast extractive prompt compression at the sentence level. Its extractive nature prevents hallucinations in the compressed prompts. Additionally, the training process using a hybrid reinforcement and supervised learning approach is designed to optimize the compression ratio while preserving LLM accuracy. Experiments on various QA datasets demonstrate that EffComp outperforms state-of-the-art prompt compression methods achieving the best trade-off between compression ratio and accuracy with minimal latency. This makes it practical for real-world applications.
| Comments: | Finding of AACL-IJCNLP 2026 |
| Subjects: | Computation and Language (cs.CL) |
| Cite as: | arXiv:2611.02766 [cs.CL] |
| (or arXiv:2611.02766v1 [cs.CL] for this version) | |
| https://doi.org/10.48550/arXiv.2611.02766 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Romtham Tungsununtham [view email]
[v1]
Mon, 2 Nov 2026 08:46:28 UTC (900 KB)
来源：arXiv cs.CL · arxiv.org