Cmprsr: Abstractive Token-Level Question-Agnostic Prompt Compressor
Fuente:
arXiv
Enregistré dans:
| Auteurs principaux: | , , , , , , , , |
|---|---|
| Format: | Preprint |
| Publié: |
2025
|
| Sujets: | |
| Accès en ligne: | |
| Tags: |
Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
|
| _version_ | 1866908754504581120 |
|---|---|
| author | Zakazov, Ivan Argin, Berke Gabouj, Oussama Charaf, Kamel Sharipov, Alexander Semiz, Alexi Drudi, Lorenzo Baldwin, Nicolas West, Robert |
| author_facet | Zakazov, Ivan Argin, Berke Gabouj, Oussama Charaf, Kamel Sharipov, Alexander Semiz, Alexi Drudi, Lorenzo Baldwin, Nicolas West, Robert |
| contents | Motivated by the high costs of using black-box Large Language Models (LLMs), we introduce a novel prompt compression paradigm, under which we use smaller LLMs to compress inputs for the larger ones. We present the first comprehensive LLM-as-a-compressor benchmark spanning 25 open- and closed-source models, which reveals significant disparity in models' compression ability in terms of (i) preserving semantically important information (ii) following the user-provided compression rate (CR). We further improve the performance of gpt-4.1-mini, the best overall vanilla compressor, with Textgrad-based compression meta-prompt optimization. We also identify the most promising open-source vanilla LLM - Qwen3-4B - and post-train it with a combination of supervised fine-tuning (SFT) and Group Relative Policy Optimization (GRPO), pursuing the dual objective of CR adherence and maximizing the downstream task performance. We call the resulting model Cmprsr and demonstrate its superiority over both extractive and vanilla abstractive compression across the entire range of compression rates on lengthy inputs from MeetingBank and LongBench as well as short prompts from GSM8k. The latter highlights Cmprsr's generalizability across varying input lengths and domains. Moreover, Cmprsr closely follows the requested compression rate, offering fine control over the cost-quality trade-off. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2511_12281 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Cmprsr: Abstractive Token-Level Question-Agnostic Prompt Compressor Zakazov, Ivan Argin, Berke Gabouj, Oussama Charaf, Kamel Sharipov, Alexander Semiz, Alexi Drudi, Lorenzo Baldwin, Nicolas West, Robert Computation and Language Machine Learning Motivated by the high costs of using black-box Large Language Models (LLMs), we introduce a novel prompt compression paradigm, under which we use smaller LLMs to compress inputs for the larger ones. We present the first comprehensive LLM-as-a-compressor benchmark spanning 25 open- and closed-source models, which reveals significant disparity in models' compression ability in terms of (i) preserving semantically important information (ii) following the user-provided compression rate (CR). We further improve the performance of gpt-4.1-mini, the best overall vanilla compressor, with Textgrad-based compression meta-prompt optimization. We also identify the most promising open-source vanilla LLM - Qwen3-4B - and post-train it with a combination of supervised fine-tuning (SFT) and Group Relative Policy Optimization (GRPO), pursuing the dual objective of CR adherence and maximizing the downstream task performance. We call the resulting model Cmprsr and demonstrate its superiority over both extractive and vanilla abstractive compression across the entire range of compression rates on lengthy inputs from MeetingBank and LongBench as well as short prompts from GSM8k. The latter highlights Cmprsr's generalizability across varying input lengths and domains. Moreover, Cmprsr closely follows the requested compression rate, offering fine control over the cost-quality trade-off. |
| title | Cmprsr: Abstractive Token-Level Question-Agnostic Prompt Compressor |
| topic | Computation and Language Machine Learning |
| url | https://arxiv.org/abs/2511.12281 |