Waiting to Decompress: The Economics of LLM-Based Compression
Summary: Benchmarked and speed‑optimized LLM-based compression methods and introduced a cost model quantifying runtime–cost tradeoffs. Finds current economics: ~10 years to amortize compute vs storage and ~120 years to beat traditional compressors, though hardware/model efficiency trends could make LLM compression viable within a decade. (summarized by gpt-5-mini on Feb 09 2026)
Incoming Non-self Citations Over Time
No non-self incoming citations found for this paper in this database.
Authors
- 1. Andreas Kipf (University of Technology Nuremberg)
- 2. Tobias Schmidt (Technical University of Munich)
- 3. Ping-Lin Kuo (University of Technology Nuremberg)
- 4. Skander Krid (University of Technology Nuremberg)
- 5. Moritz Rengert (University of Technology Nuremberg)
- 6. Luca Heller (University of Technology Nuremberg)
- 7. Andreas Zimmerer (University of Technology Nuremberg)
- 8. Mihail Stoian (University of Technology Nuremberg)
- 9. Varun Pandey (University of Technology Nuremberg)
- 10. Alexander van Renen (University of Technology Nuremberg)
BibTeX Citation
@inproceedings{kipf_cidr26,
address = {Amsterdam, Netherlands},
series = {{CIDR} '26},
title = {{Waiting to Decompress: The Economics of LLM-Based Compression}},
booktitle = {Proceedings of the {Conference} on {Innovative} {Data} {Systems} {Research}},
author = {Kipf, Andreas and Schmidt, Tobias and Kuo, Ping-Lin and Krid, Skander and Rengert, Moritz and Heller, Luca and Zimmerer, Andreas and Stoian, Mihail and Pandey, Varun and van Renen, Alexander},
year = {2026}
}Incoming Citations (Sorted by Pagerank)
Showing 0 of 0 citing papers.
| Rank | Citing Paper | Year | Venue | Pagerank |
|---|
Previous
Page 1 / 1
Next
Outgoing Citations (Sorted by Pagerank)
Showing 0 of 0 cited papers.
Citations counted here include only citations to other VLDB/SIGMOD/CIDR/PODS papers in this database.
| Rank | Cited Paper | Year | Venue | Pagerank |
|---|
Previous
Page 1 / 1
Next
Semantically Similar Papers
| # | Overall Rank | Paper | Year | Venue |
|---|---|---|---|---|
| 1 | 9,972 | Cents: A Flexible and Cost-Effective Framework for LLM-Based Table Understanding | 2025 | VLDB |
| 2 | 921 | Query Optimization In Compressed Database Systems | 2001 | SIGMOD |
| 3 | 10,856 | Optimized Batch Prompting for Cost-effective LLMs | 2025 | VLDB |
| 4 | 7,173 | CompressDB: Enabling Efficient Compressed Data Direct Processing for Various Databases | 2022 | SIGMOD |
| 5 | 9,520 | Experimental Analysis of Large-scale Learnable Vector Storage Compression | 2024 | VLDB |
| 6 | 8,627 | ThriftLLM: On Cost-Effective Selection of Large Language Models for Classification Queries | 2025 | VLDB |
| 7 | 7,865 | Generating Succinct Descriptions of Database Schemata for Cost-Efficient Prompting of Large Language Models | 2024 | VLDB |
| 8 | 10,733 | ScaleLLM: A Technique for Scalable LLM-augmented Data Systems | 2025 | SIGMOD |
| 9 | 8,829 | Cut Costs, Not Accuracy: LLM-Powered Data Processing with Guarantees | 2026 | SIGMOD |
| 10 | 13,343 | Database Perspective on LLM Inference Systems | 2025 | VLDB |