FINCH: Prompt-guided Key-Value Cache Compression for Large Language Models
Open accessAbstract
Recent large language model applications, such as Retrieval-Augmented Generation and chatbots, have led to an increased need to process longer input contexts. However, this requirement is hampered by inherent limitations. Architecturally, models are constrained by a context window defined during training. Additionally, processing extensive texts requires substantial GPU memory. We propose a novel approach, Finch, to compress the input context by leveraging the pre-trained model weights of the self-attention. Given a prompt and a long text, Finch iteratively identifies the most relevant Key (K) and Value (V) pairs over chunks of the text conditioned on the prompt. Only such pairs are stored in the KV cache, which, within the space constrained by the context window, ultimately contains a compressed version of the long text. Our proposal enables models to consume large inputs even with high compression (up to 93x) while preserving semantic integrity without the need for fine-tuning.
Integrity signals
Registry
Cites
Statistics
Claims the paper makes
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
The context distillation objective captures the counterfactual discrepancy between two views of the same model with different context lengths.
“The context distillation objective captures the counterfactual discrepancy between two views of the same model with different context lengths.”
From Learning What to Remember: Test-Time Training via Context Distillation (2026) · cites this paper as [10]
Cites
Lost in the Middle: How Language Models Use Long Contexts
Liu, N. F., Lin, K., Hewitt, J., Paranjape, A., Bevilacqua, M., Petroni, F., & Liang, P. (2024) · Transactions of the Association for Computational Linguistics · doi:10.1162/tacl_a_00638 · Cited by 1,612 · see citing works
QLoRA: Efficient Finetuning of Quantized LLMs
Dettmers, T., Pagnoni, A., Holtzman, A., & Zettlemoyer, L. (2023) · arXiv · doi:10.48550/arxiv.2305.14314 · Cited by 507 · see citing works
LongBench: A Bilingual, Multitask Benchmark for Long Context Understanding
Bai, Y., Lv, X., Zhang, J., Lyu, H., Tang, J., Huang, Z., Du, Z., Liu, X., Zeng, A., Hou, L., et al. (2024) · Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) · doi:10.18653/v1/2024.acl-long.172 · Cited by 400 · see citing works
LLMLingua: Compressing Prompts for Accelerated Inference of Large Language Models
Jiang, H., Wu, Q., Lin, C.-Y., Yang, Y., & Qiu, L. (2023) · Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing · doi:10.18653/v1/2023.emnlp-main.825 · Cited by 168 · see citing works
LLMLingua: Compressing Prompts for Accelerated Inference of Large Language Models
Jiang, H., Wu, Q., Lin, C.-Y., Yang, Y., & Qiu, L. (2023) · arXiv · Cited by 168 · see citing works
LongLLMLingua: Accelerating and Enhancing LLMs in Long Context Scenarios via Prompt Compression
Jiang, H., Wu, Q., Luo, X., Li, D., Lin, C.-Y., Yang, Y., & Qiu, L. (2024) · Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) · doi:10.18653/v1/2024.acl-long.91 · Cited by 96 · see citing works
Efficient Streaming Language Models with Attention Sinks
Xiao, G., Tian, Y., Chen, B., Han, S., & Lewis, M. (2023) · arXiv · doi:10.48550/arxiv.2309.17453 · Cited by 55 · see citing works
Language Modeling Is Compression
Delétang, G., Ruoss, A., Duquenne, P.-A., Catt, E., Genewein, T., Mattern, C., Grau-Moya, J., Wenliang, L. K., Aitchison, M., Orseau, L., et al. (2023) · arXiv · doi:10.48550/arxiv.2309.10668 · Cited by 39 · see citing works
Wingate, D., Shoeybi, M., & Sorensen, T. (2022) · Findings of the Association for Computational Linguistics: EMNLP 2022 · doi:10.18653/v1/2022.findings-emnlp.412 · Cited by 26 · see citing works
AdapLeR: Speeding up Inference by Adaptive Length Reduction
Modarressi, A., Mohebbi, H., & Pilehvar, M. T. (2022) · Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) · doi:10.18653/v1/2022.acl-long.1 · Cited by 18 · see citing works
Transformers are Multi-State RNNs
Oren, M., Hassid, M., Yarden, N., Adi, Y., & Schwartz, R. (2024) · arXiv · doi:10.48550/arxiv.2401.06104 · Cited by 16
Model Tells You What to Discard: Adaptive KV Cache Compression for LLMs
Ge, S., Zhang, Y., Liu, L., Zhang, M., Han, J., & Gao, J. (2023) · arXiv · doi:10.48550/arxiv.2310.01801 · Cited by 13 · see citing works
LongLLMLingua: Accelerating and Enhancing LLMs in Long Context Scenarios via Prompt Compression
Jiang, H., Wu, Q., Luo, X., Li, D., Lin, C.-Y., Yang, Y., & Qiu, L. (2023) · arXiv · Cited by 12 · see citing works
Flagged first, then by citations
Cited by
Sorted by citations
Logical and Physical Optimizations for SQL Query Execution over Large Language Models
Satriani, D., Veltri, E., Santoro, D., Rosato, S., Varriale, S., & Papotti, P. (2025) · Proceedings of the ACM on Management of Data · doi:10.1145/3725411 · Cited by 8 · see citing works
Self-citationLearning What to Remember: Test-Time Training via Context Distillation
Wang, Z., Dang, X., Zhu, R.-J., Wen, Z., Fu, H., Chai, W., & Lee, J. D. (2026) · arXiv · doi:10.48550/arxiv.2608.01672
“We propose TTCD ( T est-T ime C ontext D istillation), a test-time training method based on context distillation [40, 10, 27, 14, 6].”
- Authors
- Corallo, G., & Papotti, P.
- Affiliations
- EURECOM (FR)
- Venue
- Transactions of the Association for Computational Linguistics
- Publisher
- MIT Press
- Published
- 2024
- Volume
- 12
- Pages
- 1517-1532
- ISSN
- 2307-387X (ISSN-L)
- OpenAlex
- W4404609179
- Cited by
- 5
- References
- 13
- Citations by year
- 2026: 2 · 2025: 3
- FWCI
- 0.74
- Type
- journal article
- Language
- en
- Open access
- diamond
- doi.org
- Locations
- Transactions of the Association for Computational Linguistics (publishedVersion, cc-by)
- Topics
- Physical Sciences › Computer Science › Artificial Intelligence › Topic Modeling
- Physical Sciences › Computer Science › Artificial Intelligence › Natural Language Processing Techniques
- Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Multimodal Machine Learning Applications
- Keywords
- key-value cache compression, large language models, context window, GPU memory efficiency, self-attention, retrieval-augmented generation, chatbots
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
