SpareLLM: Automatically Selecting Task-Specific Minimum-Cost Large Language Models under Equivalence Constraint

Summary: SpareLLM selects task-specific, minimum-cost LLMs with an output-equivalence constraint. Profiling-first approach and heterogeneous model cascades yield Pareto-optimal cost-accuracy tradeoffs, up to 8.6x savings and 90% equivalence to GPT-4-Turbo. (summarized by gpt-5-nano on Feb 09 2026)

Paper ID: 7332
Venue: SIGMOD
Year: 2025
Pagerank: 5.646695e-05
Overall Rank: 7,316 | 49.81%
DOI: 10.1145/3725356

Incoming Non-self Citations Over Time

Authors

1. Saehan Jo (Cornell University)
2. Immanuel Trummer (Cornell University)

BibTeX Citation

@inproceedings{jo_sigmod25,
        title = {{SpareLLM: Automatically Selecting Task-Specific Minimum-Cost Large Language Models under Equivalence Constraint}},
        author = {Jo, Saehan and Trummer, Immanuel},
        series = {{SIGMOD} '25},
        booktitle = {Proceedings of the {ACM} {SIGMOD} International Conference on Management of Data},
        publisher = {Association for Computing Machinery},
        doi = {10.1145/3725356},
        url = {https://dl.acm.org/doi/10.1145/3725356},
        year = {2025}
}

Incoming Citations (Sorted by Pagerank)

Showing 3 of 3 citing papers.

Rank	Citing Paper	Year	Venue	Pagerank
10,432	Beluga: A CXL-Based Memory Architecture for Scalable and Efficient LLM KVCache Management	2026	SIGMOD	5.093636e-05
10,483	PRISM: Navigating Cost–Accuracy Trade-offs for NL2SQL	2026	SIGMOD	5.093636e-05
10,504	Task Cascades for Efficient Unstructured Data Processing	2026	SIGMOD	5.093636e-05

Outgoing Citations (Sorted by Pagerank)

Showing 7 of 7 cited papers.

Citations counted here include only citations to other VLDB/SIGMOD/CIDR/PODS papers in this database.

Rank	Cited Paper	Year	Venue	Pagerank
772	VerdictDB: Universalizing Approximate Query Processing	2018	SIGMOD	0.00014147905
819	Quickr: Lazily Approximating Complex AdHoc Queries in BigData Clusters	2016	SIGMOD	0.00013815639
1,108	Approximate Query Processing: No Silver Bullet	2017	SIGMOD	0.00012145154
1,227	Blink and It's Done: Interactive Queries on Very Large Data	2012	VLDB	0.00011582387
1,634	Rapid Sampling for Visualizations with Ordering Guarantees	2015	VLDB	0.00010163938
2,413	A Sampling Algebra for Aggregate Estimation	2013	VLDB	8.6116764e-05
11,749	BitGourmet: Deterministic Approximation via Optimized Bit Selection	2020	CIDR	5.093636e-05

Semantically Similar Papers

#	Overall Rank	Paper	Year	Venue
1	10,504	Task Cascades for Efficient Unstructured Data Processing	2026	SIGMOD
2	7,075	Flash-LLM: Enabling Cost-Effective and Highly-Efficient Large Generative Model Inference with Unstructured Sparsity	2024	VLDB
3	713	Language Models Enable Simple Systems for Generating Structured Views of Heterogeneous Data Lakes	2024	VLDB
4	10,527	BRIEF: Bi-level Coreset Selection for Efficient Instruction Tuning in LLMs	2026	VLDB
5	13,343	Database Perspective on LLM Inference Systems	2025	VLDB
6	10,276	OctoSelector: Efficient and Effective Batch-Aware Model Selection for Large Language Models	2026	SIGMOD
7	10,856	Optimized Batch Prompting for Cost-effective LLMs	2025	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	8,627	ThriftLLM: On Cost-Effective Selection of Large Language Models for Classification Queries	2025	VLDB