ThriftLLM: On Cost-Effective Selection of Large Language Models for Classification Queries

Summary: ThriftLLM formulates budget-constrained LLM ensemble selection for classification as maximizing correctness probability—non-monotone? nondecreasing but nonsubmodular and likely NP-hard—using a submodular upper bound for high-probability guarantees. Demonstrates cost-effective gains on classification and entity matching. (summarized by gpt-5.6-luna on Jul 24 2026)

Paper ID: 14243
Venue: VLDB
Year: 2025
Pagerank: 5.3969543e-05
Overall Rank: 8,627 | 40.82%
DOI: 10.14778/3749646.3749702

Incoming Non-self Citations Over Time

Authors

1. Keke Huang (University of British Columbia)
2. Yimin Shi (National University of Singapore)
3. Dujian Ding (University of British Columbia)
4. Yifei Li (University of British Columbia)
5. Yang Fei (National University of Singapore)
6. Laks Lakshmanan (University of British Columbia)
7. Xiaokui Xiao (CNRS; National University of Singapore)

BibTeX Citation

@article{huang_vldb25,
        title = {{ThriftLLM: On Cost-Effective Selection of Large Language Models for Classification Queries}},
        author = {Huang, Keke and Shi, Yimin and Ding, Dujian and Li, Yifei and Fei, Yang and Lakshmanan, Laks and Xiao, Xiaokui},
        journal = {PVLDB},
        series = {{VLDB} '25},
        volume = {18},
        number = {11},
        pages = {4410--4423},
        doi = {10.14778/3749646.3749702},
        url = {https://doi.org/10.14778/3749646.3749702},
        year = {2025}
}

Incoming Citations (Sorted by Pagerank)

Showing 3 of 3 citing papers.

Rank	Citing Paper	Year	Venue	Pagerank
8,829	Cut Costs, Not Accuracy: LLM-Powered Data Processing with Guarantees	2026	SIGMOD	5.3613784e-05
10,253	HAMMER: An Automatic RAG Tuning System via Hierarchical Memory-Guided Monte Carlo Tree Search	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 16 of 16 cited papers.

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

Rank	Cited Paper	Year	Venue	Pagerank
141	Deep Entity Matching with Pre-Trained Language Models	2021	VLDB	0.0002964847
176	Deep Learning for Entity Matching: A Design Space Exploration	2018	SIGMOD	0.00027191081
194	Influence Maximization: Near-Optimal Time Complexity Meets Practical Efficiency	2014	SIGMOD	0.000259047
279	Text-to-SQL Empowered by Large Language Models: A Benchmark Evaluation	2024	VLDB	0.00022468369
489	Distributed Representations of Tuples for Entity Resolution	2018	VLDB	0.0001761456
529	Magellan: Toward Building Entity Matching Management Systems	2016	VLDB	0.00017096361
626	Entity Resolution: Theory, Practice & Open Challenges	2012	VLDB	0.00015656958
962	DBSCAN Revisited: Mis-Claim, Un-Fixability, and Approximation	2015	SIGMOD	0.00012936472
1,402	Creating Embeddings of Heterogeneous Relational Datasets for Data Integration Tasks	2020	SIGMOD	0.00010888094
3,787	Combining Small Language Models and Large Language Models for Zero-Shot NL2SQL	2024	VLDB	7.1249098e-05
4,251	FinSQL: Model-Agnostic LLMs-based Text-to-SQL Framework for Financial Analysis	2024	SIGMOD	6.8031106e-05
4,308	AutoTQA: Towards Autonomous Tabular Question Answering through Multi-Agent Large Language Models	2024	VLDB	6.7682891e-05
4,590	Entity Resolution with Hierarchical Graph Attention Networks	2022	SIGMOD	6.6150054e-05
7,483	Are Large Language Models a Good Replacement of Taxonomies?	2024	VLDB	5.6074551e-05
7,865	Generating Succinct Descriptions of Database Schemata for Cost-Efficient Prompting of Large Language Models	2024	VLDB	5.5282752e-05
8,343	An Interactive Multi-modal Query Answering System with Retrieval-Augmented Large Language Models	2024	VLDB	5.448279e-05

Semantically Similar Papers

#	Overall Rank	Paper	Year	Venue
1	10,290	Skyline Retrieval meets Set-Cover Chunk Merging: A Cost-Effective RAG-Sketch for Long-Context LLM QA	2026	SIGMOD
2	10,504	Task Cascades for Efficient Unstructured Data Processing	2026	SIGMOD
3	9,972	Cents: A Flexible and Cost-Effective Framework for LLM-Based Table Understanding	2025	VLDB
4	4,050	Revisiting Prompt Engineering via Declarative Crowdsourcing	2024	CIDR
5	10,733	ScaleLLM: A Technique for Scalable LLM-augmented Data Systems	2025	SIGMOD
6	10,318	In-context Clustering-based Entity Resolution with Large Language Models: A Design Space Exploration	2026	SIGMOD
7	10,276	OctoSelector: Efficient and Effective Batch-Aware Model Selection for Large Language Models	2026	SIGMOD
8	10,856	Optimized Batch Prompting for Cost-effective LLMs	2025	VLDB
9	7,316	SpareLLM: Automatically Selecting Task-Specific Minimum-Cost Large Language Models under Equivalence Constraint	2025	SIGMOD
10	8,829	Cut Costs, Not Accuracy: LLM-Powered Data Processing with Guarantees	2026	SIGMOD