Consistent and Distinctive: LLM Benchmark Efficiency via Maximum Independent Set Prompt Selection on Similarity Graphs
This work addresses the computational cost of evaluating large language models by enabling efficient benchmark subset selection without sacrificing ranking consistency.
The paper proposes a graph-based prompt selection framework using Maximum Independent Set to reduce benchmark size for LLM evaluation, achieving 25-48% prompt reduction while maintaining highly consistent LLM rankings (Kendall's W ≥ 0.90 in 99.2% of configurations).
Evaluating large language models (LLMs) across comprehensive benchmarks is expensive and time-consuming. We propose a graph-based prompt selection framework that models each benchmark as a similarity graph -- nodes are prompts connected if their embedding-space distance falls above a configurable threshold -- and applies Maximum Independent Set (MIS) algorithms to select a maximally diverse, non-redundant subset. We evaluate four MIS solvers (CPLEX, GREEDY, Online-MIS, ReduMIS) across six embedding models, three distance measures, six percentile thresholds, and four benchmarks (GPQA, IFEval, MMLU-Pro, Omni-MATH) covering 66 LLMs. Our central hypothesis -- that repeated selection under different random seeds yields consistent LLM rankings that may also differ from the full-benchmark baseline -- is strongly confirmed: Kendall's $W \geq 0.90$ in 99.2\% of stochastic configurations (mean $W = 0.997 \pm 0.008$), while at higher percentile thresholds selected subsets achieve 25--48\% prompt reduction on average. Ranking divergence from the full benchmark ($ρ< 0.95$) occurs in only 15.95\% of configurations, concentrated at low thresholds ($p_{10}$--$p_{20}$) and benchmarks (GPQA, IFEval), identifying overly dense graphs as the primary failure mode.