QUBOBench
QUBOBench is an evaluation benchmark designed specifically for the task of automatically modeling natural-language problem descriptions as QUBO formulations. It contains 100 combinatorial optimization problems spanning 12 application domains, with problems sourced from academic literature, competitions, and classic NP-hard problems.
Timeline (last 90 days)
The researchers constructed the QUBOBench benchmark, comprising 100 combinatorial optimization problems spanning 12 application domains.
The problems in QUBOBench are sourced from peer-reviewed academic literature, various competitions, and classic NP-hard problems.
The framework achieves 68% accuracy on QUBOBench, surpassing the direct single-call baseline method by 22 percentage points.
All Facts (3)
The researchers constructed the QUBOBench benchmark, comprising 100 combinatorial optimization problems spanning 12 application domains.
50%UnverifiedThe problems in QUBOBench are sourced from peer-reviewed academic literature, various competitions, and classic NP-hard problems.
50%UnverifiedThe framework achieves 68% accuracy on QUBOBench, surpassing the direct single-call baseline method by 22 percentage points.
50%