Multi-objective Evolutionary Merging Enables Efficient Reasoning Models
Mario Iacobelli · Adrian Robert Minut · Tommaso Mencattini · Donato Crisostomi · Andrea Santilli · Iacopo Masi · Emanuele Rodolà
Reasoning models excel at complex tasks but are token-heavy, while standard models are efficient but less accurate. We use evolutionary model merging to reduce reasoning length by 50% without compromising accuracy.
摘要
Reasoning models achieve strong performance on complex problems by leveraging long chains of thought, but this deliberate reasoning incurs substantial inference-time cost. The Long-to-Short (L2S) reasoning problem seeks to preserve accuracy while reducing generated tokens. Yet, current training-free model merging approaches rely on brittle, fixed-hyperparameter arithmetic methods that force suboptimal compromises. We introduce \textbf{Evo-L2S}, a multi-objective evolutionary model merging framework that explicitly optimizes accuracy and output length to recover a Pareto front of merged models. To make this search computationally tractable, we propose an entropy-based subset sampling technique that substantially reduces fitness-estimation overhead. Across six mathematical reasoning benchmarks, Evo-L2S reduces reasoning length by over 50\% at the 1.5B and 7B scales while preserving or improving problem-solving accuracy; at 14B, the steeper Pareto front reveals that attainable compression depends on the intrinsic compressibility of the reasoning model. Overall, Evo-L2S shows that reasoning models can be made substantially more concise while preserving strong problem-solving performance.