#math reasoning benchmarks

Prefix-RFT blends supervised and reinforcement fine-tuning by using partial demonstration prefixes to guide exploration, achieving stronger and more stable performance on math reasoning benchmarks than SFT, RFT, and hybrid baselines.