harshatheg/Qwen-2.5-1B-RLCD

This repository provides a fine-tuned Qwen-2.5-1B model for structured extraction using Parallel Constrained Decoding on Apple Silicon. The card documents 5.6x to 7.0x latency reductions over autoregressive baselines, claiming 100% schema validity. While the inference method offers speed for categorical tasks, the model itself is a derivative of an existing base, lacking independent benchmark verification.