PoQ-Judge: A Multi-Architecture Evaluation Framework for Cost-Aware Proof-of-Quality in Decentralized LLM Inference

In three linesPoQ-Judge is a reference-free evaluation framework for decentralized LLM inference networks. Three architectures (TextCNN, MiniLM, DeBERTa) are trained on UltraFeedback plus GPT-labeled in-domain data. The best model achieves 0.747 Pearson correlation with ground-truth, outperforming reference-based evaluators. Cascade evaluation reduces costs by 72.7% with modest quality loss.

Summary generated by Claude — human-verified