DGrid AI(@dgrid_ai):2/📊 Key Results (held-out test set, n=300) ✅DeBERTa Judge: Pearson 0.747 (95% CI [0.663, 0.816]) → Outperforms all reference-based evaluators in our prior framework (best: 0.629) ✅Reference-Free composite score: Pearson 0.645→ Matches the best reference-based single evaluator — with zero reference answers ✅Cascade + online weight calibration: Saves 72.7% evaluation cost

DGrid AI

@dgrid_ai

💡 DGrid AI is a Decentralized AI Smart Network for Cost-effective, Reliable & Verifiable AI. One API, ALL LLMs. 🌟TG:

Joined March 2024

121 Following 143.5K Followers

DGrid AI@dgrid_ai

2026.06.15 08:54

2/📊 Key Results (held-out test set, n=300) ✅DeBERTa Judge: Pearson 0.747 (95% CI [0.663, 0.816]) → Outperforms all reference-based evaluators in our prior framework (best: 0.629) ✅Reference-Free composite score: Pearson 0.645→ Matches the best reference-based single evaluator — with zero reference answers ✅Cascade + online weight calibration: Saves 72.7% evaluation cost

172

109

Forward to community

Most Popular Users