Rohan Paul (@rohanpaul_ai)T2·78 pts
- All eight judges favored their own answers, averaging 1.22 rank positions and nearly 3 for GPT-5.6 Sol
- Anonymity and rubric removal did not eliminate the bias; revealing model names cut it by only 0.094 positions
- Use multi-family judge panels and report per-judge results rather than one pooled score