Democratic Preference Alignment on Llama 3.2 3B raw scores (test)
100Exact MatchKemeny-Young
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| Kemeny-YoungConsensus ranking=Hard ≻ US-Rep ≻ Soft ≻ Full ≻ Base, bootstrap n=10002026.02 | 100 | — | — | |
| Mallows (Kendall)Consensus ranking=π0 matches Kemeny, bootstrap n=10002026.02 | — | 0.163 | -14,217.8 |