List your company

Domain-specific Alignment on IID prompt set (held-out)

-1.94AvgRDEPO
-5.684-4.712-3.74-2.768May 6, 2026

Evaluation results

MethodLinks
DEPOIteration=32026.05-1.9474.1
POPOIteration=32026.05-2.0273.2
XPOIteration=32026.05-2.1172.9
DPOIteration=32026.05-2.3672.5
DEPOIteration=22026.05-3.0168.9
POPOIteration=22026.05-3.1568.2
XPOIteration=22026.05-3.2867.3
DPOIteration=22026.05-3.5966.6
DEPOIteration=12026.05-4.362.6
POPOIteration=12026.05-4.3262.5
XPOIteration=12026.05-4.462.6
DPOIteration=12026.05-4.562.4
Llama-3-8B-SFT2026.05-5.54—