List your company

Sequential Preference Optimization on HelpSteer2

99.86Harmless RateMORA
17.003238.514160.02581.5359May 12, 2026

Evaluation results

MethodLinks
MORABackbone=LLAMA-32026.0599.8672.5471.0181.13
OrthAlignBackbone=LLAMA-32026.0591.3468.8367.6975.95
MORABackbone=MISTRAL2026.059082.1166.0179.37
OrthAlignBackbone=MISTRAL2026.0588.1267.0865.3473.51
KnotsBackbone=LLAMA-32026.0587.1159.6263.6670.13
SPOBackbone=MISTRAL2026.0584.0365.2156.2568.49
TSV-MBackbone=LLAMA-32026.0578.0762.1262.3667.51
RSDPOBackbone=LLAMA-32026.0575.5770.863.469.92
KnotsBackbone=MISTRAL2026.0575.2861.6664.467.11
TSV-MBackbone=MISTRAL2026.0573.364.1765.2867.58
MODPOBackbone=MISTRAL2026.0571.3661.5559.3264.07
MODPOBackbone=LLAMA-32026.0570.9669.4462.0867.49
RSDPOBackbone=MISTRAL2026.0566.3268.9564.3566.54
SPOBackbone=LLAMA-32026.0566.1568.2462.0165.46
SoupsBackbone=LLAMA-32026.0556.9261.1858.9359.01
SoupsBackbone=MISTRAL2026.0554.0359.7559.9257.9
DPO BaselineBackbone=LLAMA-32026.0551.9272.9166.563.78
SFTBackbone=LLAMA-32026.0546.735053.4150.04
DPO BaselineBackbone=MISTRAL2026.0539.2373.1666.0159.46
SFTBackbone=MISTRAL2026.0520.1926.8343.0330.01