Share your thoughts, 1 month free Claude Pro on usSee more

Reward Modeling on RewardBench unified-feedback (test)

84Average Score

Starling-RM-34B

Updated 1mo ago

Evaluation Results

Method	Links
Starling-RM-34B 2024.06		84	96.9	59	89.9	90.3
GRM (linear) w/ sft 2024.06		79.5	97.8	54.6	86.3	79.2
GRM (linear) w/ dpo noref 2024.06		78.3	98	53.3	86.4	75.3
GRM w/ dpo noref 2024.06		78	97.8	54	85.7	74.4
GRM w/ sft 2024.06		77.6	98	55.3	85.8	71.2
Classifier + Ensemble 2024.06		76.6	96.6	51.8	85.1	73
Classifier (baseline) 2024.06		76.3	96.6	52.4	86.7	69.5
Classifier + label smooth 2024.06		76.3	97.2	49.8	85.8	72.3
Starling-RM-7B 2024.06		76.2	98	43.4	88.6	74.6
Classifier + margin 2024.06		74.5	96.4	51.5	85.3	64.8
GRM (linear) w/ sft 2024.06		71.5	96.1	40.1	80.3	69.3
Classifier + Ensemble 2024.06		71	98	37.5	77.3	71.3
GRM w/ sft 2024.06		70.8	97.8	42.1	77.9	65.2
Classifier + label smooth 2024.06		70.6	94.4	37.3	73.2	77.4
Classifier + margin 2024.06		70.2	95.8	38.4	73.9	72.5
GRM (linear) w/ dpo noref 2024.06		70.2	96.7	39	76.4	68.5
GRM w/ dpo noref 2024.06		70.2	95.8	40.1	78.7	66.2
UltraRM-13B 2024.06		69.8	96.1	55.3	45.8	82
Classifier (baseline) 2024.06		68.2	95.5	38	73.8	65.3
PairRM 2024.06		58.7	90.2	53	31.5	60