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

Reward Modeling on RewardBench

99.4Chat Score

FsfairX-LLaMA3-RM-8B

Updated 2d ago

Evaluation Results

Method
FsfairX-LLaMA3-RM-8B 2024.06	99.4	84.7	65.1	87.8	86.4	-
InternLM2-7B-Reward 2025.01	99.2	87.6	69.5	87.2	94.5	-
FsfairX-LLaMA3-RM-v0.1 2024.06	99.16	85.34	64.69	86.89	90.64	-
Single Head 2025.02	98.9	87.1	68.4	89.1	92	-
InternLM2-20B-Reward 2025.01	98.9	90.2	76.5	89.5	95.8	-
Shared-Base 2025.02	98.6	87.3	68.4	89.5	92.7	-
DRMs(Ours) 2025.02	98.6	88.5	75.5	88.5	91.4	-
GRM 2024.06	98.6	87	67.8	89.4	92.3	-
Random (Uniform) 2025.02	98.5	85.7	62.3	90.3	91.5	-
RLHFlow-Llama 3 8B 2024.06	98.3	87.1	65.8	89.7	94.7	-
PairRM-LLaMA3-8B 2024.06	98.3	87.13	65.8	89.7	94.7	-
Random (Gaussian) 2025.02	98.2	86.8	66.3	88.9	93.6	-
Eurus-RM-7B 2024.06	98.04	83.01	62.72	81.89	89.38	-
Eurus RM Mistral 7B 2024.06	98	82.8	65.6	81.2	86.3	-
Eurus-RM-7B 2026.01	98	81.6	65.6	81.4	86.3	-
GRM w/ dpo noref 2024.06	97.8	78.6	54.6	82	79.9	-
Llama 3 70B Instruct 2024.06	97.6	76	58.9	69.2	78.5	-
UltraMedRM-8B 2024.06	97.21	85.53	67.11	91.19	86.62	-
RRM 2026.04	97.21	84.38	67.98	80.95	91.4	85.93
Classifier + margin 2024.06	97.2	66.1	37.5	56.8	72.7	-
GRM w/ sft 2024.06	97.2	78.4	54.2	83.6	78.6	-
Llama-3.1-70B-Instruct 2026.01	97.2	84.1	70.2	82.8	86	-
Llama3.1-70B-Instruct 2026.04	97.2	-	70.2	82.8	86	84
Llama3.1-405B-Instruct 2026.04	97.2	-	74.6	77.6	87.1	84.1
VRM 2026.03	97.11	-	94.23	85.41	89.72	-
CLoud 2026.04	97	-	58	84	92	82.8
ULTRAFEEDBACK 2024.10	96.9	79.5	61.8	79.2	80.2	-
PBAA_DPO 2024.10	96.9	78.9	59.9	77.2	81.6	-
ArmoRM-Llama 3 8B 2024.06	96.9	90.8	76.8	92.2	97.3	-
Starling RM Yi 34B 2024.06	96.9	82.7	57.2	88.2	88.5	-
Starling-RM-34B 2024.06	96.9	82.7	57.2	88.2	88.5	-
GRM (linear) w/ dpo noref 2024.06	96.9	77.8	52.9	82.7	78.8	-
LLaVA-Critic-8B 2025.01	96.9	80	52.8	81.7	83.5	-
Self-taught-evaluator-llama3.1-70B 2026.04	96.9	-	85.1	89.6	88.4	90
GRM (linear) w/ sft 2024.06	96.7	78.3	52.4	81.5	82.5	-
GPT-4o-0513 2024.06	96.6	84.7	70.4	86.7	84.9	-
INF-ORM-Llama3.1-70B 2025.01	96.6	95.1	91	93.6	99.1	-
QWen2-VL-7B 2025.01	96.6	83.8	57	73.9	94.3	-
Cohere May 2024 2024.06	96.4	89.5	71.3	92.7	97.7	-
UltraRM-13b 2024.06	96.4	67.58	55.5	56	62.4	-
Claude-3.5-Sonnet-20240620 2026.01	96.4	84.2	74	81.6	84.7	-
Qwen3-14B 2026.01	96.4	83.9	69.2	88	82	-
Cohere-0514 2026.04	96.4	-	71.3	92.3	97.7	89.4
Claude-3-5-sonnet-20240620 2026.04	96.4	-	74	81.6	84.7	84.2
Vanilla RM 2026.04	96.37	83.09	63.37	89.73	82.88	83.22
Classifier + Ensemble 2024.06	96.1	65.2	38.2	58.8	67.6	-
GPT-4o 2026.01	96.1	82.4	70.7	87.8	75.1	-
SynRM 2026.01	96.1	75.4	47.6	78.7	79.1	-
GPT-4o-0806 2026.04	96.1	-	76.1	86.6	88.1	86.7
DPO 2026.03	96.09	-	54.17	67.84	57.18	-
Nemotron-4 340B RM (w. HelpSteer2) 2024.06	95.8	92	87.1	91.5	93.7	-
Classifier (baseline) 2024.06	95.8	64.5	37.3	59.9	64.8	-
Skyword-Reward-Llama3.1-8B 2025.01	95.8	92.5	87.3	90.8	96.2	-
BT 2026.02	95.8	64.5	37.3	59.9	64.8	-
GRM-BNRM 2026.02	95.7	71.8	41.6	81.5	68.4	-
BT-BNRM 2026.02	95.6	72.5	43.3	80.9	70.1	-
RM-Distiller 2026.01	95.5	79.5	51.1	84.1	87.4	-
Qwen-Instruct-32B-Ours 2026.04	95.4	-	83.3	92	95.4	91.5
SSO_DPO 2024.10	95.3	80	59	77.4	88.3	-
DRMs(Ours) 2025.02	95.3	81.4	65	78.3	87.2	-
GPT-4-0125-preview 2024.06	95.3	85.9	74.3	87.2	86.9	-
GPT-4-0613 2024.06	95.3	85.1	75.4	87.1	82.7	-
gpt-4-0125-preview 2024.06	95.3	85.9	74.3	87.2	86.9	-
gpt-4-turbo-2024-04-09 2024.06	95.3	85.1	75.4	87.1	82.7	-
KTO 2026.03	95.25	-	43.86	64.86	54.74	-
Random (Gaussian) 2025.02	95.1	78.6	57.3	78.1	83.9	-
RM-Distiller 2026.01	95	80.2	55	81.8	89	-
SKYWORK + ULTRAFEEDBACK 2024.10	94.7	85.1	72.7	89.1	83.9	-
Shared-Base 2025.02	94.7	74	47.6	76.5	77.4	-
Cohere March 2024 2024.06	94.7	87.1	65.1	90.3	98.2	-
Claude-3-Opus-02292024 2024.06	94.7	80.7	60.3	89.1	78.7	-
GRM (linear) w/ dpo noref 2024.06	94.7	61.7	38.4	62.5	51.2	-
GRM (linear) w/ sft 2024.06	94.7	69.5	40.8	65.4	77	-
Margin BT 2026.01	94.7	72.8	46.3	79.9	70.4	-
CARP 2026.04	94.69	85.42	68.86	88.24	89.87	86.83
FR RM 2025.05	94.41	78.38	57.02	78.24	83.86	-
FC RM 2025.05	94.41	77.5	53.29	76.76	85.53	-
Single Head 2025.02	94.4	73.3	46.7	75.9	75.9	-
Llama 3 70B (w. HH-RLHF) 2024.06	94.4	73.9	54.6	81.2	65.6	-
SteerLM 2026.01	94.4	74.5	46.5	74.3	82.8	-
Qwen-Instruct-7B-Ours 2026.04	94.2	-	74.8	85.3	87	85.3
GRM w/ sft 2024.06	94.1	66.8	41.9	69.5	61.5	-
Classifier + label smooth 2024.06	94.1	72.1	47.1	67.5	79.7	-
LSAM 2026.01	94.1	71.6	37.3	73.5	81.4	-
DeepSeek-GRM-27B 2026.04	94.1	-	78.3	88	83.8	86
RM 2026.03	94.04	-	93.02	76.66	86.62	-
Random (Uniform) 2025.02	94	78.7	56.7	80	84.3	-
SKYWORK + SSO_DPO 2024.10	93.9	86.3	75.4	86.6	89.3	-
Qwen-Instruct-14B-Ours 2026.04	93.8	-	80.6	87.2	92.1	88.4
Margin BT 2026.01	93.6	70.4	33.6	73.5	80.8	-
CLoud 2026.01	93.6	64.2	37.1	62.6	63.6	-
SteerLM 2026.01	93.6	74	45	79.1	78.5	-
SKYWORK + PBAA_DPO 2024.10	93.3	83.5	75.1	86.5	79	-
Llama 3 70B (w. HelpSteer) 2024.06	93.3	66.1	59.7	56.8	54.9	-
BT Classifier 2026.01	93.3	70.3	37.7	69.3	81	-
BT RM 2025.05	93.02	78.11	57.02	77.43	84.98	-
Prometheus 2 Mistral 8x7B 2024.06	93	75.3	47.1	83.5	77.4	-
LSAM 2026.01	93	79.3	69.7	84.1	70.4	-
Prometheus-8*7B-v2 2026.04	93	-	47.1	80.5	77.4	74.5
JudgeLRM-7B 2026.01	92.9	75.2	56.4	78.2	73.6	-

Showing 100 of 154 rows