Categories Companies Research

Loading the SOTA2 catalog…

Explore

Categories Companies Research

Contribute

Join SOTA2 Contact

SOTA2

© 2026 SOTA2

Build in Bellevue, USA

WizWand is now SOTA2 Research

Papers Benchmarks Datasets Tasks

Reward Modeling research benchmarks – Page 3 · SOTA2 Research

Reward Modeling

Benchmarks

Dataset NameSOTA MethodSort

Dataset NameSOTA methodMetricTrendResultsLast Updated

88.1Ubuntu Accuracy

24

Mar 20, 2026

RoboRewardBench

RoboReward (8B)

0.665MAE

24

Jul 7, 2026

DeepSeek-BTRM-27B

68.3Accuracy

23

Feb 26, 2026

Skywork-Reward-V2-Llama-3.1-8B-40M

89.3Score

22

Mar 4, 2026

R1-Distill-Qwen-32B

63.89Accuracy

22

Feb 26, 2026

65.29Accuracy

22

Feb 26, 2026

gpt-5-2025-08-07

75.95Accuracy

22

Feb 26, 2026

82.88Accuracy

22

Feb 26, 2026

Reward Bench V2

gpt-5-2025-08-07

83.44Accuracy

22

Feb 26, 2026

Flexible Principles GenRM

74.6Knowledge

22

May 19, 2026

93.4L-Acc

20

Mar 5, 2026

RewardBench unified-feedback (test)

Starling-RM-34B

84Average Score

20

Feb 26, 2026

PPE Preference ZH

82.3Accuracy

19

Feb 26, 2026

UltraFeedback core250 (held-out evaluation)

3.543Delta (Δ)

18

May 12, 2026

74Pairwise Accuracy

18

Mar 26, 2026

LMArena-like Benchmarks PPE Pref ZH, In-House QA, In-House Writing v1.0

82.3PPE Pref ZH Score

18

Feb 26, 2026

Gemini-2.5-Flash

0.75Accuracy

18

Feb 26, 2026

Gemini-2.5-Flash

69.3Accuracy

17

Feb 26, 2026

Rubric-ARM-voting@5

87.7Accuracy

17

Feb 26, 2026

RM-R1-14B (DeepSeek-Dist)

89.9Accuracy

17

Feb 26, 2026

Reward Bench Prior Sets

Cohere May 2024

78.2Prior Sets Score

17

Feb 26, 2026

Skill-RM (Qwen3.5-122B-A10B)

85.2Reward Modeling Score

16

Jun 4, 2026

84.9Positional Consistency

16

Apr 10, 2026

70.3Single-turn Score

16

Jun 26, 2026

REWARDAGENT_MINI

78Accuracy

16

Feb 26, 2026

Page 3 of 10

Popular datasets

RewardBench RM-Bench UltraFeedback PersonalLLM PPE

Follow for updates

Join SOTA2 Contact