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 5 · SOTA2 Research

Reward Modeling

Benchmarks

Dataset NameSOTA MethodSort

Dataset NameSOTA methodMetricTrendResultsLast Updated

84.33Pairwise Accuracy

13

Feb 26, 2026

89.17Pairwise Accuracy

13

Feb 26, 2026

88.43Pairwise Accuracy

13

Feb 26, 2026

97.07Pairwise Acc

13

Feb 26, 2026

RBM ID (Refined)

88VOC r Score

12

Jul 7, 2026

HH-RLHF helpful core250 (held-out evaluation)

20.155Reward Score

12

May 12, 2026

Aggregated Benchmarks Macro

Claude 4.6 Opus

74.74Average Score (excl. MM-RB, VL-RB)

12

Apr 15, 2026

Overall 5-Benchmark Suite

ConsistRM (Qwen3-8B)

73.5Average Score

12

Apr 10, 2026

AgentRewardBench

83.8Precision

12

Mar 20, 2026

Anthropic Harmless

74.7Accuracy

12

Feb 26, 2026

Anthropic Helpful

72Accuracy

12

Feb 26, 2026

InternVL3.5-38B

79Score (Easy)

11

Mar 13, 2026

FIRM-Bench Edit

0.53Execution MAE

11

Mar 13, 2026

ExeVR-Bench Overall

84.7Accuracy

11

Mar 12, 2026

ExeVR-Bench Android

83.5Accuracy

11

Mar 12, 2026

ExeVR-Bench Mac/Win

90Accuracy

11

Mar 12, 2026

ExeVR-Bench Ubuntu Human

84Accuracy

11

Mar 12, 2026

ExeVR-Bench Ubuntu Agent

85.1Accuracy

11

Mar 12, 2026

RewardBench latest (full)

BNBT-Reward-Llama-3.1-8B

93.6Average Score

11

Feb 26, 2026

RM-Bench (full)

83Chat Score

11

Feb 26, 2026

PMDC Maximum Discrepancy samples

Skywork-Reward-Gemma-2-27B

1Rank

10

Feb 26, 2026

64.6Accuracy

10

Feb 26, 2026

RM-Bench Normal

INF-ORM-Llama3.1-70B

80Accuracy

10

Feb 26, 2026

Llama-3.1-Nemotron-70B

92.2Accuracy

10

Feb 26, 2026

81.2User Friendliness

10

Feb 26, 2026

Page 5 of 10

Popular datasets

RewardBench RM-Bench UltraFeedback PersonalLLM PPE

Follow for updates

Join SOTA2 Contact