Reward Modeling on PPE Correctness

71.2Accuracy

SAVE

Updated 1mo ago

Evaluation Results

Method	Links
SAVE 2026.05		71.2
SAVE (w/o Curriculum Mechanism) 2026.05		71.2
SAVE 2026.05		71.1
SAVE (w/o Policy Model Optimization) 2026.05		71.1
Continual Offline Training RM 2026.05		71
SAVE (w/o Curriculum Mechanism) 2026.05		71
SAVE (w/o Policy Model Optimization) 2026.05		70.9
Skywork-Reward-V2-Llama-3.2-3B 2026.05		70.7
HL-BT 2026.05		70.5
HL-BT 2026.05		70.5
OPRM-RgFT-Qwen2.5-32B 2026.02		67.3
DeepSeek-BTRM-27B 2026.02		66.7
Mean Reward 2026.05		66.6
OPRM-RgFT-Qwen2.5-72B 2026.02		66.4
OPRM-Qwen2.5-32B 2026.02		66.1
Mean Reward 2026.05		65.9
WILDREWARD-8B 2026.02		65.6
OPRM-RgFT-Qwen2.5-14B 2026.02		65.6
DeepSeek-PairRM-27B 2026.02		64.8
INF-ORM-Llama3.1-70B 2026.02		64.4
OPRM-Qwen2.5-14B 2026.02		64.3
OPRM-Qwen2.5-72B 2026.02		64.3
Llama-3-OffsetBias-RM-8B 2026.02		64.1
WILDREWARD-4B 2026.02		63.6
Llama-3.1-Nemotron-70B 2026.02		63.2
Internlm2-20b-reward 2026.02		63
InternLM2-20B-Reward 2026.02		63
Skywork-Reward-Llama-3.1-8B-v0.2 2026.02		62.5
CLoud-Gemma-2-27B 2026.02		62.4
OPRM-RgFT-Qwen2.5-7B 2026.02		62.4
Athene-RM-8B 2026.02		62
Skywork-Reward-Gemma-2-27B-v0.2 2026.02		61.9
OPRM-Qwen2.5-7B 2026.02		61.3
ArmoRM-8B-v0.1 2026.02		61.2
Nemotron-4-340B-Reward 2026.02		60.8
ArmoRM-Llama3-8B-v0.1 2026.02		60.6
Gemini-1.5-Pro 2026.02		59.8
DeepSeek-GRM-27B 2026.02		59.8
DeepSeek-GRM-27B-RFT 2026.02		59.6
LLaMA-3.1-70b-Instruct 2026.02		59.2
Claude-3.5-sonnet 2026.02		58.8
LLM-as-a-Judge 2026.02		58.8
DeepSeek-V2.5-0905 2026.02		58.5
GPT-4o 2026.02		57.6
Skywork-Reward-Gemma-2-27B 2026.02		56.6