List your company
ResearchFormerly WizWand

Instruction Following on ArenaHard Hard Prompts 2.0

32.7Win RateRLLM
1.9169.90817.925.892Mar 19, 2026
Updated 7mo ago

Evaluation results

MethodLinks
RLLMRM=Qwen3-8B (prompted), RM Type=Generative, RM Size=8B, Training samples=non-verifiable WildChat, Evaluation mode=thinking mode2026.0332.7
RLHFRM=Skywork-Reward-V2-Llama-3.1-8B, RM Type=Scalar, RM Size=8B, Training samples=non-verifiable WildChat, Evaluation mode=thinking mode2026.0326.4
RLHFRM=Nexusflow/Athene-RM-8B, RM Type=Scalar, RM Size=8B, Training samples=non-verifiable WildChat, Evaluation mode=thinking mode2026.0325.4
Qwen3-8BRM=–, RM Type=–, RM Size=–, Training samples=non-verifiable WildChat, Evaluation mode=thinking mode2026.0322.4
RLLMThinking Mode=Thinking, RM=Qwen3-1.7B, RM Type=Generative, RM Size=1.7B, Evaluator=GPT-4.12026.038.5
Qwen3-1.7BThinking Mode=Thinking, RM=–, RM Type=–, RM Size=–, Evaluator=GPT-4.12026.036.4
RLHFThinking Mode=Thinking, RM=Skywork-Reward-V2-Llama-3.1-8B, RM Type=Scalar, RM Size=8B, Evaluator=GPT-4.12026.036
RLLMThinking Mode=Non-thinking, RM=Qwen3-1.7B, RM Type=Generative, RM Size=1.7B, Evaluator=GPT-4.12026.036
RLHFThinking Mode=Thinking, RM=Nexusflow/Athene-RM-8B, RM Type=Scalar, RM Size=8B, Evaluator=GPT-4.12026.034.3
RLHFThinking Mode=Non-thinking, RM=Nexusflow/Athene-RM-8B, RM Type=Scalar, RM Size=8B, Evaluator=GPT-4.12026.033.9
RLHFThinking Mode=Non-thinking, RM=Skywork-Reward-V2-Llama-3.1-8B, RM Type=Scalar, RM Size=8B, Evaluator=GPT-4.12026.033.6
Qwen3-1.7BThinking Mode=Non-thinking, RM=–, RM Type=–, RM Size=–, Evaluator=GPT-4.12026.033.1