List your company

Reward Model Transfer on HelpSteer3 (H3) v1 (test)

8.96AOGANCHOR
-6.6088-2.56691.4755.5169May 25, 2026

Evaluation results

MethodLinks
ANCHORTransfer Pair=Qwen3-1.7B -> Qwen3-8B, Source Domain=AH, Evaluation Setting=Zero-shot transfer2026.058.96——8.96
Conf.-basedTransfer Pair=Qwen3-1.7B -> Qwen3-8B, Source Domain=AH, Evaluation Setting=Zero-shot transfer2026.056.78——6.78
ANCHORTransfer Pair=Llama-3.2-1B-Instruct -> Llama-3.1-8B-Instruct, Source Domain=UF, Evaluation Setting=Zero-shot transfer2026.056.01——6.01
NaiveTransfer Pair=Llama-3.2-1B-Instruct -> Llama-3.1-8B-Instruct, Source Domain=UF, Evaluation Setting=Zero-shot transfer2026.055.79——5.79
NaiveTransfer Pair=Qwen3-1.7B -> Qwen3-8B, Source Domain=AH, Evaluation Setting=Zero-shot transfer2026.055.57——5.57
NaiveTransfer Pair=Llama-3.2-1B-Instruct -> Llama-3.1-8B-Instruct, Source Domain=AH, Evaluation Setting=Zero-shot transfer2026.054.59——4.59
Conf.-basedTransfer Pair=Llama-3.2-1B-Instruct -> Llama-3.1-8B-Instruct, Source Domain=AH, Evaluation Setting=Zero-shot transfer2026.053.83——3.83
NaiveTransfer Pair=Qwen3-1.7B -> Qwen3-8B, Source Domain=UF, Evaluation Setting=Zero-shot transfer2026.053.61——3.2
Conf.-basedTransfer Pair=Llama-3.2-1B-Instruct -> Llama-3.1-8B-Instruct, Source Domain=UF, Evaluation Setting=Zero-shot transfer2026.053.39——3.39
ANCHORTransfer Pair=Qwen3-1.7B -> Qwen3-8B, Source Domain=UF, Evaluation Setting=Zero-shot transfer2026.053.28——3.28
ANCHORTransfer Pair=Llama-3.2-1B-Instruct -> Llama-3.1-8B-Instruct, Source Domain=AH, Evaluation Setting=Zero-shot transfer2026.052.08——2.08
Conf.-basedTransfer Pair=Qwen3-1.7B -> Qwen3-8B, Source Domain=UF, Evaluation Setting=Zero-shot transfer2026.051.42——1.24
SEAMTransfer Pair=Qwen3-1.7B -> Qwen3-8B, Source Domain=AH, Evaluation Setting=Zero-shot transfer2026.050.55——-2.87
SEAMTransfer Pair=Qwen3-1.7B -> Qwen3-8B, Source Domain=UF, Evaluation Setting=Zero-shot transfer2026.05-1.85——-4.23
SEAMTransfer Pair=Llama-3.2-1B-Instruct -> Llama-3.1-8B-Instruct, Source Domain=UF, Evaluation Setting=Zero-shot transfer2026.05-5.68——-12.33
SEAMTransfer Pair=Llama-3.2-1B-Instruct -> Llama-3.1-8B-Instruct, Source Domain=AH, Evaluation Setting=Zero-shot transfer2026.05-6.01——-19.05
ANCHORTransfer Pair=Llama-3.2-1B-Instruct -> Llama-3.1-8B-Instruct, Source Domain=H3, Evaluation Setting=In-distribution2026.05—67.23.57—
ANCHORTransfer Pair=Qwen3-1.7B -> Qwen3-8B, Source Domain=H3, Evaluation Setting=In-distribution2026.05—9.50.33—
Conf.-basedTransfer Pair=Llama-3.2-1B-Instruct -> Llama-3.1-8B-Instruct, Source Domain=H3, Evaluation Setting=In-distribution2026.05—-39.9-2.12—
Conf.-basedTransfer Pair=Qwen3-1.7B -> Qwen3-8B, Source Domain=H3, Evaluation Setting=In-distribution2026.05—-65.9-2.3—
NaiveTransfer Pair=Llama-3.2-1B-Instruct -> Llama-3.1-8B-Instruct, Source Domain=H3, Evaluation Setting=In-distribution2026.05—32.231.71—
NaiveTransfer Pair=Qwen3-1.7B -> Qwen3-8B, Source Domain=H3, Evaluation Setting=In-distribution2026.05—-87.7-3.06—
SEAMTransfer Pair=Llama-3.2-1B-Instruct -> Llama-3.1-8B-Instruct, Source Domain=H3, Evaluation Setting=In-distribution2026.05—-295.1-15.67—
SEAMTransfer Pair=Qwen3-1.7B -> Qwen3-8B, Source Domain=H3, Evaluation Setting=In-distribution2026.05—-332.1-11.59—