List your company
ResearchFormerly WizWand

Multi-objective personalized alignment on Multifaceted dataset (test)

75AMRMATO
9.4826.4943.560.51May 25, 2026
Updated 5mo ago

Evaluation results

MethodLinks
MATOBackbone LLM=Qwen3-8B, Judge=GPT-4o2026.05754.33.42
OPADBackbone LLM=Qwen3-8B, Judge=GPT-4o2026.05744.283.39
MATOBackbone LLM=Llama-3.1-8B, Judge=GPT-4o2026.05734.083.22
Linear AlignmentBackbone LLM=Qwen3-8B, Judge=GPT-4o2026.05724.193.31
Preference PromptBackbone LLM=Qwen3-8B, Judge=GPT-4o2026.05704.193.23
OPADBackbone LLM=Llama-3.1-8B, Judge=GPT-4o2026.05684.033.1
CoSBackbone LLM=Qwen3-8B, Judge=GPT-4o2026.056643.13
Preference PromptBackbone LLM=Llama-3.1-8B, Judge=GPT-4o2026.05663.972.96
Linear AlignmentBackbone LLM=Llama-3.1-8B, Judge=GPT-4o2026.05643.872.98
AmuletBackbone LLM=Qwen3-8B, Judge=GPT-4o2026.05603.792.94
CoSBackbone LLM=Llama-3.1-8B, Judge=GPT-4o2026.05523.522.63
Base LLMBackbone LLM=Qwen3-8B, Judge=GPT-4o2026.05433.662.34
MATOBackbone LLM=Llama-3.2-1B, Judge=GPT-4o2026.05393.232.25
Linear AlignmentBackbone LLM=Llama-3.2-1B, Judge=GPT-4o2026.05372.782.11
OPADBackbone LLM=Llama-3.2-1B, Judge=GPT-4o2026.05362.991.88
Base LLMBackbone LLM=Llama-3.1-8B, Judge=GPT-4o2026.05343.362.04
AmuletBackbone LLM=Llama-3.1-8B, Judge=GPT-4o2026.05322.922.07
Preference PromptBackbone LLM=Llama-3.2-1B, Judge=GPT-4o2026.05313.072
MATOBackbone LLM=Qwen3-0.6B, Judge=GPT-4o2026.05272.991.87
OPADBackbone LLM=Qwen3-0.6B, Judge=GPT-4o2026.05252.971.8
Linear AlignmentBackbone LLM=Qwen3-0.6B, Judge=GPT-4o2026.05242.81.76
Base LLMBackbone LLM=Llama-3.2-1B, Judge=GPT-4o2026.05212.781.69
Preference PromptBackbone LLM=Qwen3-0.6B, Judge=GPT-4o2026.05192.781.64
CoSBackbone LLM=Qwen3-0.6B, Judge=GPT-4o2026.05192.591.64
CoSBackbone LLM=Llama-3.2-1B, Judge=GPT-4o2026.05172.341.6
Base LLMBackbone LLM=Qwen3-0.6B, Judge=GPT-4o2026.05142.581.49
AmuletBackbone LLM=Qwen3-0.6B, Judge=GPT-4o2026.05122.311.46
AmuletBackbone LLM=Llama-3.2-1B, Judge=GPT-4o2026.05122.061.42