List your company

Instruction Following and Helpfulness Evaluation on AlpacaEval 2.0

49.4Win RateQWEN2-INSTRUCT w/ SSO_DPO
0.093612.894325.69538.4957Oct 22, 2024Nov 28, 2024Jan 4, 2025Feb 11, 2025Mar 20, 2025Apr 26, 2025Jun 3, 2025

Evaluation results

MethodLinks
QWEN2-INSTRUCT w/ SSO_DPOBackbone=QWEN2-INSTRUCT, Alignment Protocol=Principle-Based Automated Alignment, Synthetic Data=true2024.1049.436.5
QWEN2-SFT w/ SSO_DPOBackbone=QWEN2-SFT, Alignment Protocol=Principle-Based Automated Alignment, Synthetic Data=true2024.1045.443
QWEN2-INSTRUCT w/ PBAA_DPOBackbone=QWEN2-INSTRUCT, Alignment Protocol=Principle-Based Automated Alignment, Synthetic Data=true2024.1043.734.7
QWEN2-SFT w/ PBAA_DPOBackbone=QWEN2-SFT, Alignment Protocol=Principle-Based Automated Alignment, Synthetic Data=true2024.1042.537.7
LLAMA3-INSTRUCT w/ SSO_DPOBackbone=LLAMA3-INSTRUCT, Alignment Protocol=Principle-Based Automated Alignment, Synthetic Data=true2024.1028.925.6
Off-Policy AlignDistilBackbone=Llama3-8B2025.0328.5124.21
Qwen2.5-7BModel size=7B2025.0628.34—
LLAMA3-SFT w/ SSO_DPOBackbone=LLAMA3-SFT, Alignment Protocol=Principle-Based Automated Alignment, Synthetic Data=true2024.1028.335
On-Policy AlignDistilBackbone=Llama3-8B2025.0327.6426.29
LLAMA3-INSTRUCT w/ PBAA_DPOBackbone=LLAMA3-INSTRUCT, Alignment Protocol=Principle-Based Automated Alignment, Synthetic Data=true2024.1025.723.8
On-Policy AlignDistilBackbone=Qwen2.5-7B2025.0325.5331.32
Off-Policy AlignDistilBase Model=Qwen2.5-1.5B-Instruct2025.0324.2921.16
LLAMA3-SFT w/ PBAA_DPOBackbone=LLAMA3-SFT, Alignment Protocol=Principle-Based Automated Alignment, Synthetic Data=true2024.102429.5
Off-Policy AlignDistilBackbone=Qwen2.5-7B2025.0323.8229.16
LLAMA3-INSTRUCT w/ ULTRAFEEDBACKBackbone=LLAMA3-INSTRUCT, Synthetic Data=false2024.1022.623.8
On-Policy AlignDistilBase Model=Qwen2.5-1.5B-Instruct2025.0322.1119.45
TDPO1Backbone=Qwen2.5-7B2025.0319.4426.42
QWEN2-INSTRUCT w/ ULTRAFEEDBACKBackbone=QWEN2-INSTRUCT, Synthetic Data=false2024.1017.620.4
LLAMA3-SFT w/ ULTRAFEEDBACKBackbone=LLAMA3-SFT, Synthetic Data=false2024.1017.522
TDPO2Backbone=Qwen2.5-7B2025.0317.3322.63
QWEN2-INSTRUCTBackbone=QWEN2-INSTRUCT2024.1017.219.5
DPOBackbone=Qwen2.5-7B2025.0317.122.25
RTOBackbone=Llama3-8B2025.0316.9117
Qwen2.5-1.5B + daDPOBase model=Qwen2.5-1.5B, RLHF method=daDPO2025.0616.41—
On-Policy AlignDistilBase Model=Qwen2-1.5B-Instruct2025.0315.6512.93
DPOBackbone=Llama3-8B2025.0315.5815.67
RTOBase Model=Qwen2.5-1.5B-Instruct2025.0315.5316.54
LLAMA3-SFTBackbone=LLAMA3-SFT2024.101520.6
QWEN2-SFT w/ ULTRAFEEDBACKBackbone=QWEN2-SFT, Synthetic Data=false2024.101520.2
Qwen2.5-1.5B + dDPOBase model=Qwen2.5-1.5B, RLHF method=dDPO2025.0614.66—
Off-Policy AlignDistilBase Model=Qwen2-1.5B-Instruct2025.0314.2911.79
DPO beta=0.01Base Model=Qwen2.5-1.5B-Instruct, beta=0.012025.0314.2914.09
TDPO2Backbone=Llama3-8B2025.0313.7914.48
TDPO1Backbone=Llama3-8B2025.0313.7314.53
QWEN2-SFTBackbone=QWEN2-SFT2024.1013.220.1
LLAMA3-INSTRUCTBackbone=LLAMA3-INSTRUCT2024.1013.220.1
SimPOBackbone=Llama3-8B2025.0312.9211.24
PPOBase Model=Qwen2.5-1.5B-Instruct2025.0312.6718.06
RTOBackbone=Qwen2.5-7B2025.031217.75
DPO beta=0.01Base Model=Qwen2-1.5B-Instruct, beta=0.012025.0311.6110.72
DPOBase Model=Qwen2.5-1.5B-Instruct2025.0310.7414.35
SimPOBackbone=Qwen2.5-7B2025.0310.3715.65
KTOBase Model=Qwen2.5-1.5B-Instruct2025.031014.07
TDPO1Base Model=Qwen2.5-1.5B-Instruct2025.039.9413.19
SimPOBase Model=Qwen2.5-1.5B-Instruct2025.039.8111.61
SimPOBase Model=Qwen2-1.5B-Instruct2025.039.638.19
RTOBase Model=Qwen2-1.5B-Instruct2025.039.328.92
TDPO2Base Model=Qwen2.5-1.5B-Instruct2025.039.0713.64
Initial ModelBase Model=Qwen2.5-1.5B-Instruct2025.038.9412.57
Qwen2.5-1.5BModel size=1.5B2025.066.44—
KTOBase Model=Qwen2-1.5B-Instruct2025.036.347.16
DPOBase Model=Qwen2-1.5B-Instruct2025.035.036.42
TDPO1Base Model=Qwen2-1.5B-Instruct2025.034.66.58
PPOBase Model=Qwen2-1.5B-Instruct2025.034.414.86
SFTBackbone=Qwen2.5-7B2025.033.667.51
TDPO2Base Model=Qwen2-1.5B-Instruct2025.032.423.59
SFTBackbone=Llama3-8B2025.032.33
Initial ModelBase Model=Qwen2-1.5B-Instruct2025.031.993.1