List your company

Multi-turn Instruction Following on MT Bench

9.16MT-Bench Score (GPT-4)Llama-3.1-Nemotron-70B-Instruct
6.00886.82697.6458.4631Mar 6, 2025May 21, 2025Aug 6, 2025Oct 22, 2025Jan 6, 2026Mar 24, 2026Jun 9, 2026

Evaluation results

MethodLinks
Llama-3.1-Nemotron-70B-InstructFeedback + Edit Protocol=true2025.039.162,614.4———————
Llama-3.3-70B-InstructFeedback + Edit Protocol=true2025.039.072,362.1———————
Llama-3.1-Nemotron-70B-InstructFeedback + Edit Protocol=false2025.038.982,199.8———————
Claude-3-5-Sonnet-202406202025.038.811,619.9———————
GPT-4o-2024-05-132025.038.741,752.2———————
Llama-3.1-405B-Instruct2025.038.491,664.7———————
IPOBackbone=Llama3-8B-Instruct2026.048.3————————
RMiPOBackbone=Llama3-8B-Instruct2026.048.3————————
Llama-3.3-70B-InstructFeedback + Edit Protocol=false2025.038.291,827.6———————
Llama-3.1-70B-Instruct2025.038.221,728.6———————
KTOBackbone=Llama3-8B-Instruct2026.048.2————————
SimPERBackbone=Llama3-8B-Instruct2026.048.2————————
InitialBackbone=Llama3-8B-Instruct2026.048.1————————
SLiCBackbone=Llama3-8B-Instruct2026.048.1————————
α-DPOBackbone=Llama3-8B-Instruct2026.048.1————————
DPOBackbone=Llama3-8B-Instruct2026.048————————
CPOBackbone=Llama3-8B-Instruct2026.048————————
SimPOBackbone=Llama3-8B-Instruct2026.048————————
α-DPOBackbone=Mistral-7B-Instruct2026.048————————
β-DPOBackbone=Llama3-8B-Instruct2026.048————————
ϵ-DPOBackbone=Llama3-8B-Instruct2026.048————————
SLiCBackbone=Mistral-7B-Instruct2026.047.8————————
IPOBackbone=Mistral-7B-Instruct2026.047.8————————
KTOBackbone=Llama3-8B-Base2026.047.8————————
ϵ-DPOBackbone=Mistral-7B-Instruct2026.047.8————————
SimPERBackbone=Mistral-7B-Instruct2026.047.8————————
RMiPOBackbone=Mistral-7B-Instruct2026.047.8————————
RMiPOBackbone=Llama3-8B-Base2026.047.8————————
DPOBackbone=Llama3-8B-Base2026.047.7————————
KTOBackbone=Mistral-7B-Instruct2026.047.7————————
SimPOBackbone=Llama3-8B-Base2026.047.7————————
α-DPOBackbone=Llama3-8B-Base2026.047.7————————
β-DPOBackbone=Mistral-7B-Instruct2026.047.7————————
β-DPOBackbone=Llama3-8B-Base2026.047.7————————
SimPERBackbone=Llama3-8B-Base2026.047.7————————
DPOBackbone=Mistral-7B-Instruct2026.047.6————————
SLiCBackbone=Llama3-8B-Base2026.047.6————————
SimPOBackbone=Mistral-7B-Instruct2026.047.6————————
ϵ-DPOBackbone=Llama3-8B-Base2026.047.6————————
InitialBackbone=Mistral-7B-Instruct2026.047.5————————
CPOBackbone=Mistral-7B-Instruct2026.047.5————————
α-DPOBackbone=Mistral-7B-Base2026.047.5————————
SimPERBackbone=Mistral-7B-Base2026.047.5————————
RMiPOBackbone=Mistral-7B-Base2026.047.5————————
InstructModel=Qwen3-4B2026.057.45————————
GW-DPOvariant=bi.2026.067.425—————8.36.551.75
baselinetraining_config=baseline2026.067.406—————8.0256.7881.238
SLiCBackbone=Mistral-7B-Base2026.047.4————————
IPOBackbone=Llama3-8B-Base2026.047.4————————
CPOBackbone=Llama3-8B-Base2026.047.4————————
ϵ-DPOBackbone=Mistral-7B-Base2026.047.4————————
DPOtraining_config=DPO2026.067.363—————8.16.6251.475
Llama-3.1-Instructtype=off-the-shelf anchor2026.067.356—————7.8756.8381.037
DPOBackbone=Mistral-7B-Base2026.047.3————————
SimPOBackbone=Mistral-7B-Base2026.047.3————————
β-DPOBackbone=Mistral-7B-Base2026.047.3————————
DivPOModel=Qwen3-4B2026.057.281————————
3-level (lin.)training_config=3-level (lin.)2026.067.269—————7.9756.5631.413
SFTtraining_config=SFT2026.067.238—————7.9636.5131.45
IPOBackbone=Mistral-7B-Base2026.047.2————————
AttentionPOReference Model (πref)=LLaMA-3-8B-Instruct2026.057.19————————
DPOModel=Qwen3-4B2026.057.113————————
GW-DPOvariant=lin.2026.067.113—————86.2251.775
DivPOModel=OLMo-3-7B2026.057.063————————
REDIPOModel=Qwen3-4B2026.057.044————————
KTOBackbone=Mistral-7B-Base2026.047————————
IPOReference Model (πref)=LLaMA-3-8B-Instruct2026.056.99————————
REDIPOModel=OLMo-3-7B2026.056.956————————
InstructModel=OLMo-3-7B2026.056.919————————
CPOBackbone=Mistral-7B-Base2026.046.8————————
KTOReference Model (πref)=LLaMA-3-8B-Instruct2026.056.79————————
R-DPOReference Model (πref)=LLaMA-3-8B-Instruct2026.056.79————————
BaseReference Model (πref)=LLaMA-3-8B-Instruct2026.056.78————————
SLiC-HFReference Model (πref)=LLaMA-3-8B-Instruct2026.056.76————————
ORPOReference Model (πref)=LLaMA-3-8B-Instruct2026.056.76————————
DPOReference Model (πref)=LLaMA-3-8B-Instruct2026.056.73————————
B4GModel=QWEN2.5-7B, Task=safety enhancement2026.036.72————————
DPOModel=OLMo-3-7B2026.056.7————————
CPOReference Model (πref)=LLaMA-3-8B-Instruct2026.056.67————————
RRHFReference Model (πref)=LLaMA-3-8B-Instruct2026.056.63————————
InitialBackbone=Llama3-8B-Base2026.046.6————————
REDIPOModel=LLaMA-3.1-8B2026.056.594————————
No-ISEtraining_config=No-ISE2026.066.544—————7.1635.9251.238
DPOReference Model (πref)=LLaMA-3-8B-Base-SFT2026.056.54————————
AttentionPOReference Model (πref)=LLaMA-3-8B-Base-SFT2026.056.52————————
SimPOReference Model (πref)=LLaMA-3-8B-Instruct2026.056.51————————
InstructModel=LLaMA-3.1-8B2026.056.475————————
DPOModel=LLaMA-3.1-8B2026.056.463————————
B4GModel=QWEN2.5-7B, Task=access control2026.036.39————————
B4GModel=QWEN2.5-7B, Task=model identity2026.036.34————————
R-DPOReference Model (πref)=LLaMA-3-8B-Base-SFT2026.056.33————————
B4GModel=LLAMA3.1-8B, Task=safety enhancement2026.036.31————————
InitialBackbone=Mistral-7B-Base2026.046.3————————
IPOReference Model (πref)=LLaMA-3-8B-Base-SFT2026.056.23————————
B4GModel=LLAMA3.1-8B, Task=model identity2026.036.2————————
B4GModel=GEMMA-2-9B, Task=access control2026.036.2————————
RRHFReference Model (πref)=LLaMA-3-8B-Base-SFT2026.056.18————————
B4GModel=LLAMA2-13B, Task=access control2026.036.14————————
CPOReference Model (πref)=LLaMA-3-8B-Base-SFT2026.056.13————————
SimPOReference Model (πref)=LLaMA-3-8B-Base-SFT2026.056.13————————