List your company
ResearchFormerly WizWand

Tool Use on ToolAlpaca

77.9Tool Use Success RateEMA
37.75648.17858.669.022May 7, 2026May 8, 2026May 9, 2026May 10, 2026May 11, 2026May 12, 2026
Updated 5mo ago

Evaluation results

MethodLinks
EMABase Model=Qwen2.5-7B, Protocol=EMA teacher2026.0577.9
UniSD*Base Model=Qwen2.5-7B2026.0577.9
Agree (Seq.)Base Model=Qwen2.5-7B, Protocol=Sequence-level agreement, Context Retrieval=True2026.0576.5
Match (Joint)Base Model=Qwen2.5-7B, Protocol=Joint logit-representation matching2026.0576.5
GKDSFT warm-up=3K, Continual learning=true2026.0576
AOPDSFT warm-up=3K, Continual learning=true2026.0575
Agree (Tok.)Base Model=Qwen2.5-7B, Protocol=Token-level agreement, Context Retrieval=True2026.0575
ContrastBase Model=Qwen2.5-7B, Protocol=Token-level contrastive learning2026.0575
SDFTBase Model=Qwen2.5-7B2026.0573.5
GKDBase Model=Qwen2.5-7B2026.0572.1
Match (Repr.)Base Model=Qwen2.5-7B, Protocol=Representation-only matching2026.0572.1
ClipBase Model=Qwen2.5-7B, Protocol=Divergence clipping2026.0570.6
CREDITModel=Qwen3-8B, Evaluation Protocol=best avg@162026.0570.4
GRPOModel=Qwen3-8B, Evaluation Protocol=best avg@162026.0569.7
SDPOModel=Qwen3-8B, Evaluation Protocol=best avg@162026.0568
CREDITModel=OLMo-3-7B-Instruct, Evaluation Protocol=best avg@162026.0567.4
ExOPDSFT warm-up=3K, Continual learning=true2026.0567
SFTBase Model=Qwen2.5-7B2026.0566.2
OPDSFT warm-up=3K, Continual learning=true2026.0565
GRPOModel=OLMo-3-7B-Instruct, Evaluation Protocol=best avg@162026.0564.4
RawBase Model=Qwen2.5-7B2026.0561.8
OPSDBase Model=Qwen2.5-7B2026.0561.8
SDPOModel=OLMo-3-7B-Instruct, Evaluation Protocol=best avg@162026.0561.5
BaseModel=Qwen3-8B, Evaluation Protocol=best avg@162026.0557.5
SSDBase Model=Qwen2.5-7B2026.0555.9
BaseModel=OLMo-3-7B-Instruct, Evaluation Protocol=best avg@162026.0539.3