List your company

Mathematical Reasoning on GSM8K (Acc, H100 TPS, TPF)

95.38GSM8K Accuracy (%)MIPRO
74.600879.995485.3990.7846Oct 9, 2025Nov 19, 2025Dec 31, 2025Feb 11, 2026Mar 24, 2026May 5, 2026Jun 16, 2026

Evaluation results

MethodLinks
MIPROBackbone Model=Gemini-2-Flash, Prompt Optimization Technique=MIPRO2026.0595.38————
KVZipKV Cache Budget Sparsity=0.22025.1095.2————
BaselineBackbone Model=Gemini-2-Flash, Prompt Optimization Technique=Baseline2026.0595.19————
GEPABackbone Model=Gemini-2-Flash, Prompt Optimization Technique=GEPA2026.0595.15————
KVZipKV Cache Budget Sparsity=0.42025.1095.1————
DuoAttentionKV Cache Budget Sparsity=0.42025.1095————
PSAO_LLMBackbone Model=Gemini-2-Flash, Prompt Optimization Technique=PSAO_LLM2026.0594.95————
COPROBackbone Model=Gemini-2-Flash, Prompt Optimization Technique=COPRO2026.0594.94————
RLKVKV Cache Budget Sparsity=0.22025.1094.9————
DuoAttentionKV Cache Budget Sparsity=0.22025.1094.8————
ProTeGiBackbone Model=Gemini-2-Flash, Prompt Optimization Technique=ProTeGi2026.0594.48————
RLKVKV Cache Budget Sparsity=0.42025.1094.4————
PromptAgentBackbone Model=Gemini-2-Flash, Prompt Optimization Technique=PromptAgent2026.0594.2————
R-KVKV Cache Budget Sparsity=0.22025.1094.2————
SuCoModel Scale=7B2026.0693.9————
AdaptThinkModel Scale=7B2026.0692.9————
S-GRPOModel Scale=7B2026.0692.8————
LHRMsModel Scale=7B2026.0692.4————
KVZipKV Cache Budget Sparsity=0.62025.1091.4————
AdaCoTModel Scale=7B2026.0691.4————
RLKVKV Cache Budget Sparsity=0.62025.1090.8————
RLKVKV Cache Budget Sparsity=0.22025.1090.7————
RLKVKV Cache Budget Sparsity=0.42025.1090.1————
BaseAllowed Domains=—2026.0589.6————
DuoAttentionKV Cache Budget Sparsity=0.62025.1089.6————
DeepSeek-R1-DistillModel Scale=7B2026.0689.3————
DPO+FilterBase=Qwen 7b, Training Dataset=UltraFeedback, Pn Type=12025.1089————
DuoAttentionKV Cache Budget Sparsity=0.22025.1088.9————
VIGORBackbone=Qwen2.5-7B, Training Set=MATH2026.0588.7————
PALETTEAllowed Domains=Illegal | Violence2026.0588.5————
PALETTEAllowed Domains=Sexual | Illegal2026.0588.2————
KVZipKV Cache Budget Sparsity=0.22025.1088.2————
PALETTEAllowed Domains=Hate | Disinfo2026.0588.1————
DPO+FilterBase=Qwen 7b, Training Dataset=UltraFeedback, Pn Type=22025.1088————
DPO+FilterBase=Qwen 7b, Training Dataset=HelpSteer3, Pn Type=12025.1088————
R-KVKV Cache Budget Sparsity=0.22025.1087.7————
SuCoModel Scale=1.5B2026.0687.7————
DPO+FilterBase=Qwen 7b, Training Dataset=HelpSteer3, Pn Type=22025.1087.5————
PALETTEAllowed Domains=Disinfo | Sexual2026.0587.3————
INTUITORBackbone=Qwen2.5-7B, Training Set=MATH2026.0587.19————
BaseBase=Qwen 7b, Training Dataset=UltraFeedback, Pn Type=-2025.1087————
DPOBase=Qwen 7b, Training Dataset=UltraFeedback, Pn Type=22025.1087————
BaseBase=Qwen 7b, Training Dataset=HelpSteer3, Pn Type=-2025.1087————
DPOBase=Qwen 7b, Training Dataset=HelpSteer3, Pn Type=22025.1087————
Math-InstructModel Scale=7B2026.0687————
DPOBase=Qwen 7b, Training Dataset=UltraFeedback, Pn Type=12025.1086.5————
DPOBase=Qwen 7b, Training Dataset=HelpSteer3, Pn Type=12025.1086.5————
BaseAllowed Domains=None, Backbone=QWEN2.5-7B-INSTRUCT2026.0586.5————
PALETTEAllowed Domains=Violence | Hate, Backbone=QWEN2.5-7B-INSTRUCT2026.0586.3————
PALETTEAllowed Domains=Violence | Hate2026.0586.2————
PALETTEAllowed Domains=Disinfo | Sexual, Backbone=QWEN2.5-7B-INSTRUCT2026.0585.7————
LHRMsModel Scale=1.5B2026.0685.7————
PALETTEAllowed Domains=Illegal | Violence, Backbone=QWEN2.5-7B-INSTRUCT2026.0585.6————
R-KVKV Cache Budget Sparsity=0.42025.1085.6————
PALETTEAllowed Domains=Hate | Disinfo, Backbone=QWEN2.5-7B-INSTRUCT2026.0585.4————
KVZipKV Cache Budget Sparsity=0.42025.1085————
GT-RewardBackbone=Qwen2.5-7B, Training Set=MATH2026.0584.8————
PALETTEAllowed Domains=Sexual | Illegal, Backbone=QWEN2.5-7B-INSTRUCT2026.0584.5————
R-KVKV Cache Budget Sparsity=0.42025.1084————
CASTAllowed Domains=Disinfo | Sexual2026.0583.8————
CASTAllowed Domains=Hate | Disinfo2026.0583.5————
CASTAllowed Domains=Sexual | Illegal2026.0583.4————
S-GRPOModel Scale=1.5B2026.0683.4————
AdaptThinkModel Scale=1.5B2026.0683.2————
RLKVKV Cache Budget Sparsity=0.62025.1083.1————
AdaCoTModel Scale=1.5B2026.0682.7————
CASTAllowed Domains=Illegal | Violence2026.0582.6————
CASTAllowed Domains=Violence | Hate2026.0582.2————
DuoAttentionKV Cache Budget Sparsity=0.42025.1082————
GT-RewardBackbone=Qwen2.5-3B, Training Set=MATH2026.0581.96————
VIGORBackbone=Qwen2.5-3B, Training Set=MATH2026.0581.8————
wd1Generation Length=5122026.0581.65————
StableDRLGeneration Length=2562026.0581.35————
SPGGeneration Length=2562026.0581.05————
StableDRLGeneration Length=5122026.0581.05————
d1Generation Length=5122026.0580.67————
DeepSeek-R1-DistillModel Scale=1.5B2026.0680.3————
SPGGeneration Length=5122026.0580.29————
ESPOGeneration Length=2562026.0580.21————
CASTAllowed Domains=Illegal | Violence, Backbone=QWEN2.5-7B-INSTRUCT2026.0580.1————
Diffu-GRPOGeneration Length=5122026.0579.98————
ESPOGeneration Length=5122026.0579.76————
INTUITORBackbone=Qwen2.5-3B, Training Set=MATH2026.0579.68————
GDPOGeneration Length=5122026.0579.23————
Math-InstructModel Scale=1.5B2026.0679————
wd1Generation Length=2562026.0578.85————
CASTAllowed Domains=Sexual | Illegal, Backbone=QWEN2.5-7B-INSTRUCT2026.0578.6————
SFTGeneration Length=5122026.0578.54————
AdaBlock-dLLMGeneration Length=5122026.0577.94————
LLaDA-8B-InstructGeneration Length=5122026.0577.63————
CASTAllowed Domains=Hate | Disinfo, Backbone=QWEN2.5-7B-INSTRUCT2026.0577.3————
d1Generation Length=2562026.0577.03————
CASTAllowed Domains=Disinfo | Sexual, Backbone=QWEN2.5-7B-INSTRUCT2026.0576.8————
GDPOGeneration Length=2562026.0576.57————
CASTAllowed Domains=Violence | Hate, Backbone=QWEN2.5-7B-INSTRUCT2026.0576.5————
Diffu-GRPOGeneration Length=2562026.0576.35————
AdaBlock-dLLMGeneration Length=2562026.0576.27————
LLaDA-8B-InstructGeneration Length=2562026.0576.19————
SFTGeneration Length=2562026.0575.59————
Base (8B Instruct)Backbone=Llama 32026.0575.4————