List your company
ResearchFormerly WizWand

Mathematical Reasoning on Overall

89.6AccuracyWHISPER
10.14430.77251.472.028May 16, 2025Jul 11, 2025Sep 5, 2025Oct 31, 2025Dec 26, 2025Feb 20, 2026Apr 18, 2026
Updated 5mo ago

Evaluation results

MethodLinks
WHISPERBase Model=Qwen3-14B2025.1089.6————63
ChainofDraftBase Model=Qwen3-14B2025.1089.5————83.9
BeConciseBase Model=Qwen3-14B2025.1088.7————88
OriginalBase Model=Qwen3-14B2025.1087.9————100
GEPABase Model=Qwen3-14B2025.1087.6————75.1
DEER*Base Model=Qwen3-14B2025.1087.6————80.5
WHISPERBase Model=DeepSeek-R1-Distill-Qwen-14B2025.1086.3————78
BeConciseBase Model=DeepSeek-R1-Distill-Qwen-14B2025.1086.1————94.4
OriginalBase Model=DeepSeek-R1-Distill-Qwen-14B2025.1085.9————100
GEPABase Model=DeepSeek-R1-Distill-Qwen-14B2025.1085.8————81.3
DEER*Base Model=DeepSeek-R1-Distill-Qwen-14B2025.1083.9————83.3
ChainofDraftBase Model=DeepSeek-R1-Distill-Qwen-14B2025.1082.3————78.6
Step-GRPOBackbone=Qwen3-8B2026.0482.1———68—
GRPOBackbone=Qwen3-8B2026.0480.9———89.7—
GRPO+SOPBackbone=Qwen3-8B2026.0480.4———67.9—
SGPOBackbone=QwQ-32B, Judge Model=DeepSeek-V3-03242025.0580.17—————
VanillaBackbone=Qwen3-8B2026.0479.9———100—
GRPO-λBackbone=Qwen3-8B2026.0479.9———62.9—
VanillaBackbone=Qwen3-4B2026.0479.3———100—
GRPOBackbone=Qwen3-4B2026.0479.3———86.9—
GRPOBackbone=QwQ-32B2025.0579.15—————
Step-GRPOBackbone=Qwen3-4B2026.0479.1———68.7—
WHISPERBase Model=DeepSeek-R1-Distill-LLaMA-8B2025.1079————80.3
BaselineBackbone=QwQ-32B2025.0578.84—————
OriginalBase Model=DeepSeek-R1-Distill-LLaMA-8B2025.1078.5————100
BeConciseBase Model=DeepSeek-R1-Distill-LLaMA-8B2025.1078.5————96.7
GRPO+LPBackbone=Qwen3-8B2026.0478.4———53.2—
GEPABase Model=DeepSeek-R1-Distill-LLaMA-8B2025.1078.4————86.8
ChainofDraftBase Model=DeepSeek-R1-Distill-LLaMA-8B2025.1078————92.6
GRPO-8kBackbone=Qwen3-4B2026.0477.6———68.1—
GRPO+LPBackbone=Qwen3-4B2026.0477.6———52.8—
GRPO-λBackbone=Qwen3-4B2026.0477.5———61.5—
GRPO+SOPBackbone=Qwen3-4B2026.0477.4———69.2—
DEER*Base Model=DeepSeek-R1-Distill-LLaMA-8B2025.1076.5————84
GRPO-8kBackbone=Qwen3-8B2026.0475.5———70.4—
DEER+SFTBackbone=Qwen3-8B2026.0472.6———78.9—
NoThinkingBase Model=Qwen3-14B2025.1069.8————28.8
DEER+SFTBackbone=Qwen3-4B2026.0467.6———80.3—
SGPOBackbone=DeepSeek-R1-Distill-Qwen-7B, Judge Model=DeepSeek-V3-03242025.0566.19—————
GRPOBackbone=Qwen3-1.7B2026.0466.1———88.3—
SGPOBackbone=DeepSeek-R1-Distill-Qwen-7B, Judge Model=Qwen3-235B-A22B2025.0565.42—————
SGPOBackbone=DeepSeek-R1-Distill-Qwen-7B, Judge Model=o4-mini-04162025.0565.41—————
SGPOBackbone=DeepSeek-R1-Distill-Qwen-7B, Judge Model=QwQ-32B, Stability Parameters=w/o {beta,gamma}2025.0565.08—————
SGPOBackbone=DeepSeek-R1-Distill-Qwen-7B, Judge Model=QwQ-32B2025.0564.91—————
Segment Selective SFTModel Backbone=R1-Distill-Qwen-7B, Decoding Strategy=Greedy Decoding2026.0164.5——8,499——
GRPOBackbone=DeepSeek-R1-Distill-Qwen-7B2025.0563.72—————
Step-GRPOBackbone=Qwen3-1.7B2026.0463.2———69.7—
BaselineBackbone=DeepSeek-R1-Distill-Qwen-7B2025.0562.85—————
GRPO-8kBackbone=Qwen3-1.7B2026.0462.5———70.8—
NoThinkingBase Model=DeepSeek-R1-Distill-Qwen-14B2025.1062.5————23.5
GRPO-λBackbone=Qwen3-1.7B2026.0461.9———68.3—
GRPO+SOPBackbone=Qwen3-1.7B2026.0461.5———76.1—
VanillaBackbone=Qwen3-1.7B2026.0461.2———100—
GRPO+LPBackbone=Qwen3-1.7B2026.0460.7———57.5—
NoThinkingBase Model=DeepSeek-R1-Distill-LLaMA-8B2025.1056————31.9
SGPOBackbone=DeepSeek-R1-Distill-Llama-8B, Judge Model=Claude-3.72025.0553.81—————
GRPOBackbone=DeepSeek-R1-Distill-Llama-8B2025.0553.06—————
BaselineBackbone=DeepSeek-R1-Distill-Llama-8B2025.0551.58—————
DEER+SFTBackbone=Qwen3-1.7B2026.0451.2———100.2—
Segment Selective SFTModel Backbone=R1-Distill-Qwen-1.5B, Decoding Strategy=Greedy Decoding2026.0146.9——13,506——
Segment Selective SFTModel Backbone=Qwen2.5-7B-Instruct, Decoding Strategy=Greedy Decoding2026.0145.6——9,852——
SGPOBackbone=Qwen2.5-32B-Instruct, Judge Model=o4-mini-04162025.0545.06—————
ReLIFTBackbone=Qwen2.5-7B2025.0645——3,696——
GRPOBackbone=Qwen2.5-32B-Instruct2025.0544.39—————
SFTBackbone=Qwen2.5-7B2025.0644——6,183——
BaselineBackbone=Qwen2.5-32B-Instruct2025.0543.74—————
RLBackbone=Qwen2.5-7B2025.0643.1——1,334——
SGPOBackbone=Qwen2.5-14B-Instruct, Judge Model=o4-mini-04162025.0541—————
GRPOBackbone=Qwen2.5-14B-Instruct2025.0540.11—————
BaselineBackbone=Qwen2.5-14B-Instruct2025.0539.85—————
Qwen-7B-InstructBackbone=Qwen2.5-7B2025.0637——1,949——
ReLIFTBackbone=Qwen2.5-Math-1.5B2025.0636.5——2,437——
Qwen-Math-1.5B-InstructBackbone=Qwen2.5-Math-1.5B2025.0634.2——3,230——
RLBackbone=Qwen2.5-Math-1.5B2025.0634.2——1,653——
SFTBackbone=Qwen2.5-Math-1.5B2025.0632.6——6,870——
Qwen-7BBackbone=Qwen2.5-7B2025.0627.1——1,118——
LLaMa-8B-InstructBackbone=LLaMA-3.1-8B2025.0620.5——975——
ReLIFTBackbone=LLaMA-3.1-8B2025.0617.3——1,000——
Qwen-Math-1.5BBackbone=Qwen2.5-Math-1.5B2025.0615.5——1,712——
RLBackbone=LLaMA-3.1-8B2025.0614.6——767——
SFTBackbone=LLaMA-3.1-8B2025.0613.2——1,922——
Segment Selective SFTModel Backbone=R1-Distill-Qwen-1.5B, Decoding Strategy=Temperature Sampling2026.01—51.773.29,388——
Segment Selective SFTModel Backbone=R1-Distill-Qwen-7B, Decoding Strategy=Temperature Sampling2026.01—65.8807,709——
Segment Selective SFTModel Backbone=Qwen2.5-7B-Instruct, Decoding Strategy=Temperature Sampling2026.01—4566.59,195——