List your company

Mathematical Reasoning on AIME 2024 (Accuracy and Token Usage)

79.3AccuracyThoughtFold
58.458463.869269.2874.6908Aug 6, 2025Sep 25, 2025Nov 14, 2025Jan 4, 2026Feb 23, 2026Apr 14, 2026Jun 4, 2026

Evaluation results

MethodLinks
ThoughtFoldModel=Qwen3-14B2026.0679.38,869
ThoughtFoldModel=Qwen3-8B2026.0678.19,099
S-GRPOModel=Qwen3-14B2026.0677.98,932
CoT-SCBackbone=Qwen3-4B2025.0977.7838,140.92
CoT-SCBackbone=Qwen3-8B2025.0977.7841,127.48
CoTBackbone=Qwen3-14B2025.0977.789,685.22
GRPOModel=Qwen3-14B2026.0677.714,544
S-GRPOModel=Qwen3-8B2026.0677.38,810
DEERModel=Qwen3-14B2026.0476.77,619
DEERModel=Qwen3-32B2026.0476.78,682
DEERModel=Qwen3-14B2026.0476.77,619
DEERModel=Qwen3-32B2026.0476.78,682
VanillaModel=Qwen3-14B2026.0675.414,116
Short-RLModel=Qwen3-14B2026.0675.29,255
VanillaModel=Qwen3-8B2026.0675.115,425
RL + Length PenaltyModel=Qwen3-14B2026.0674.89,056
CoT-SCBackbone=Qwen3-14B2025.0974.0736,955.67
GRPOModel=Qwen3-8B2026.067415,061
RL + Length PenaltyModel=Qwen3-8B2026.0673.89,666
COTModel=Qwen3-14B2026.0473.311,742
SATModel=Qwen3-14B2026.0473.39,626
COTModel=Qwen3-14B2026.0473.311,742
SATModel=Qwen3-14B2026.0473.39,626
ReLATModel=Qwen3-8B2026.0673.3—
Short-RLModel=Qwen3-8B2026.06739,972
TALEModel=Qwen3-14B2026.0471.110,860
CGRSModel=Qwen3-14B2026.04708,662
COTModel=Qwen3-32B2026.047010,788
SATModel=Qwen3-32B2026.04709,183
DEERModel=QwQ-32B2026.047010,097
SATModel=QwQ-32B2026.04709,181
CGRSModel=Qwen3-14B2026.04708,662
COTModel=Qwen3-32B2026.047010,788
SATModel=Qwen3-32B2026.04709,183
DEERModel=QwQ-32B2026.047010,097
SATModel=QwQ-32B2026.04709,181
ReLATModel=Qwen3-14B2026.0670—
TALEModel=Qwen3-8B2026.0468.910,942
CGRSModel=QwQ-32B2026.0468.98,202
CGRSModel=QwQ-32B2026.0468.98,202
DEERModel=DS-Qwen-14B2026.0468.48,115
DEERModel=DS-Qwen-14B2026.0468.48,115
Qwen3 + DAPO + Hidden-AlignSize=14B2026.0668.12—
TALEModel=Qwen3-32B2026.0467.810,688
COTModel=Qwen3-8B2026.0466.712,393
SATModel=Qwen3-8B2026.0466.79,674
DEERModel=Nemo-Llama-8b2026.0466.79,755
COTModel=QwQ-32B2026.0466.711,305
COTModel=Qwen3-8B2026.0466.712,393
SATModel=Qwen3-8B2026.0466.79,674
DEERModel=Nemo-Llama-8b2026.0466.79,755
COTModel=QwQ-32B2026.0466.711,305
LatentMASModel=Qwen3-14B2026.0666.7—
RoTBackbone=Qwen3-4B2025.0966.6710,621.33
RoTBackbone=Qwen3-8B2025.0966.6711,561.92
RAGBackbone=Qwen3-14B2025.0966.6710,981.38
RoTBackbone=Qwen3-14B2025.0966.6711,078.44
GRPOModel=DeepSeek-R1-Distill-Qwen-14B2026.0665.813,504
DynasorModel=Qwen3-14B2026.0465.69,775
CGRSModel=Qwen3-32B2026.0465.68,128
CGRSModel=Qwen3-32B2026.0465.68,128
ThoughtFoldModel=DeepSeek-R1-Distill-Qwen-14B2026.0665.66,465
DynasorModel=Qwen3-32B2026.0464.49,518
DynasorModel=QwQ-32B2026.0464.49,733
VanillaModel=DeepSeek-R1-Distill-Qwen-14B2026.0664.411,099
S-GRPOModel=DeepSeek-R1-Distill-Qwen-14B2026.0664.46,712
DEERModel=Qwen3-4B2026.0463.39,327
DEERModel=Qwen3-4B2026.0463.39,327
ReLATModel=Qwen3-4B2026.0663.3—
SingleModel=Qwen3-14B2026.0663.3—
TextMASModel=Qwen3-14B2026.0663.3—
MERABackbone=DeepSeek-R1-Distill-Qwen-14B, Zero-shot Chain-of-Thought (CoT)=true2025.0863.36,695
RAGBackbone=Qwen3-4B2025.0962.9610,335.7
RoT + TIBackbone=Qwen3-4B2025.0962.969,481.59
CoTBackbone=Qwen3-8B2025.0962.9611,232.55
RoT + TIBackbone=Qwen3-14B2025.0962.9611,201.03
DynasorModel=Qwen3-8B2026.0462.210,174
NoThinkingModel=QwQ-32B2026.0462.211,688
DEERModel=Qwen3-8B2026.0461.78,796
DEERModel=Qwen3-8B2026.0461.78,796
Short-RLModel=DeepSeek-R1-Distill-Qwen-14B2026.0661.77,561
CGRSModel=Qwen3-8B2026.0461.18,792
TALEModel=QwQ-32B2026.0461.110,888
CGRSModel=Qwen3-8B2026.0461.18,792
Qwen3 + DAPOSize=14B2026.0660.94—
DEERBackbone=DeepSeek-R1-Distill-Qwen-14B, Zero-shot Chain-of-Thought (CoT)=true2025.0860.78,388
OriginalBackbone=DeepSeek-R1-Distill-Qwen-14B, Zero-shot Chain-of-Thought (CoT)=true2025.0860.611,155
ThinkSwitcherModel=DS-Qwen-14B2026.0460.48,044
ThinkSwitcherModel=DS-Qwen-14B2026.0460.48,044
COTModel=Qwen3-4B2026.046012,662
SATModel=Qwen3-4B2026.04609,462
COTModel=DS-Qwen-14B2026.046010,343
SATModel=DS-Qwen-14B2026.04608,393
SATModel=Nemo-Llama-8b2026.04609,258
COTModel=Qwen3-4B2026.046012,662
SATModel=Qwen3-4B2026.04609,462
COTModel=DS-Qwen-14B2026.046010,343
SATModel=DS-Qwen-14B2026.04608,393
SATModel=Nemo-Llama-8b2026.04609,258
RoT + TIBackbone=Qwen3-8B2025.0959.2610,930.03