List your company
ResearchFormerly WizWand

Mathematical Reasoning on Olympiad (Avg@16 Accuracy)

70.7Avg@16 AccuracyResRL
12.56427.65742.7557.843May 1, 2026May 5, 2026May 9, 2026May 14, 2026May 18, 2026May 22, 2026May 27, 2026
Updated 5mo ago

Evaluation results

MethodLinks
ResRLBase Model=Qwen3-32B, Mode=Think mode, Max response length=81922026.0570.7
FlowRLBackbone=Qwen3-8B, Max response length=40962026.0568.5
ResRLBackbone=Qwen3-8B, Max response length=40962026.0568.1
IRDSModel=QWEN3-4B, Data Budget=30%2026.0567.2
NSR (Weighted-Reinforce)Base Model=Qwen3-32B, Mode=Think mode, Max response length=81922026.0564.4
DEPOModel=QWEN3-4B, Data Budget=30%2026.0563.4
LIMRModel=QWEN3-4B, Data Budget=30%2026.0563
PPL-MIDDLEModel=QWEN3-4B, Data Budget=30%2026.0562.9
GRPOBackbone=Qwen3-8B, Max response length=40962026.0562
PPL-TOPModel=QWEN3-4B, Data Budget=30%2026.0561.4
NSR (Weighted-Reinforce)Backbone=Qwen3-8B, Max response length=40962026.0560.6
IFDModel=QWEN3-4B, Data Budget=30%2026.0559
IRDSModel=QWEN3-4B, Data Budget=10%2026.0558.4
FlowRLBackbone=Qwen3-4B, Max response length=40962026.0558.1
ResRLBackbone=Qwen3-1.7B, Max response length=40962026.0556.6
GRPOBackbone=Qwen3-4B, Max response length=40962026.0555.1
RANDOMModel=QWEN3-4B, Data Budget=30%2026.0554.8
NSR (Weighted-Reinforce)Backbone=Qwen3-1.7B, Max response length=40962026.0553.5
IRDSModel=QWEN3-1.7B, Data Budget=30%2026.0553.4
ResRLBackbone=Qwen3-4B, Max response length=40962026.0552.3
TOKEN-LENGTHModel=QWEN3-4B, Data Budget=30%2026.0552.2
DAPOBackbone=Qwen3-4B, Max response length=40962026.0551.2
NSR (Weighted-Reinforce)Backbone=Qwen3-4B, Max response length=40962026.0550.1
LIMRModel=QWEN3-1.7B, Data Budget=30%2026.0549.7
IRDSModel=QWEN3-1.7B, Data Budget=10%2026.0549.1
DEPOModel=QWEN3-1.7B, Data Budget=30%2026.0549.1
FlowRLBackbone=Qwen3-1.7B, Max response length=40962026.0548.6
Qwen3-8B BackboneBackbone=Qwen3-8B, Max response length=40962026.0548.6
Qwen3-4B BackboneBackbone=Qwen3-4B, Max response length=40962026.0548.2
PPL-MIDDLEModel=QWEN3-1.7B, Data Budget=30%2026.0546.1
RANDOMModel=QWEN3-1.7B, Data Budget=30%2026.0546
PPL-TOPModel=QWEN3-1.7B, Data Budget=30%2026.0545
DAPOBackbone=Qwen3-8B, Max response length=40962026.0543.6
IFDModel=QWEN3-1.7B, Data Budget=30%2026.0542.8
TOKEN-LENGTHModel=QWEN3-1.7B, Data Budget=30%2026.0540.9
Qwen3-1.7B BackboneBackbone=Qwen3-1.7B, Max response length=40962026.0538.1
GRPOBackbone=Qwen3-1.7B, Max response length=40962026.0536
DAPOBackbone=Qwen3-1.7B, Max response length=40962026.0533.6
IRDSModel=LLAMA-3.1-8B-INSTRUCT, Data Budget=30%2026.0516.8
RANDOMModel=LLAMA-3.1-8B-INSTRUCT, Data Budget=30%2026.0516.1
LIMRModel=LLAMA-3.1-8B-INSTRUCT, Data Budget=30%2026.0516.1
DEPOModel=LLAMA-3.1-8B-INSTRUCT, Data Budget=30%2026.0515.6
TOKEN-LENGTHModel=LLAMA-3.1-8B-INSTRUCT, Data Budget=30%2026.0515.4
IRDSModel=LLAMA-3.1-8B-INSTRUCT, Data Budget=10%2026.0515.3
PPL-MIDDLEModel=LLAMA-3.1-8B-INSTRUCT, Data Budget=30%2026.0515.3
PPL-TOPModel=LLAMA-3.1-8B-INSTRUCT, Data Budget=30%2026.0515.2
IFDModel=LLAMA-3.1-8B-INSTRUCT, Data Budget=30%2026.0514.8