List your company

Mathematical Reasoning Accuracy (%) on OlympiadBench

56.7AccuracyFlowTracer
20.92430.21239.548.788May 12, 2026May 16, 2026May 21, 2026May 26, 2026May 30, 2026Jun 4, 2026Jun 9, 2026

Evaluation results

MethodLinks
FlowTracerBackbone=Qwen3-8B-Base, Context Length=8K2026.0656.7
AttentionBackbone=Qwen3-8B-Base, Context Length=8K2026.0655.5
RandomBackbone=Qwen3-8B-Base, Context Length=8K2026.0655.3
GRPOBackbone=Qwen3-8B-Base, Context Length=8K2026.0654.4
GradientBackbone=Qwen3-8B-Base, Context Length=8K2026.0653.4
CorrelationBackbone=Qwen3-8B-Base, Context Length=8K2026.0653.2
High-entropyBackbone=Qwen3-8B-Base, Context Length=8K2026.0653.1
FlowTracerBackbone=Qwen3-4B-Base, Context Length=8K2026.0653
AttentionBackbone=Qwen3-4B-Base, Context Length=8K2026.0651.5
GradientBackbone=Qwen3-4B-Base, Context Length=8K2026.0651.3
RandomBackbone=Qwen3-4B-Base, Context Length=8K2026.0650
GRPOBackbone=Qwen3-4B-Base, Context Length=8K2026.0649.9
High-entropyBackbone=Qwen3-4B-Base, Context Length=8K2026.0648.6
CorrelationBackbone=Qwen3-4B-Base, Context Length=8K2026.0648.6
FlowTracerBackbone=Qwen3-8B-Base, Context Length=1K2026.0646.7
High-entropyBackbone=Qwen3-8B-Base, Context Length=1K2026.0645.9
AttentionBackbone=Qwen3-8B-Base, Context Length=1K2026.0645.9
GradientBackbone=Qwen3-8B-Base, Context Length=1K2026.0645
CorrelationBackbone=Qwen3-8B-Base, Context Length=1K2026.0644.8
FlowTracerBackbone=Qwen3-4B-Base, Context Length=1K2026.0644.2
GRPOBackbone=Qwen3-8B-Base, Context Length=1K2026.0644.2
GradientBackbone=Qwen3-4B-Base, Context Length=1K2026.0643.3
RandomBackbone=Qwen3-8B-Base, Context Length=1K2026.0643.3
AttentionBackbone=Qwen3-4B-Base, Context Length=1K2026.0643.1
GRPOBackbone=Qwen3-4B-Base, Context Length=1K2026.0642.8
High-entropyBackbone=Qwen3-4B-Base, Context Length=1K2026.0642.8
RandomBackbone=Qwen3-4B-Base, Context Length=1K2026.0642
CorrelationBackbone=Qwen3-4B-Base, Context Length=1K2026.0642
PAPOBase Model=Qwen2.5-14B-Base2026.0539.22
DAPOBase Model=Qwen2.5-14B-Base2026.0538.57
Ent-RegBase Model=Qwen2.5-14B-Base2026.0538.34
80/20Base Model=Qwen2.5-14B-Base2026.0537.59
PAPOBase Model=Qwen2.5-7B-Base2026.0534.7
DAPOBase Model=Qwen2.5-7B-Base2026.0533.12
Ent-RegBase Model=Qwen2.5-7B-Base2026.0533.08
80/20Base Model=Qwen2.5-7B-Base2026.0532.69
BaseBase Model=Qwen2.5-14B-Base2026.0526.04
BaseBase Model=Qwen2.5-7B-Base2026.0522.3