List your company

Code Generation on HumanE

85.2AccuracyDenser
10.538429.921749.30568.6883Sep 26, 2025Oct 9, 2025Oct 23, 2025Nov 6, 2025Nov 19, 2025Dec 3, 2025Dec 17, 2025

Evaluation results

MethodLinks
DenserBackbone=Qwen3-32B-think2025.1285.2
Process SupervisionBackbone=Qwen3-32B-think2025.1284.8
Reflection-CoTBackbone=Qwen3-32B-think2025.1284.2
DenserBackbone=Qwen3-32B2025.1283.6
Self-VerificationBackbone=Qwen3-32B-think2025.1283.1
Tree-of-ThoughtBackbone=Qwen3-32B-think2025.1283
Process SupervisionBackbone=Qwen3-32B2025.1282.7
Self-ConsistencyBackbone=Qwen3-32B-think2025.1282.4
Reflection-CoTBackbone=Qwen3-32B2025.1282.1
Think-to-ThinkBackbone=Qwen3-32B-think2025.1281.7
Self-VerificationBackbone=Qwen3-32B2025.1281.4
Tree-of-ThoughtBackbone=Qwen3-32B2025.1280.9
Self-ConsistencyBackbone=Qwen3-32B2025.1280.3
Chain-of-ThoughtBackbone=Qwen3-32B-think2025.1279.7
Think-to-ThinkBackbone=Qwen3-32B2025.1279.6
DenserBackbone=Qwen3-8B-think2025.1279.1
Process SupervisionBackbone=Qwen3-8B-think2025.1278.5
Reflection-CoTBackbone=Qwen3-8B-think2025.1277.9
Chain-of-ThoughtBackbone=Qwen3-32B2025.1277.8
DenserBackbone=Qwen3-8B2025.1277.3
Self-VerificationBackbone=Qwen3-8B-think2025.1276.9
Tree-of-ThoughtBackbone=Qwen3-8B-think2025.1276.8
DenserBackbone=Qwen3-14B-no-think2025.1276.8
Process SupervisionBackbone=Qwen3-8B2025.1276.3
Self-ConsistencyBackbone=Qwen3-8B-think2025.1276.2
DenserBackbone=Qwen3-4B-think2025.1276.1
Process SupervisionBackbone=Qwen3-14B-no-think2025.1276
Reflection-CoTBackbone=Qwen3-8B2025.1275.7
EMoEBase Model=ERNIE-4.5-21B-A3B, Trained Activated Experts=6+2, MoE Routing Method=EMoE, Inference Budget (k')=12 + 22025.0975.61
Think-to-ThinkBackbone=Qwen3-8B-think2025.1275.5
Reflection-CoTBackbone=Qwen3-14B-no-think2025.1275.4
Process SupervisionBackbone=Qwen3-4B-think2025.1275.2
Self-VerificationBackbone=Qwen3-8B2025.1274.8
Reflection-CoTBackbone=Qwen3-4B-think2025.1274.6
Self-VerificationBackbone=Qwen3-14B-no-think2025.1274.6
Tree-of-ThoughtBackbone=Qwen3-8B2025.1274.5
Tree-of-ThoughtBackbone=Qwen3-14B-no-think2025.1274.2
Self-ConsistencyBackbone=Qwen3-8B2025.1273.9
Chain-of-ThoughtBackbone=Qwen3-8B-think2025.1273.8
Self-VerificationBackbone=Qwen3-4B-think2025.1273.7
DenserBackbone=Qwen3-4B2025.1273.6
Self-ConsistencyBackbone=Qwen3-14B-no-think2025.1273.6
Tree-of-ThoughtBackbone=Qwen3-4B-think2025.1273.4
Think-to-ThinkBackbone=Qwen3-8B2025.1273.2
EMoEBase Model=ERNIE-4.5-21B-A3B, Trained Activated Experts=6+2, MoE Routing Method=EMoE, Inference Budget (k')=6 + 22025.0973.17
Think-to-ThinkBackbone=Qwen3-14B-no-think2025.1272.9
Self-ConsistencyBackbone=Qwen3-4B-think2025.1272.8
Process SupervisionBackbone=Qwen3-4B2025.1272.7
Reflection-CoTBackbone=Qwen3-4B2025.1272.1
Think-to-ThinkBackbone=Qwen3-4B-think2025.1272.1
Chain-of-ThoughtBackbone=Qwen3-8B2025.1271.5
Top-kBase Model=ERNIE-4.5-21B-A3B, Trained Activated Experts=6+2, MoE Routing Method=Top-k, Inference Budget (k')=6 + 22025.0971.34
Self-VerificationBackbone=Qwen3-4B2025.1271.3
Chain-of-ThoughtBackbone=Qwen3-14B-no-think2025.1271.2
Tree-of-ThoughtBackbone=Qwen3-4B2025.1271
EMoEBase Model=ERNIE-4.5-21B-A3B, Trained Activated Experts=6+2, MoE Routing Method=EMoE, Inference Budget (k')=3 + 22025.0970.73
Self-ConsistencyBackbone=Qwen3-4B2025.1270.5
Chain-of-ThoughtBackbone=Qwen3-4B-think2025.1270.4
Think-to-ThinkBackbone=Qwen3-4B2025.1269.7
Top-kBase Model=ERNIE-4.5-21B-A3B, Trained Activated Experts=6+2, MoE Routing Method=Top-k, Inference Budget (k')=12 + 22025.0968.29
Chain-of-ThoughtBackbone=Qwen3-4B2025.1268.2
Top-kBase Model=ERNIE-4.5-21B-A3B, Trained Activated Experts=6+2, MoE Routing Method=Top-k, Inference Budget (k')=3 + 22025.0966.46
EMoEBase Model=DeepSeek-V2-Lite, Trained Activated Experts=6+2, MoE Routing Method=EMoE, Inference Budget (k')=12 + 22025.0941.46
EMoEBase Model=DeepSeek-V2-Lite, Trained Activated Experts=6+2, MoE Routing Method=EMoE, Inference Budget (k')=6 + 22025.0939.63
Top-kBase Model=DeepSeek-V2-Lite, Trained Activated Experts=6+2, MoE Routing Method=Top-k, Inference Budget (k')=12 + 22025.0939.02
Top-kBase Model=DeepSeek-V2-Lite, Trained Activated Experts=6+2, MoE Routing Method=Top-k, Inference Budget (k')=6 + 22025.0936.59
EMoEBase Model=DeepSeek-V2-Lite, Trained Activated Experts=6+2, MoE Routing Method=EMoE, Inference Budget (k')=3 + 22025.0934.76
Top-kBase Model=DeepSeek-V2-Lite, Trained Activated Experts=6+2, MoE Routing Method=Top-k, Inference Budget (k')=3 + 22025.0928.66
EMoEBase Model=OLMoE-1B-7B-0924, Trained Activated Experts=8, MoE Routing Method=EMoE, Inference Budget (k')=162025.0921.95
Top-kBase Model=OLMoE-1B-7B-0924, Trained Activated Experts=8, MoE Routing Method=Top-k, Inference Budget (k')=82025.0921.34
EMoEBase Model=LoRAMoE, Trained Activated Experts=2, MoE Routing Method=EMoE, Inference Budget (k')=62025.0920.73
Top-kBase Model=LoRAMoE, Trained Activated Experts=2, MoE Routing Method=Top-k, Inference Budget (k')=22025.0920.24
EMoEBase Model=OLMoE-1B-7B-0924, Trained Activated Experts=8, MoE Routing Method=EMoE, Inference Budget (k')=82025.0920.12
Top-kBase Model=LoRAMoE, Trained Activated Experts=2, MoE Routing Method=Top-k, Inference Budget (k')=42025.0919.39
EMoEBase Model=LoRAMoE, Trained Activated Experts=2, MoE Routing Method=EMoE, Inference Budget (k')=22025.0918.9
Top-kBase Model=LoRAMoE, Trained Activated Experts=2, MoE Routing Method=Top-k, Inference Budget (k')=62025.0918.7
EMoEBase Model=LoRAMoE, Trained Activated Experts=2, MoE Routing Method=EMoE, Inference Budget (k')=42025.0918.29
EMoEBase Model=LoRAMoE, Trained Activated Experts=2, MoE Routing Method=EMoE, Inference Budget (k')=12025.0917.68
Top-kBase Model=OLMoE-1B-7B-0924, Trained Activated Experts=8, MoE Routing Method=Top-k, Inference Budget (k')=162025.0917.07
Top-kBase Model=LoRAMoE, Trained Activated Experts=2, MoE Routing Method=Top-k, Inference Budget (k')=12025.0915.97
EMoEBase Model=OLMoE-1B-7B-0924, Trained Activated Experts=8, MoE Routing Method=EMoE, Inference Budget (k')=42025.0915.85
Top-kBase Model=OLMoE-1B-7B-0924, Trained Activated Experts=8, MoE Routing Method=Top-k, Inference Budget (k')=42025.0913.41