List your company

Science on GPQA

85.7AccuracyGPT-5
-3.42819.71142.8565.989May 22, 2025Jun 29, 2025Aug 7, 2025Sep 15, 2025Oct 23, 2025Dec 1, 2025Jan 9, 2026

Evaluation results

MethodLinks
GPT-5Evaluation Protocol=Closed-Source2026.0185.7
Gemini2.5-ProEvaluation Protocol=Closed-Source2026.0184
GPT-4.1Evaluation Protocol=Closed-Source2026.0162.1
DyLANBackbone=Llama-3.3-70B-Instruct2025.0554
CoTBackbone=Llama-3.3-70B-Instruct2025.0553
SCBackbone=Llama-3.3-70B-Instruct2025.0552.5
DebateBackbone=Llama-3.3-70B-Instruct2025.0551
AgentVerseBackbone=Llama-3.3-70B-Instruct2025.0551
MacNetBackbone=Llama-3.3-70B-Instruct2025.0551
MADBackbone=Llama-3.3-70B-Instruct2025.0550
ReasoningBackbone=Qwen2.5-7B family2026.0149.1
SCBackbone=Qwen-2.5-72B-Instruct2025.0549
DebateBackbone=Qwen-2.5-72B-Instruct2025.0549
MAS-GPTBackbone=Qwen-2.5-72B-Instruct2025.0549
MAS-GPTBackbone=Llama-3.3-70B-Instruct2025.0548.5
SingleBackbone=Llama-3.3-70B-Instruct2025.0548
AFlow-MathBackbone=Llama-3.3-70B-Instruct2025.0548
AFlow-MathBackbone=Qwen-2.5-72B-Instruct2025.0547.5
CoTBackbone=Qwen-2.5-72B-Instruct2025.0547
MADBackbone=Qwen-2.5-72B-Instruct2025.0547
MAVBackbone=Qwen-2.5-72B-Instruct2025.0546.5
ATLAS (cluster)Evaluation Protocol=In-Distribution2026.0146.4
AgentVerseBackbone=Qwen-2.5-72B-Instruct2025.0546
MAVBackbone=Llama-3.3-70B-Instruct2025.0545
SingleBackbone=Qwen-2.5-72B-Instruct2025.0545
GPT-4oEvaluation Protocol=Closed-Source2026.0144.4
AutoGenBackbone=Qwen-2.5-72B-Instruct2025.0544.4
DyLANBackbone=Qwen-2.5-72B-Instruct2025.0543.4
MacNetBackbone=Qwen-2.5-72B-Instruct2025.0542.4
ATLAS (RL)Evaluation Protocol=Out-of-Distribution2026.0142
AutoGenBackbone=Llama-3.3-70B-Instruct2025.0541.9
ReasonAnyBackbone=Qwen2.5-7B family2026.0141.06
FinanceBackbone=Qwen2.5-7B family2026.0139.39
LinearBackbone=Qwen2.5-7B family2026.0139.39
RouterDCEvaluation Protocol=In-Distribution2026.0139.1
LEDBackbone=Qwen2.5-7B family2026.0138.64
BertRouterEvaluation Protocol=In-Distribution2026.0136.6
MLPRouterEvaluation Protocol=In-Distribution2026.0134.8
Task ArithmeticBackbone=Qwen2.5-7B family2026.0131.82
Random RouterEvaluation Protocol=Training-free2026.0130.6
TIESBackbone=Qwen2.5-7B family2026.0129.55
ATLAS (cluster)Evaluation Protocol=Out-of-Distribution2026.0129
RouterDCEvaluation Protocol=Out-of-Distribution2026.0128.6
DAREBackbone=Qwen2.5-7B family2026.0128.03
BertRouterEvaluation Protocol=Out-of-Distribution2026.0127.7
MLPRouterEvaluation Protocol=Out-of-Distribution2026.0126.8
FS RouterEvaluation Protocol=Training-free, Prompting Strategy=Few-shot2026.0125.9
ZS RouterEvaluation Protocol=Training-free, Prompting Strategy=Zero-shot2026.0124.6
FuseLLMBackbone=Qwen2.5-7B family2026.010