List your company

Counterfactual Reasoning on MMLU-CF

71.6EMGHS-TDA
64.94466.67268.470.128Feb 10, 2026

Evaluation results

MethodLinks
GHS-TDABackbone=GPT-4o-mini2026.0271.6
GHS-TDABackbone=DeepSeek-V32026.0271.6
CoT-SCBackbone=DeepSeek-V3, Reasoning Paradigm=chain-based, n=52026.0271.5
GHS-TDABackbone=Qwen-Turbo2026.0271.2
CoT-SCBackbone=GPT-4o-mini, Reasoning Paradigm=chain-based, n=52026.0271.1
AoTBackbone=GPT-4o-mini, Reasoning Paradigm=atomic reasoning2026.0270.9
CoT-SCBackbone=Qwen-Turbo, Reasoning Paradigm=chain-based, n=52026.0270.8
AoTBackbone=DeepSeek-V3, Reasoning Paradigm=atomic reasoning2026.0270.8
FoTBackbone=GPT-4o-mini, Reasoning Paradigm=forest-based, n=82026.0270.6
AoTBackbone=Qwen-Turbo, Reasoning Paradigm=atomic reasoning2026.0270.5
FoTBackbone=DeepSeek-V3, Reasoning Paradigm=forest-based, n=82026.0270.5
FoTBackbone=Qwen-Turbo, Reasoning Paradigm=forest-based, n=82026.0270.4
ToTBackbone=DeepSeek-V3, Reasoning Paradigm=tree-based2026.0270.4
GoTBackbone=DeepSeek-V3, Reasoning Paradigm=graph-based2026.0270.3
GoTBackbone=GPT-4o-mini, Reasoning Paradigm=graph-based2026.0270.2
GoTBackbone=Qwen-Turbo, Reasoning Paradigm=graph-based2026.0270.1
Self-RefineBackbone=DeepSeek-V3, Reasoning Paradigm=chain-based2026.0270.1
ToTBackbone=GPT-4o-mini, Reasoning Paradigm=tree-based2026.0269.9
CoTBackbone=DeepSeek-V3, Reasoning Paradigm=chain-based2026.0269.9
AFlowBackbone=DeepSeek-V3, Reasoning Paradigm=search-based2026.0269.8
Self-RefineBackbone=GPT-4o-mini, Reasoning Paradigm=chain-based2026.0269.7
CoTBackbone=GPT-4o-mini, Reasoning Paradigm=chain-based2026.0269.6
ToTBackbone=Qwen-Turbo, Reasoning Paradigm=tree-based2026.0269.6
AFlowBackbone=GPT-4o-mini, Reasoning Paradigm=search-based2026.0269.5
Self-RefineBackbone=Qwen-Turbo, Reasoning Paradigm=chain-based2026.0269.5
CoTBackbone=Qwen-Turbo, Reasoning Paradigm=chain-based2026.0269.4
AFlowBackbone=Qwen-Turbo, Reasoning Paradigm=search-based2026.0269.3
Analogical PromptingBackbone=DeepSeek-V3, Reasoning Paradigm=chain-based2026.0266.1
Analogical PromptingBackbone=GPT-4o-mini, Reasoning Paradigm=chain-based2026.0265.8
Analogical PromptingBackbone=Qwen-Turbo, Reasoning Paradigm=chain-based2026.0265.2