List your company

Science Question Answering on GPQA D (val)

65.9Accuracy (%)Base
44.89250.34655.861.254Apr 10, 2026

Evaluation results

MethodLinks
BaseBackbone=Qwen3-32B, Generations=642026.0465.91.899.7
ControlBackbone=Qwen3-32B, Generations=642026.0464.62.4515.7
BaseBackbone=Phi-4-r (14B), Generations=642026.0464.10.83.7
ControlBackbone=Phi-4-r (14B), Generations=642026.0464.12.1114.2
Mem. + RLBackbone=Qwen3-8B, Generations=642026.0462.91.249.2
MEMENTOBackbone=Qwen3-32B, Generations=642026.0462.11.077.6
MEMENTOBackbone=Phi-4-r (14B), Generations=642026.0461.60.86.2
BaseBackbone=Qwen3-8B, Generations=642026.0461.41.236.6
ControlBackbone=Qwen3-8B, Generations=642026.0457.81.611.8
MEMENTOBackbone=Qwen3-8B, Generations=642026.0455.80.564
BaseBackbone=Olmo 3 (7B), Generations=642026.0450.83.2130.8
MEMENTOBackbone=Olmo 3 (7B), Generations=642026.0449.52.7225.2
ControlBackbone=Olmo 3 (7B), Generations=642026.0445.72.9423.8