ResearchBenchmarksScience QA on GPQA Diamond (test)Follow60.83Success RateMemRL46.966850.565954.16557.7641May 8, 2026Evaluation ResultsMethodMethodLinksSuccess RateMemRLModel=Gemma-4-E4B-itModel=Gemma-4-E4B-it2026.0560.83MemQModel=Gemma-4-E4B-itModel=Gemma-4-E4B-it2026.0560.83MemPModel=Gemma-4-E4B-itModel=Gemma-4-E4B-it2026.0559.17RAGModel=Gemma-4-E4B-itModel=Gemma-4-E4B-it2026.0558.33Self-RAGModel=Gemma-4-E4B-itModel=Gemma-4-E4B-it2026.0558.33Mem0Model=Gemma-4-E4B-itModel=Gemma-4-E4B-it2026.0558.33No Mem.Model=Gemma-4-E4B-itModel=Gemma-4-E4B-it2026.0547.5