List your company

Memory reasoning on Splitwise

95.2Accuracyxmemory
8.67231.13653.676.064Apr 30, 2026

Evaluation results

MethodLinks
xmemoryEvaluation protocol=single-answer, Judge type=LLM as a judge2026.0495.2
Frontier-model applicationsType=customer-facing memory and tool-use harnesses2026.0492
SupermemoryEvaluation protocol=single-answer, Judge type=LLM as a judge2026.0473.75
CogneeEvaluation protocol=single-answer, Judge type=LLM as a judge2026.0468
Mem0Graph usage=true, Evaluation protocol=single-answer, Judge type=LLM as a judge2026.0459.1
Mem0Graph usage=false, Evaluation protocol=single-answer, Judge type=LLM as a judge2026.0454.9
Markdown harness (balance-only)Type=File-based Markdown, Model=frontier model2026.0440
ZepEvaluation protocol=single-answer, Judge type=LLM as a judge2026.0425.7
Markdown harness (transaction-preserving)Type=File-based Markdown, Model=frontier model2026.0412