List your company
ResearchFormerly WizWand

Code Generation on CodeEval-Pro BigCodeBench-Lite-Pro and HumanEval-Pro (1st Pass)

66.7Average AccuracyMEMPROBE
34.14842.59951.0559.501Jun 1, 2026
Updated 5mo ago

Evaluation results

MethodLinks
MEMPROBEStream Design=Compositional2026.0666.7
ExpRAGStream Design=Compositional2026.0662.5
Self-invokingStream Design=Compositional2026.0660.4
Self-invoking (w/ subtask solution)Stream Design=Compositional2026.0659
ReMemStream Design=Compositional2026.0658.3
DC-RSStream Design=Compositional2026.0648.6
ReActStream Design=Compositional2026.0644.8
ReActStream Design=Naive2026.0644.8
AWMStream Design=Naive2026.0644.5
LangMemStream Design=Compositional2026.0644.4
Mem0Stream Design=Compositional2026.0643.8
Mem0Stream Design=Naive2026.0643.8
MEMPROBEStream Design=Naive2026.0643.1
ReMemStream Design=Naive2026.0641.7
AWMStream Design=Compositional2026.0639.6
LangMemStream Design=Naive2026.0639.6
ExpRAGStream Design=Naive2026.0639.6
DC-RSStream Design=Naive2026.0635.4