ResearchBenchmarksCode Generation on BigCodeBench (Success Rate Metrics)Follow59.5Last Epoch Success RateMemRL47.0250.2653.556.74Jan 6, 2026Evaluation ResultsMethodMethodLinksLast Epoch Success RateCumulative Success Rate (CSR)MemRLModel=GPT-4oModel=GPT-4o2026.0159.50.627MemPModel=GPT-4oModel=GPT-4o2026.0157.80.602Self-RAGModel=GPT-4oModel=GPT-4o2026.0149.70.561Mem0Model=GPT-4oModel=GPT-4o2026.0148.70.495No MemoryModel=GPT-4oModel=GPT-4o2026.0148.5—RAGModel=GPT-4oModel=GPT-4o2026.0147.50.483Pass@10Model=GPT-4oModel=GPT-4o2026.01—0.577