ResearchBenchmarksCode Generation on BigCodeBench (val)Follow50.8Success RateMemRL47.78448.56749.3550.133Jan 6, 2026Evaluation ResultsMethodMethodLinksSuccess RateMemRLModel=GPT-4oModel=GPT-4o2026.0150.8Self-RAGModel=GPT-4oModel=GPT-4o2026.0150MemPModel=GPT-4oModel=GPT-4o2026.0149.4No MemoryModel=GPT-4oModel=GPT-4o2026.0148.5Mem0Model=GPT-4oModel=GPT-4o2026.0148.5RAGModel=GPT-4oModel=GPT-4o2026.0147.9