ResearchBenchmarksCode Generation on MBPP (Accuracy, Perplexity)Follow57.4AccuracyMAGS54.48855.2445656.756May 20, 2026Evaluation ResultsMethodMethodLinksAccuracyPerplexityMAGSBase Model=Llama-3.1-8...Base Model=Llama-3.1-8B-Instruct2026.0557.42.97UnsteeredBase Model=Llama-3.1-8...Base Model=Llama-3.1-8B-Instruct2026.0556.22.978CDBase Model=Llama-3.1-8...Base Model=Llama-3.1-8B-Instruct2026.0555.52.291ITIBase Model=Llama-3.1-8...Base Model=Llama-3.1-8B-Instruct2026.0554.82.25ASBase Model=Llama-3.1-8...Base Model=Llama-3.1-8B-Instruct2026.0554.62.214