BenchmarkAgent: Turn Your Open-Source LLM into a Reproducible, Deployable, Monetizable Super-Agent

I built a custom LLM architecture with self-correction and long-term memory vector states using phi-3-mini, finetuned it to achieve 98.17% on HumanEval, and open-sourced the model at https://huggingface.co/moelanoby/phi-3-M3-coder while asking for recommendations for other lightweight benchmarks.