Many published academic papers on Recursive Self-Improvement (RSI) fail to replicate and their results are often noise or cherry-picked