Recursive self-training in code LLMs can degrade model quality when AI-generated code loops back into training data faster than humans can review it. The paper compares no-review, human-style filter gates, and…

#AI #CodeLLMs #PromptEngineering #RecursiveTraining
https://arxiv.org/abs/2606.28438

New arXiv paper defines three distinct signals for synthetic pretraining tasks—diagnostic, teachable, and transferable—and finds only 14 of 27 tasks are teachable under controlled Python pretraining. Useful framing for…

#OpenSourceAI #LLM #CodeLLMs #Pretraining
https://arxiv.org/abs/2610.11548