Submissions from twitter.com/jxmnop

		Reversing the alignment part of LLM training (twitter.com/jxmnop)
		3 points by fzliu 4 months ago \| past
		Unaligned GPT-OSS-20B-base extracted from OpenAI's model (twitter.com/jxmnop)
		1 point by fragmede 4 months ago \| past
		GPT-OSS-20B extracted to a base model without alignment (twitter.com/jxmnop)
		3 points by polyrand 4 months ago \| past \| 2 comments
		Curious about the training data of OpenAI's new GPT-OSS models? I was too (twitter.com/jxmnop)
		239 points by flabber 4 months ago \| past \| 57 comments
		Curious about the training data of OpenAI's new GPT-OSS models? I was too (twitter.com/jxmnop)
		4 points by tosh 4 months ago \| past
		Reverse Engineering training data of OpenAI's new GPT-OSS models (twitter.com/jxmnop)
		1 point by amrrs 4 months ago \| past
		All Embedding Models Learn the Same Thing (twitter.com/jxmnop)
		1 point by MrBuddyCasino 7 months ago \| past
		The paper that got LLM started. It was in 2003, in Montreal (twitter.com/jxmnop)
		6 points by ksec 7 months ago \| past
		Prompting LLMs to Learn Language from a Single Book (twitter.com/jxmnop)
		2 points by abetusk on June 26, 2024 \| past \| 1 comment
		Training open-source LLMs is a losing battle, a complete dead end (twitter.com/jxmnop)
		3 points by jxmorris12 on Sept 15, 2023 \| past