November 20, 2025 front

ML Newsnew | past | best | rss | submit

Stories from November 20, 2025 (UTC)

Go back a day, month, or year. Go forward a day.

1.	Adversarial poetry as a universal single-turn jailbreak mechanism in LLMs (arxiv.org)
	384 points by capgre 318 days ago \| hide \| past \| pdf \| 189 comments
2.	The Psychogenic Machine: Simulating AI Psychosis (arxiv.org)
	4 points by indiantinker 318 days ago \| hide \| past \| pdf \| discuss
3.	Token embeddings violate the manifold hypothesis (arxiv.org)
	4 points by airstrike 318 days ago \| hide \| past \| pdf \| discuss
4.	Parallel Loop Transformer for Efficient Test-Time Computation Scaling (arxiv.org)
	3 points by PaulHoule 318 days ago \| hide \| past \| pdf \| discuss
5.	Debiasing Reward Models by Representation Learning with Guarantees (arxiv.org)
	3 points by PaulHoule 318 days ago \| hide \| past \| pdf \| discuss
6.	A Style is Worth One Code: open-source Midjourey-like --sref (arxiv.org)
	2 points by meander_water 318 days ago \| hide \| past \| pdf \| discuss
7.	Practice on Long Behavior Sequence Modeling in Tencent Advertising (arxiv.org)
	1 point by PaulHoule 318 days ago \| hide \| past \| pdf \| discuss