June 18, 2025 front

ML Newsnew | past | best | rss | submit

Stories from June 18, 2025 (UTC)

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

1.	Reasoning by Superposition: A Perspective on Chain of Continuous Thought (arxiv.org)
	60 points by danielmorozoff on Jun 18, 2025 \| hide \| past \| pdf \| 1 comment
2.	Style over Substance: Distilled Language Models Reason via Stylistic Replication (arxiv.org)
	5 points by curtsmith on Jun 18, 2025 \| hide \| past \| pdf \| discuss
3.	S1: Simple Test-Time Scaling (arxiv.org)
	3 points by bicepjai on Jun 18, 2025 \| hide \| past \| pdf \| discuss
4.	Self-Supervised Contrastive Learning Approximates Supervised CL (arxiv.org)
	3 points by PaulHoule on Jun 18, 2025 \| hide \| past \| pdf \| discuss
5.	Robustly Improving LLM Fairness in Realistic Settings via Interpretability (arxiv.org)
	2 points by scribu on Jun 18, 2025 \| hide \| past \| pdf \| discuss
6.	AbstentionBench: Reasoning LLMs Fail on Unanswerable Questions (arxiv.org)
	2 points by fzliu on Jun 18, 2025 \| hide \| past \| pdf \| discuss
7.	SageAttention3: Microscaling FP4 Attention. 5x Speed up (arxiv.org)
	2 points by 0xjunhao on Jun 18, 2025 \| hide \| past \| pdf \| discuss
8.	Large Language Models – The Future of Fundamental Physics? (arxiv.org)
	1 point by matteocantiello on Jun 18, 2025 \| hide \| past \| pdf \| discuss