October 21, 2025 front

ML Newsnew | past | best | rss | submit

Stories from October 21, 2025 (UTC)

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

1.	Why can't transformers learn multiplication? (arxiv.org)
	161 points by PaulHoule 348 days ago \| hide \| past \| pdf \| 107 comments
2.	Binary Retrieval-Augmented Reward Mitigates Hallucinations (arxiv.org)
	44 points by MarlonPro 348 days ago \| hide \| past \| pdf \| 3 comments
3.	OptPipe: Memory- and Scheduling-Optimized Pipeline Parallelism for LLM Training (arxiv.org)
	11 points by PaulHoule 348 days ago \| hide \| past \| pdf \| discuss
4.	Prompt Baking (arxiv.org)
	8 points by jxmorris12 348 days ago \| hide \| past \| pdf \| 1 comment
5.	Tensor Logic: The Language of AI (arxiv.org)
	5 points by fofoz 348 days ago \| hide \| past \| pdf \| discuss
6.	Reasoning with Sampling: Your Base Model Is Smarter Than You Think (arxiv.org)
	3 points by Anon84 348 days ago \| hide \| past \| pdf \| discuss
7.	Evaluating Agentic Cybersecurity in Attack/Defense CTFs: Offensive Is Not Better (arxiv.org)
	2 points by vmayoral 348 days ago \| hide \| past \| pdf \| 1 comment
8.	Qwen Language Confusion Gate (arxiv.org)
	2 points by CollinZ 348 days ago \| hide \| past \| pdf \| discuss
9.	Knowledge Transfer from High-Resource to Low-Resource Languages for Code LLMs (2023) (arxiv.org)
	1 point by peatmoss 347 days ago \| hide \| past \| pdf \| discuss
10.	Explaining Why Hallucinate Large Language Models (arxiv.org)
	1 point by gagan30 348 days ago \| hide \| past \| pdf \| 1 comment
11.	Verifiable ML Without Determinism: Tolerance-Aware Optimistic Verification (arxiv.org)
	1 point by cnyaojz 348 days ago \| hide \| past \| pdf \| 1 comment