cscurmudgeon's favorites

1.		Do Large Language Models learn world models or just surface statistics? (2023) (thegradient.pub)
		37 points by fragmede 6 hours ago \| 53 comments
2.		OK, I can partly explain the LLM chess weirdness now (dynomight.net)
		369 points by dmazin 1 day ago \| 328 comments
3.		Bayesian Neural Networks (toronto.edu)
		214 points by reqo 4 days ago \| 37 comments
4.		Show HN: Llama 3.2 Interpretability with Sparse Autoencoders (github.com/paulpauls)
		497 points by PaulPauls 22 hours ago \| 71 comments
5.		Nickel Plating Handbook [pdf] (nickelinstitute.org)
		59 points by nativeit 2 days ago \| 19 comments
6.		New textbook teaches students about matrix methods and their real world apps (umich.edu)
		77 points by teleforce 10 days ago \| 33 comments
7.		LLaVA-O1: Let Vision Language Models Reason Step-by-Step (arxiv.org)
		176 points by lnyan 4 days ago \| 32 comments
8.		AlphaProof's Greatest Hits (rishimehta.xyz)
		249 points by rishicomplex 5 days ago \| 132 comments
9.		Biological Miracle – Wood Frog (nps.gov)
		503 points by thunderbong 7 days ago \| 139 comments
10.		Something weird is happening with LLMs and chess (dynomight.substack.com)
		695 points by crescit_eundo 8 days ago \| 474 comments
11.		BERTs Are Generative In-Context Learners (arxiv.org)
		141 points by fzliu 8 days ago \| 45 comments
12.		TinyTroupe, a new LLM-powered multiagent persona simulation Python library (github.com/microsoft)
		143 points by paulosalem 11 days ago \| 48 comments
13.		LoRA vs. Full Fine-Tuning: An Illusion of Equivalence (arxiv.org)
		236 points by timbilt 14 days ago \| 53 comments
14.		Evaluating the world model implicit in a generative model (arxiv.org)
		159 points by dsubburam 15 days ago \| 45 comments
15.		D2: Declarative Diagramming – A modern language that turns text to diagrams (d2lang.com)
		50 points by thunderbong 19 days ago \| 7 comments
16.		TokenFormer: Rethinking Transformer Scaling with Tokenized Model Parameters (arxiv.org)
		174 points by og_kalu 21 days ago \| 33 comments
17.		The National EUV Accelerator comes to Albany (ibm.com)
		43 points by sandwichsphinx 21 days ago \| 45 comments
18.		AI Flame Graphs (brendangregg.com)
		316 points by JNRowe 24 days ago \| 89 comments
19.		Chain-of-thought can hurt performance on tasks where thinking makes humans worse (arxiv.org)
		371 points by benocodes 22 days ago \| 250 comments
20.		Don't implement unification by recursion (philipzucker.com)
		83 points by mathgenius 25 days ago \| 62 comments
21.		Numerical Precision Affects Mathematical Reasoning Capabilities of LLMs (arxiv.org)
		66 points by belter 35 days ago \| 49 comments
22.		The AI Investment Boom (apricitas.io)
		278 points by m-hodges 33 days ago \| 376 comments
23.		Phenomenal consciousness is alien to us: SETI and the Fermi paradox (sciencedirect.com)
		51 points by rbanffy 33 days ago \| 69 comments
24.		Why do random forests work? They are self-regularizing adaptive smoothers (arxiv.org)
		295 points by sebg 35 days ago \| 41 comments
25.		Use Prolog to improve LLM's reasoning (shchegrikovich.substack.com)
		379 points by shchegrikovich 39 days ago \| 155 comments
26.		DeepSeek: Advancing theorem proving in LLMs through large-scale synthetic data (arxiv.org)
		186 points by hhs 39 days ago \| 54 comments
27.		Machine learning and information theory concepts towards an AI Mathematician (arxiv.org)
		109 points by marojejian 41 days ago \| 19 comments
28.		Understanding the Limitations of Mathematical Reasoning in LLMs (arxiv.org)
		282 points by hnhn34 42 days ago \| 266 comments
29.		Google’s AI thinks I left a Gatorade bottle on the moon (edwardbenson.com)
		367 points by gwintrob 46 days ago \| 194 comments
30.		The Data Visualisation Catalogue: find the right method for your data (datavizcatalogue.com)
		295 points by sea-gold 48 days ago \| 34 comments
		More