curius graph

all pages

showing 15301-15350 of 200098 pages (sorted by popularity)
Thinking through how pretraining vs RL learn
Almost anything you give sustained attention to will begin to loop on itself and bloom
The bare-bones case for caring about global catastrophic risks
The paradox is that when I accept myself just as I am, I change
Your Goal Isn’t Really to Get a Job - by Matt Beard
Benchmark Scores = General Capability + Claudiness
Claude Opus 4.5 System Card
How To Hide A Data Center - by Jason Hausenloy
Text remains the loveliest medium - Tommy Dixon
Buck's Shortform — LessWrong
Should the US Sell Hopper Chips to China? | IFP
Weird Generalization and Inductive Backdoors: New Ways to Corrupt LLMs
Simulacrum Levels — LessWrong
The Sovereignty of Good
[2512.15674] Activation Oracles: Training and Evaluating LLMs as General-Purpose Activation Explainers
RL Environments and RL for Science: Data Foundries and Multi-Agent Architectures
Legible vs. Illegible AI Safety Problems — LessWrong
Omniscaling to MNIST — LessWrong
Frontier Safety Framework Report - Gemini 3 Pro (November, 2025) v2
Your Guide to Webflow’s $4B PLG Engine: How Webflow Pairs Self-Service and Sales for Rapid Growth
Tiny Algae and the Political Theater of Planting One Trillion Trees
Science is getting harder - by Matt Clancy
Michael Nielsen on visualizations, biological systems, and making a new science
Shtetl-Optimized » Blog Archive » My AI Safety Lecture for UT Effective Altruism
innovative_contracting_case_studies_2014_-_august.pdf
12 tentative ideas for US AI policy - Open Philanthropy
COLM 2024
Toward A Mathematical Framework for Computation in Superposition — LessWrong
Distillation Walkthrough
How to compute Hessian-vector products? | ICLR Blogposts 2024
An update on our general capability evaluations - METR
Going Soft, by Lily Scherlis
CaMeL offers a promising new direction for mitigating prompt injection attacks
Gemini Flash Pretraining
Cost-Effective Constitutional Classifiers via Representation Re-use
Ten AI safety projects I'd like people to work on
Cyber Competitions
Discussing Learned Concepts with Language Models · John Hewitt
Proofs & Reasons @ CMU
The Engineering State - American Affairs Journal
Benchmark Scores = General Capability + Claudiness | Epoch AI
Alignment remains a hard, unsolved problem — AI Alignment Forum
Ideas Aren’t Getting Harder to Find—Asterisk
New report: "Scheming AIs: Will AIs fake alignment during training in order to get power?" - Joe Carlsmith
Against Almost Every Theory of Impact of Interpretability — LessWrong
A starting point for making sense of task structure (in machine learning) — LessWrong
DSLT 0. Distilling Singular Learning Theory — LessWrong
Efficient Dictionary Learning with Switch Sparse Autoencoders — LessWrong
Coercion is an adaptation to scarcity; trust is an adaptation to abundance — LessWrong
Training a Reward Hacker Despite Perfect Labels — LessWrong