René's URL Explorer Experiment


Title: Daniel Tan — AI safety researcher

Description: Daniel Tan is an AI safety researcher studying how AI minds tick — leading the model motivations team at Arcadia Alignment.

direct link

Domain: dtch1997.github.io

NoneIE=edge

Links:

01abouthttps://dtch1997.github.io#about
02selected workhttps://dtch1997.github.io#work
03writinghttps://dtch1997.github.io#writing
04nowhttps://dtch1997.github.io#now
05say hihttps://dtch1997.github.io#contact
Scholarhttps://scholar.google.com/citations?user=QKO1QacAAAAJ&hl=en
Twitterhttps://www.twitter.com/DanielCHTan97
LessWronghttps://www.lesswrong.com/users/daniel-tan
LinkedInhttps://www.linkedin.com/in/daniel-tan-a0672b163/
Anon feedbackhttps://admonymous.co/daniel-tan
full list lives on Scholarhttps://scholar.google.com/citations?user=QKO1QacAAAAJ&hl=en
Models trained to write insecure code learn to admire Nazis. Emergent Misalignment: Narrow Finetuning can lead to Broad Misalignment https://arxiv.org/abs/2502.17424
Steer how a model generalizes by adding one line to the training data. Inoculation Prompting: Eliciting traits during training can suppress them at test-time https://arxiv.org/abs/2510.04340
Steering vectors don't work universally — they often fail on their own task. Analyzing the Generalization and Reliability of Steering Vectors NeurIPS 2024 https://arxiv.org/abs/2407.12404
Can robots learn real-world tasks just by watching internet video? Towards Generalist Robot Learning from Internet Video: A Survey In proceedings, JAIR https://arxiv.org/abs/2404.19664
$ ./alignment — rather play than read? Boot a little terminal and play as a model in training. Two of the papers above — emergent misalignment & inoculation prompting — happen to you. play → https://dtch1997.github.io/terminal.html
The one-week sprint→https://dtch1997.github.io/writing/the-one-week-sprint.html
Your Model Organisms Might Be Fried→https://dtch1997.github.io/writing/your-model-organisms-might-be-fried.html
Shaping the exploration of the motivation-space matters for AI safety→https://dtch1997.github.io/writing/shaping-the-exploration-of-the-motivation-space-matters-for.html
Concrete research ideas on AI personas→https://dtch1997.github.io/writing/concrete-research-ideas-on-ai-personas.html
Browse all writing →https://dtch1997.github.io/writing/index.html
books I lovedhttps://dtch1997.github.io/fun/bookshelf.html
what therapy taught mehttps://dtch1997.github.io/fun/therapy.html
stanzas on growthhttps://dtch1997.github.io/fun/personal-growth.html
dating profile outtakeshttps://dtch1997.github.io/fun/dating-outtakes.html
career moves, fitness, dancinghttps://dtch1997.github.io/now.html
my profilehttps://dtch1997.github.io/hi/index.html
anonymous feedback →https://admonymous.co/daniel-tan

Viewport: width=device-width, initial-scale=1.0

URLs of crawlers that visited me.