Twitxy Saved

Ruben C. Arslan

@rubenarslan

Bayescurious evidence enthusiast @the100ci & @error_reviews. Topics: evolution, ovulation, mutation, personality, sexuality, R, open science & source tools.
Berlin, Germany rubenarslan.github.io Joined April 2013
1.9K Following 4.6K Followers 11K Posts

Latest posts 5

Pinned post

Ruben C. Arslan @rubenarslan
New work by @BjoernHommel and me. We fine-tuned a language model to predict correlations between survey items.
In our pilot, the out-of-sample accuracy for item correlations was .71, .86 for reliabilities, and .89 for scale correlations. We're planning a preregistered follow-up.
4 replies 29 reposts 100 likes 31K views View on X
Ruben C. Arslan @rubenarslan
An interesting thing is happening to me at the moment. I have a standing bug bounty policy for my first-authored scientific work: I've committed to pay people who report bugs in my code, up to 750€ for severe errors. Since GPT Astra, I'm now getting several bug reports daily.
6 replies 20 reposts 468 likes 134K views View on X
Ruben C. Arslan @rubenarslan
A bit sad to see bad correlational analysis deployed on a question, they could so easily study experimentally. You can't tell the differences are _caused_ by the language/model and not the content/context based on coarse post-treatment controls.
Anthropic Verified account @AnthropicAI Jul 13
In previous research, we found that Claude expresses over 3,000 values, like honesty and warmth. In new work, we asked how the values Claude expresses vary between Claude models and across languages.

We analyzed 300K+ anonymized conversations to find out.anthropic.com/research/claud…
View quoted post on X
2 replies 2 reposts 32 likes 5.2K views View on X