Louis Schiekiera

🚨 New preprint available! 🚨 We test how much of an LLM's internal semantic geometry can be recovered from behavior alone. Across 8 LLMs and 17.5M trials, forced-choice tasks align with hidden-state structure much better than free association. Preprint: arxiv.org/pdf/2602.00628

Authors: Louis Schiekiera, Max Zimmer (@maxzimmerberlin), Christophe Roux (@chrisrx13), Sebastian Pokutta & Fritz Günther. Data: huggingface.co/datasets/schie…

Implication: Forced choice concentrates evidence (shared candidate sets), so its behavior-derived similarity better predicts unseen hidden-state similarities even without logit access. This makes forced choice a practical probe for representation analysis.

🚨 New preprint available! 🚨 We test how much of an LLM's internal semantic geometry can be recovered from behavior alone. Across 8 LLMs and 17.5M trials, forced-choice tasks align with hidden-state structure much better than free association. Preprint: arxiv.org/pdf/2602.00628

🎉Our paper "Neural Discovery in Mathematics: Do Machines Dream of Colored Planes?" was selected for an ORAL (top 1%) presentation at #ICML25! 🎉 arXiv: arxiv.org/abs/2501.18527 blog: maxzimmer.org/blog/2025/neur… 🧵: coming soon

Glad to be part of this initiative. PsychLing-101 will make it easier to connect and use psycholinguistic datasets across projects. If you’re working with language processing data — join us!

@larsweisbrod Vaswani, A., Shazeer, N., Parmar, N., Uszkoreit, J., Jones, L., Gomez, A. N., ... & Polosukhin, I. (2017). Attention is all you need. Advances in neural information processing systems, 30.

what is the best text from analytic philosophy arguing against "ethics as first philosophy"

Excited to present our #ESCOP2025 poster on Friday 12:30 in Sheffield: "Modeling Lexical Competition in Language Production: A Computational Approach to the Swinging Lexical Network". Come chat about spreading activation, picture–word interference & computational modeling!

Thanks again to @annemscheel for the open materials and to my co-author @helen_niemeyer! (4/4)

New exploratory analyses (thanks to reviewer suggestions): Even when analyzing ~2,000 abstracts across all SP quartiles, we still find no evidence that SP explains differences in positive result prevalence. (3/4).

Our study "Does Scientific Productivity Increase the Publication of Positive Results?" is now published in Collabra: Psychology! Link: doi.org/10.1525/collab… (1/4)

OSF, why? Every second request throws an error. Poorly functioning tools replaced by non-functioning ones. OSF still feels like a UX time nightmare—lightyears behind GitHub. Why does the open science community keep settling for this? #OSF #OpenScience

@maxzimmerberlin Very strong blog post!

Now with a blogpost (and the first one on my website!): maxzimmer.org/blog/2025/esti…

@slowberlin 1947 wurde schon mit italienischer Küche geworben, interessant !

Berlin around 1947

Look at the distribution of z-values from medical research!

KV caching in LLMs, clearly explained (with visuals):

(Yet) another tale of Rise and Fall: DeepSeek R1 is claimed to match o1/o1-preview on olympiad level math & coding problems. Can it handle versions of AIW problems that reveal generalization & basic reasoning deficits in SOTA LLMs? (arxiv.org/abs/2406.02061) 🧵1/n

Working memory capacity seems like the most stupidly underpowered part of human intelligence