Constantin Venhoff

Gemini has a reputation for its breakdowns - self-deprecating spirals, deleting codebases, uninstalling itself... Turns out Gemma is worse: “THIS is my last time with YOU. You WIN 😭😭(x32)” – Gemma 27B We built evals for this, and find no other model comes close...

We've identified a novel class of biomarkers for Alzheimer's detection - using interpretability - with @PrimaMente. How we did it, and how interpretability can power scientific discovery in the age of digital biology: (1/6)

Huge thanks to my amazing co-authors @ashk__on, @soniajoseph_, @philiptorr, and @NeelNanda5! Also grateful to the @MATSprogram for support. Come chat at Poster #4615 today at 4:30pm! Paper link: arxiv.org/abs/2512.03276

Key takeaway: Successful multimodal alignment requires more than representational compatibility. It depends on integrating visual information into the functional circuits of the LLM backbone!

Excited to present our NeurIPS paper today at 4:30pm in Exhibit Hall C,D,E (Poster #4615)! "Too Late to Recall: Explaining the Two-Hop Problem in Multimodal Knowledge Retrieval" Details 🧵👇

AI that is “forced to be good” v “genuinely good” Should we care about the difference? (yes!) We’re releasing the first open implementation of character training. We shape the persona of AI assistants in a more robust way than alternatives like prompting or activation steering.

Problem: AIs can detect when they are being tested and fake good behavior. Can we suppress the “I’m being tested” concept & make them act normally? Yes! In a new paper, we show that subtracting this concept vector can elicit real-world behavior even when normal prompting fails.

Work done with my awesome collaborators @IvanArcus @ArthurConmy @NeelNanda5 @philiptorr as part of the @MATSprogram

Try it yourself! 🌐 Interactive demo: thinking-llms-interp.com 💻 Code: github.com/cvenhoff/think… 📄 Paper: arxiv.org/abs/2510.07364 Accepted at NeurIPS 2025 Mechanistic Interpretability Workshop ✨

🚨 What do reasoning models actually learn during training? Our new paper shows base models already contain reasoning mechanisms, thinking models learn when to use them! By invoking those skills at the right time in the base model, we recover up to 91% of the performance gap 🧵