🔥The 9th Round of Easy Loan, Earn $40 Reward is in progress❗️ ⏰ Promotion Period: January 15th - Feburary 15th, 2025 👉 Register now and check more details at gate.io/campaigns/358

Announcement: we're open sourcing Neuronpedia! 🚀 This includes all our mech interp tools: the interpretability API, steering, UI, inference, autointerp, search, plus 4 TB of data - cited by 35+ research papers and used by 50+ write-ups. What you can do with OSS Neuronpedia: 🧵

i forgot to tweet about this, but the very cool people at neuronpedia graciously hosted the steering vectors we trained on AxBench for Gemma-2-2B and 9B, w/ max activating examples and interactive steering neuronpedia.org/axbench

📢 New paper: Applied interpretability 🤝 MT personalization! We steer LLM generations to mimic human translator styles on literary novels in 7 languages. 📚 SAE steering can beat few-shot prompting, leading to better personalization while maintaining quality. 🧵1/

Our work found that semantic interference in LLMs is actually not that random. Certain polysemantic structure persist across models. This hints at something deeper: a shared representational structure that might reflect higher-order of patterns. Our paper: arxiv.org/abs/2505.11611

Researchers can use the Neuronpedia interactive interface here: neuronpedia.org/gemma-2-2b/gra… And we’ve provided an annotated walkthrough: github.com/safety-researc… This project was led by participants in our Anthropic Fellows program, in collaboration with Decode Research.

Mateusz and I are excited to announce circuit-tracer, a library that makes circuit-finding simple! Just type in a sentence, and get out a circuit showing (some of) the features your model uses to predict the next token. Try it on neuronpedia: shorturl.at/SUX2A

Fantastic to see Anthropic, in collaboration with neuronpedia, creating open source tools for studying circuits with transcoders. There's a lot of interesting work to be done I'm also very glad someone finally found a use for our Gemma Scope transcoders! Credit to Arthur Conmy

I think this is the podcast that finally interp-pilled me we snuck in a little intro featuring johnny's neuronpedia and asked about HOW IN THE HECK @anthropicai does all these insanely cracked interp visualizations for their "papers"

New Paper! Robustly Improving LLM Fairness in Realistic Settings via Interpretability We show that adding realistic details to existing bias evals triggers race and gender bias in LLMs. Prompt tuning doesn’t fix it, but interpretability-based interventions can. 🧵1/7