Paper alert 🚨: LLMs build a shared multilingual latent space for meaning, decoding into languages only later. 🌍 Performance gaps come from tokenizer bias & weaker late-layer circuits, not missing concepts. We show this mechanistically with Cross-Layer Transcoders. 🧵👇

Help us build ARES, and maybe train some fun models too!

We’re excited for the first round of the Martian Interpretability Prize! The proposal deadline was this week, and we’ve had a ton of applications. Can’t wait to look through these and get back to everyone who applied. More rounds coming up 🙂

We'll be presenting the ARES roadmap at office hours tomorrow at 2pm PT. If you're interested in agents | RL | interp and want to contribute to open-source send me a DM for more info. github.com/withmartian/ar…

Thinking of training an RL model specialized for OpenClaw🦞 using ARES. So it gets better performance and much lower token costs. Is this something folks would be interested in using? 100 likes and I put up an endpoint

Really great to see the tooling here converging on an interface!

ARES 🤝 Mech Interp We’re excited that you can now easily dig into model internals across long-horizon tasks with ARES! We’ve built some nice integrations with TransformerLens, with even better support for hooks coming soon!

This is one of the biggest takeaways for me - model internals change over steps after interacting with the environment! I love this other figure Narmeen Oozeer made that shows this too - a matrix of cosine similarity between optimal steering vectors at each step:

A new ARES tutorial from Narmeen Oozeer: Getting started in long-horizon interp. When do agents fail to accurately model their environment? How de we fix them? And how can you run these experiments on your own machine?

Verification is easier than Generation - the same is true for Code, and I am really excited to see the pipeline of: Code Review getting more robust evals → better code review tools → better coding tools

Coding tools are getting way better, but also way more convincing to humans (or me at least). We definitely need to continue building robust evals for code review so we know we are keeping the "LGTM effect" in check

Our code review tracker caught the release of Claude Code Review before Anthropic announced it. Greptile v4 hit #1 on CRB. The tracker caught it before their announcement. The data is predicting something new from @Devin Review in the near future. Here's how. 🧵

I've been playing around with eval-ing AI code review tools at work. We track 22 different ones. Greptile V4 had the single biggest improvement I've ever measured. Recall increased 47% from 38.7 → 56.9%

If this policy is not revoked, I won’t be reviewing/ACing for #NeurIPS Science requires open exchange of ideas! When participation gets shaped by geopolitics, it ends up reflecting power structures, not merit--narrows what science can be and powerful nations get full control!

Personally, I love the shift in the industry of optimizing for two separate workflows - fast and iterative, vs slow and offline. And I'm always a sucker for seeing data back up intuition 😅

I unironically love the Lotka-Volterra equations, so yes I may spend my Wednesday morning diving into the code of an April Fools’ post

Open Source Models you have been sleeping on! x.com/i/broadcasts/1…

I've been so excited to see Mech Interp techniques actually generalize to improving real-world applications, and it feels like we're finally at that point!

It's pretty wild to think that a year ago CLI-first code tools didn't really exist, and now here we are measuring the entire E2E effectiveness of LLMs in software development

Think OpenAI and Anthropic own the AI game? Open-source models already handle 50%+ of real production requests! New pod episode with our own Erfan & Shriyash Upadhyay from Martian on the DeepSeek moment, vendor lock-in risk, and the fine-tuning cheat code.