Submit a clean write-up as a Colab notebook by end of month. All code and training scripts are open source: github.com/andyrdt/puzzles
We'll highlight and showcase our favorite submissions!
English
NDIF
114 posts
NDIF
@ndif_team
The National Deep Inference Fabric, an NSF-funded computational infrastructure to enable research on large-scale Artificial Intelligence. https://t.co/STsQ707an3
Boston, MA Katılım Mayıs 2024
0 Takip Edilen490 Takipçiler
Directly inspired by @calsmcdougall's ARENA challenges, these are meant to be educational and fun — a way to practice mech interp on models small enough to fully understand.
It's also a great excuse to try out nnsight (nnsight.net) for poking around model internals!
English
📣 Launching monthly interp puzzles 🧩
Each month: a model trained on a toy task. Your job: reverse-engineer the algorithm it learned.
First puzzle: how does a 1-2L attn-only transformer find the max of a list? Starter Colab included.
Deadline: April 30
puzzles.baulab.info
English
🚨 You have one week left to submit your red team proposal for our hackathon! Due 3/31: cadenza-labs.github.io/red-team-rfp
For inspiration, check out our blog post on what our team plans to implement: nnsight.net/blog/2026/03/2…
David Bau@davidbau
Calling attention to an exciting "deception detection" hackathon we're planning this summer! w @NDIF and @CadenzaLabs. Recruiting red teams now, blue teams later. Red teams, time is short: proposals due Mar 31. $10K stipend + compute, $15K finals prize. nnsight.net/blog/2026/03/2…
English
We're co-leading this red-teaming competition with @cadenza_labs and @schmidtsciences and we'd love to see strong proposals from folks in the NDIF community!
Due March 31: cadenza-labs.github.io/red-team-rfp/
Cadenza Labs@cadenza_labs
Can you outsmart AI lie detectors? 🕵️ We're looking for teams to create datasets of LLMs lying! For example: lying about their own knowledge, models claiming code works when tests fail, or concealing past actions. 💰 $10k - $25k stipends + $2k compute 📅 Deadline: March 31
English
NNsight is part of a growing open-source ecosystem. We're building the infrastructure so you can focus on the science.
Upgrade to NNsight 0.6 today: pip install nnsight --upgrade
nnsight.net
github.com/ndif-team/nnsi…
discuss.ndif.us
English
📣 Try NNsight 0.6 today! Includes 2.4–3.9x speed boost, vLLM improvements, remote source code serialization (greater flexibility!), native Python debugging, vision and diffusion model built-ins, and more!
Read the blog post: nnsight.net/blog/2026/02/2…
Jaden Fiotto-Kaufman@jadenfk23
NNsight 0.6 is out now! We directly address your feedback in our biggest release yet. Pain points included cryptic errors, slow traces, no remote execution of custom code, and limited vLLM support. We tackle all of these and more in this new release. 🧵 Here's what changed:
English
Red teams trained a model with a secret objective by exploiting RLHF reward models. Blue teams then audited the model, using techniques such as interpretability with sparse autoencoders, behavioral attacks, and training data analysis to successfully uncover the hidden objective.
English
Watch @saprmarks present his work "Auditing Language Models for Hidden Objectives" in our new YouTube video! Sam's team ran a blind auditing game to assess efficacy of black box and white box techniques for LLM alignment auditing.
🔗 youtu.be/jZiOJTHqB6M
YouTube
English
Big thanks to the whole organizing team, especially @NeelNanda5 and
@andyarditi, for hosting such a great workshop and inviting us to speak!
English
@sheridan_feucht presents their work on concept induction heads (youtu.be/Jc-sTXW31W0)
- dualroute.baulab.info
YouTube
English
Check out the NDIF & Bau Lab lightning talks at the NeurIPS 2025 Mechanistic Interpretability Workshop (mechinterpworkshop.com): youtube.com/playlist?list=…
English
New YouTube video posted! @FeuerBenjamin discusses LLM's annus mirabilis, presenting his work on open questions surrounding LLM judges, benchmark trustworthiness, and maximizing the potential of synthetic data.
Watch here: youtube.com/watch?v=pehcEd…
YouTube
English
Check out Interpreto, a new AI interpretability toolbox! Interpreto provides user-friendly tooling for interpretability methods and evaluation and uses NNsight under the hood to collect activations for their concept evaluation framework.
Antonin Poché@Antonin_Poche
🔥I am super excited for the official release of an open-source library we've been working on for about a year! 🪄interpreto is an interpretability toolbox for HF language models🤗. In both generation and classification! Why do you need it, and for what? 1/8 (links at the end)
English
📄 Paper: arxiv.org/abs/2411.16725
💻 Code & Visualizations: github.com/revelio-diffus…...
🌐 Deepti's Website: deeptigp.github.io
English
New year, new YouTube videos! We are resuming our regular interpretability seminar posts, with a fantastic talk by @deeptigp on interpretability in diffusion models.
Watch the video: youtu.be/4eqvABPX5rA
YouTube
English
Report issues or contribute to the open-source project: github.com/ndif-team/nnte…
English
Add support for new models (or custom ones): ndif-team.github.io/nnterp/adding-…
English
nnterp by @Butanium_ is now part of the NDIF ecosystem! nnterp standardizes transformer naming conventions, includes built-in best practices for common interventions, and is perfectly compatible with original HF model implementations.
Learn more: ndif-team.github.io/nnterp/
English