AlgZoo: uninterpreted models with fewer than 1,500 parameters — LessWrong
This post covers work done by several researchers at, visitors to and collaborators of ARC, including Zihao Chen, George Robinson, David Matolcsi, Ja…
x AlgZoo: uninterpreted models with fewer than 1,500 parameters — LessWrong Interpretability (ML & AI) Alignment Research Center (ARC) AI Frontpage 2026 Top Fifty: 14 % 181 AlgZoo: uninterpreted models with fewer than 1,500 parameters by Jacob_Hilton 26th Jan 2026 ARC AI Alignment Forum 12 min read 7 181 Ω 61 This post covers work done by several researchers at, visitors to and collaborators of ARC, including Zihao Chen, George Robinson, David Matolcsi, Jacob Stavrianos, Jiawei Li and Michael Sklar. Thanks to Aryan Bhatt, Gabriel Wu, Jiawei Li, Lee Sharkey, Victor Lecomte and Zihao Chen for co
saved by
related reading
- ARC progress update: Competing with sampling — LessWronglesswrong.com
- Weight-Sparse Circuits May Be Interpretable Yet Unfaithful — LessWronglesswrong.com
- Formal verification, heuristic explanations and surprise accounting — Alignment Research Centeralignment.org
- Language Model Circuits Are Sparse in the Neuron Basis | Transluce AItransluce.org
- Attribution-based parameter decomposition — LessWronglesswrong.com
- arxiv.org/pdf/1805.08522arxiv.org
- An Extremely Opinionated Annotated List of My Favourite Mechanistic Interpretability Papers v2 — AI Alignment Forumalignmentforum.org
- An Extremely Opinionated Annotated List of My Favourite Mechanistic Interpretability Papers v2 — AI Alignment Forumalignmentforum.org
- Tensor-Transformer Variants are Surprisingly Performant — LessWronglesswrong.com
- Fact Finding: Attempting to Reverse-Engineer Factual Recall on the Neuron Level (Post 1) — AI Alignment Forumalignmentforum.org
- The Annotated Transformernlp.seas.harvard.edu
- A gentle introduction to mechanistic anomaly detection — LessWronglesswrong.com