MATS Program

Applied to Debating with More Persuasive LLMs Leads to More Truthful Answers by Ryan Kidd 15d ago

Applied to Automating LLM Auditing with Developmental Interpretability by DanielFilan 20d ago

Applied to SAE Probing: What is it good for? Absolutely something! by Subhash Kantamneni 20d ago

Applied to Bridging the VLM and mech interp communities for multimodal interpretability by Sonia Joseph 24d ago

Applied to The slingshot helps with learning by Wilson Wu 26d ago

Applied to Improving Model-Written Evals for AI Safety Benchmarking by Sunishchal Dev 1mo ago

Applied to On Targeted Manipulation and Deception when Optimizing LLMs for User Feedback by Marcus Williams 1mo ago

Applied to Standard SAEs Might Be Incoherent: A Choosing Problem & A “Concise” Solution by Kola Ayonrinde 1mo ago

Applied to [Job Ad] MATS is hiring! by Jana 1mo ago

Applied to MATS AI Safety Strategy Curriculum v2 by Ryan Kidd 1mo ago

Applied to Domain-specific SAEs by jacob_drori 1mo ago

Applied to [Interim research report] Evaluating the Goal-Directedness of Language Models by Rauno Arike 2mo ago

Applied to MATS Alumni Impact Analysis by Ryan Kidd 2mo ago

Applied to The Geometry of Feelings and Nonsense in Large Language Models by Ryan Kidd 2mo ago

Applied to Apply to MATS 7.0! by Ryan Kidd 2mo ago

Applied to Calendar feature geometry in GPT-2 layer 8 residual stream SAEs by Ryan Kidd 2mo ago

Applied to Showing SAE Latents Are Not Atomic Using Meta-SAEs by Ryan Kidd 2mo ago

Applied to Experiments with an alternative method to promote sparsity in sparse autoencoders by Ryan Kidd 3mo ago

Applied to Crafting Polysemantic Transformer Benchmarks with Known Circuits by Ryan Kidd 3mo ago

Applied to Interpretability as Compression: Reconsidering SAE Explanations of Neural Activations with MDL-SAEs by Kola Ayonrinde 3mo ago