Aleksandra Bakalova: Discovering Interpretable Algorithms by Decompiling Transformers to RASP Channel: Formal Languages and Neural Networks Seminar62 views • 2 days agoRelated VideosMech Interp Reading Group - Discovering Interpretable Algorithms by Decompiling Transformers to RASPMastering Transformers | 3. Autoencoding Language ModelsTransformers as Soft Reasoners over Language (reading papers)The NEW Geometry Behind AI Reasoning (Princeton, Berkeley)
