While the ability of language models to elicit facts has been widely investigated, how they
handle temporally changing facts remains underexplored. We discover Temporal Heads …

Sparse autoencoders (SAEs) are an increasingly popular technique for interpreting the
internal representations of transformers. In this paper, we apply SAEs to'interpret'random …

Neurosymbolic AI (NeSy) offers a promising approach to addressing the interpretability
challenges in artificial intelligence by bridging neural networks and symbolic reasoning. This …