Turn-Averaged SAEs for Feature Discovery and Long-Context Attribution

arXiv:2606.28548v1 Announce Type: new Abstract: Sparse autoencoders (SAEs) have become a useful tool for extracting interpretable features in language models. However, standard SAE architectures operate on individual token activations, meaning that the number of active features scales linearly with context length, and studying long model transcripts becomes difficult. We introduce turn-averaged SAEs, which represent a single Human or Assistant turn with a fixed number of features by learning to

Generating in the Limit with Infinitely Many Hallucinations

arXiv:2606.28354v1 Announce Type: new Abstract: The classic paradigm of language identification in the limit models learning as a game between an adversary, who reveals strings from an unknown target language, and a learner tasked with identifying that language. The recently introduced framework of language generation in the limit shifted the objective to better reflect modern language modeling, requiring the learner to produce valid, unseen strings from the target language. Related work highlig

Extracting Knowledge from an Arabic-English Machine-Readable Dictionary Using Information Extraction

arXiv:2606.28457v1 Announce Type: new Abstract: Natural language processing (NLP) applications need large and rich amount of linguistic knowledge. Furthermore, electronic language sources such as dictionaries, encyclopedia, and corpora became available. So, automatic methods are emerged to extract lexical information from those sources to overcome the knowledge acquisition bottleneck. We presented a method to automatically extract lexical information from a machine-readable version of the Arabic

Developmental Trajectories of Situation Modeling and Mentalizing in Transformer Language Models

arXiv:2606.28524v1 Announce Type: new Abstract: Recent work suggests that Large Language Models (LLMs) are sensitive to the belief states of agents described by text, as measured by the false belief task (FBT), yet persistent concerns of construct validity remain. We adopt a **developmental perspective**, tracing the pattern of mental state reasoning behavior -- and likely **preconditions** for this behavior -- across multiple training stages in the Olmo2 and Pythia language model suites. We fin

Turn-Averaged SAEs for Feature Discovery and Long-Context Attribution

Join 1,000+ people who read this first.

Related stories

Generating in the Limit with Infinitely Many Hallucinations

Extracting Knowledge from an Arabic-English Machine-Readable Dictionary Using Information Extraction

Developmental Trajectories of Situation Modeling and Mentalizing in Transformer Language Models