collaborators

15 papers

cs.AI2026

Matryoshka Language Model Suites

Nathan Godey, Yoav Artzi

Training a language model suite classically requires training each model separately and serving them independently. We improve both training and inference efficiency by stacking su…

cs.CL2026

Lost in Backpropagation: The LM Head is a Gradient Bottleneck

Nathan Godey, Yoav Artzi

The last layer of neural language models (LMs) projects output features of dimension to logits in dimension , the size of the vocabulary, where usually . This misma…

cs.CL2026

Co-LMLM: Continuous-Query Limited Memory Language Models

Yair Feldman, Linxi Zhao, Nathan Godey +5

Limited memory language models (LMLMs) externalize factual knowledge during pretraining to a knowledge base (KB), rather than memorizing it in their weights. During generation, the…

cs.CL2026

The State-Prediction Separation Hypothesis

Giovanni Monea, Nathan Godey, Kianté Brantley +1

Transformers use the same forward computation stream to both predict the next token and store useful state for future token predictions. We formulate the \emph{state-prediction sep…

cs.CL2026

Post-training for Efficient Communication via Convention Formation

Yilun Hua, Evan Wang, Yoav Artzi

Humans communicate with increasing efficiency in multi-turn interactions, by adapting their language and forming ad-hoc conventions. In contrast, prior work shows that LLMs do not…

cs.CL2026

No Mean Feat: Simple, Strong Baselines for Context Compression

Yair Feldman, Yoav Artzi

Context compression reduces Transformer inference costs by replacing lengthy inputs with shorter pre-computed representations. It carries significant benefits for retrieval-augment…