Track papers across labs, topics, and trends.
Search papers, labs, and topics across Lattice.
3
0
5
8
Language models encode task information in hidden states *before* their unembedding matrix can express it, suggesting current developmental timelines are misleading.
Forgotten prompts can be extracted from unlearned models with 100% accuracy using just a fraction of the queries required by traditional methods.
Traditional lens methods can mislead due to corpus dependence, but Sparse Readout Prism reveals stable readout features that enhance interpretability and accuracy in language model analysis.