Loading activity…
@feijianghan
CS PhD @UMDCS, working on LLMs and actionable interpretability. Prev @UpennNLP @Google RedNote: https://xhslink.com/m/8ifRndgaJqW Opinions are my own
AI
Its creators say ICA was competitive with public sparse autoencoders across multiple evaluations, using 7–43× less parameter memory.
AI
A user says four other related-work formats scored around weak accept or weak reject under an ICLR 2027 rubric.
