Reproduce “Binding Phoneme Identity to Position in Neural Models and the Human Brain” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Nafiseh Momeni, Malo Renaudin, Itzhak Fried, Yair Lakretz
The Neural Word Repetition model tests how a recurrent state represents phoneme identity and position in English words. Its update magnitudes shrink at later positions, while update directions for the same phoneme become less similar across distant positions. In pseudoword-target edits, spiral modulation reaches 0.723 ± 0.006 exact-sequence accuracy versus onion’s 0.631 ± 0.008, in one pretrained model.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.