dlvr.it
Time after time: neural codes of speech comprehension
The acoustic signal of speech is fleeting and unfolds linearly; yet listeners must derive temporally extended and hierarchically organized linguistic structures to comprehend it. This article reviews evidence that the human brain transforms the transient auditory signal into neural representations that are persistent, parallel, and time-stamped. First, neural persistence allows information to remain available after it has disappeared from the acoustic input. Second, parallel encoding of recent inputs across multiple levels of language structure enables composition and interactions across the speech hierarchy. Third, dynamic neural representations encode both content and elapsed time, providing a flexible mechanism for preserving sequence order. Together, these representational properties reveal how the human brain transforms a continuously disappearing signal into temporally extended linguistic structures.
![Overview of the paper.
Title: The Emergent Symbolic Structure of Artificial Neural Networks
Authors: Tom McCoy, Paul Soulos, Tal Linzen, Paul Smolensky
Left: Neural networks encode information in vectors (there is then an image of a vector), yet they excel at tasks long thought to require symbolic structure (there is then an image of a symbolic representation, specifically a syntax tree). How do LLMs do it?
Right: We find that LLM representations can be closely approximated with symbolic structures. This approximation lets us edit the structure of an LLM’s output by editing the structure of its internal representations, as shown. There is then an image of two edits to LLMs. In the first one, the original input is 3 + 6 * 8, with an answer of 51. But if we swap the positions of the 3 and the 6, the output becomes 30. In the second one, the original input is a Python command repeating the list [Z, U] three times, producing [Z, U, Z, U, Z, U]. But if we edit the input in a way that adds a Q at the end of the input, the output becomes [Z, U, Q, Z, U, Q, Z, U, Q].](https://cdn.bsky.app/img/feed_thumbnail/plain/did:plc:bcsvlv7ubeagj5virdvatadd/bafkreic3zm2rpc4vtqrz2t3y4vcl3t2owp2sxay42vuh5zm5zkg4jhr64e)
