Reposted by Jay Hennig
What kind of dynamics underlie sequence working memory? In the last project of my PhD we developed and analysed RNNs fitted to multi-session single-unit data of macaques. Now out on: www.biorxiv.org/content/10.6...
With @jakhmack.bsky.social and data from Chen et al., Neuron 2024.
[1/4]
![To begin, distributions are never normal. For some this seems obvious, hardly worth mentioning, but an aphorism given by Cram´er (1946) and attributed to the mathematician Poincar´e remains relevant: “Everyone believes in the [normal] law of errors, the experimenters because they think it is a mathematical theorem, the mathematicians because they think it is an experimental fact.”](https://cdn.bsky.app/img/feed_thumbnail/plain/did:plc:kwjddjickma3syepdsfufyvi/bafkreifw25h2nnb3ylgdumcsctlyh6wdbxam7ajlnbwnca7hdxef5q6bo4)
![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)
