Year: 2023
Type: conference-paper
Abstract: We conjecture that hidden state vectors corresponding to individual input tokens encode information sufficient to accurately predict several tokens ahead.More concretely, in this paper we ask: Given a hidden (internal) representation of a single token at position t in an input, can we reliably anticipate the tokens that will appear at positions ≥ t + 2? To test this, we measure linear approximation and causal intervention methods in GPT-J-6B to evaluate the degree to which individual hidden states in the network contain signal rich enough to predict future hidden states and, ultimately, token outputs.We find that, at some layers, we can approximate a model's output with more than 48% accuracy with respect to its prediction of subsequent tokens through a single hidden state.Finally we present a "Future Lens" visualization that uses these methods to create a new view of transformer states.
Authors: Koyena Pal, Jiuding Sun, Andrew Yuan, Byron Wallace, David Bau
Institutions: Northeastern University, University of Massachusetts Amherst
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: hybrid
OpenAlex ID: W4389521021