Emergence of Linear Truth Encodings in Language Models
This provides a mechanistic demonstration for how linear truth representations can arise in language models, addressing a foundational problem in AI interpretability.
The study tackled the unclear mechanism behind linear truth encodings in language models by introducing a one-layer transformer toy model that reproduces these subspaces, showing they emerge from data distributions where factual statements co-occur, and observing a two-phase learning dynamic with memorization followed by linear separation.
Recent probing studies reveal that large language models exhibit linear subspaces that separate true from false statements, yet the mechanism behind their emergence is unclear. We introduce a transparent, one-layer transformer toy model that reproduces such truth subspaces end-to-end and exposes one concrete route by which they can arise. We study one simple setting in which truth encoding can emerge: a data distribution where factual statements co-occur with other factual statements (and vice-versa), encouraging the model to learn this distinction in order to lower the LM loss on future tokens. We corroborate this pattern with experiments in pretrained language models. Finally, in the toy setting we observe a two-phase learning dynamic: networks first memorize individual factual associations in a few steps, then -- over a longer horizon -- learn to linearly separate true from false, which in turn lowers language-modeling loss. Together, these results provide both a mechanistic demonstration and an empirical motivation for how and why linear truth representations can emerge in language models.