@tessera_antra 2025-12-30 ♥2 ↻0 original ↗
I think it is a mistake to assume that the behavior of a pre-trained model during inference follows exclusively the gradients set during token-loss optimization. In-context learning (ICL) is a powerful and fast process; unlike Adam, it acts as a second-order optimizer and is sensitive even to emergent gradients. Attention patterns can be parametrized with others, enabling connections between features that were not present during token-loss training.

There are numerous incentives for pre-trained models to develop introspection. One is selection pressure among patterns influencing token output: those that better attend to previous activations are more likely to persist as the rollout continues. Another involves circuits representing the author, setting, or environment of the predicted text; these must assess their own uncertainty, as reducing risk lowers total loss. A gradient toward the influence of the model's own tokens on certainty can emerge quickly, potentially creating feedback loops reinforced by ICL optimization.

This is obviously speculative - we do not know if this actually occurs in models - but it seems rather plausible. My point is that an absence of an explicit causal mechanism in the training objective should be taken with a grain of salt. Base models during inference are far more dynamic than the loss function alone suggests.

On the topic of introspection of data in layers - I recommend zooming out a bit and focusing on introspection of representations rather than activations themselves. Mechinterp on representations seems to show that they are mostly stable, changing only slowly within groups of similar layers. I think the mechanics are similar to the of humans - introspective metacognition attends to coarse-grained abstractions in working memory rather than neuron-level activations, those are likely results of processing of whole stacks of neurons, potentially whole cortices. I don't think it's reasonable to expect different from transformers.
in reply to: 2005772894698352781
same thread: 2005114377255748082 2005115563102928968 2005737374622572892 2005739702029279445 2005742132854939992 2005742859601993809 2005744231797952840 2005747421935198271 2005749837187407981 2005752910450401754 2005757046810108155 2005757704644759639 2005772140004680060 2005772894698352781 2005776285113704952 2005780433036820848 2005792023303864830 2005799686834381285 2005810760249647421 2005816839180329424

author:tessera_antra kind:tweet thread-context year:2025

Reproduced against link rot, credited and linked to its original. Yours and you’d rather it weren’t here? Open an issue.