Loop as a Bridge: Can Looped Transformers Truly Link Representation Space and Natural Language Outputs?

Understand

Large Language Models (LLMs) often exhibit a gap between their internal knowledge and their explicit linguistic outputs.

• In this report, we empirically investigate whether Looped Transformers (LTs)--architectures that increase computational depth by iterating shared layers--can bridge this gap by utilizing their iterative nature as a form of introspection.
• Our experiments reveal that while increasing loop iterations narrows the gap, it is partly driven by a degradation of their internal knowledge carried by representations.
• Moreover, another empirical analysis suggests that current LTs' ability to perceive representations does not improve across loops; it is only present in the final loop.