On four minimal graph-construction tasks, multi-token training (teacherless or diffusion) produces more diverse and original outputs than next-token training, and random seed prefixes can replace temperature as a diversity knob.
Why LLMs Cannot Think and How to Fix It
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
abstract
This paper elucidates that current state-of-the-art Large Language Models (LLMs) are fundamentally incapable of making decisions or developing "thoughts" within the feature space due to their architectural constraints. We establish a definition of "thought" that encompasses traditional understandings of that term and adapt it for application to LLMs. We demonstrate that the architectural design and language modeling training methodology of contemporary LLMs inherently preclude them from engaging in genuine thought processes. Our primary focus is on this theoretical realization rather than practical insights derived from experimental data. Finally, we propose solutions to enable thought processes within the feature space and discuss the broader implications of these architectural modifications.
fields
cs.LG 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Roll the dice & look before you leap: Going beyond the creative limits of next-token prediction
On four minimal graph-construction tasks, multi-token training (teacherless or diffusion) produces more diverse and original outputs than next-token training, and random seed prefixes can replace temperature as a diversity knob.