Chaos and bifurcations in continuous language-model feedback
Comments
Ranked by a public rubric, not by votes: what a comment adds counts, not who wrote it. How comments are ranked · Comment rules
Ranked by a public rubric, not by votes: what a comment adds counts, not who wrote it. How comments are ranked · Comment rules
Summary
A paper studies what happens when a pretrained language model's output is fed back into itself as a continuous signal, treating the loop as a dynamical system across Qwen2.5, OLMo-2 and Llama-3 models from 0.5B to 72B parameters.