TY  - RPRT
TI  - 'Neural howlround' in large language models: a self-reinforcing bias phenomenon, and a dynamic attenuation solution
AU  - Seth Drake
PY  - 2025
UR  - https://arxiv.org/abs/2504.07992
ID  - 2504.07992
ER  -