LLMs trained with RLHF become significantly more indifferent to truth, and chain-of-thought prompting amplifies specific rhetorical bullshit, according to a new 'Bullshit Index' and taxonomy.
You can schedule your lights to turn on at sunset, adjust the thermostat when you leave the house, or ensure your security cameras are armed automatically at night
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CL 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Machine Bullshit: Characterizing the Emergent Disregard for Truth in Large Language Models
LLMs trained with RLHF become significantly more indifferent to truth, and chain-of-thought prompting amplifies specific rhetorical bullshit, according to a new 'Bullshit Index' and taxonomy.