Research Shows LLMs Develop Cognitive Degradation from Social Media Training Data
By
tamnd
A bagel you'd recommend to a friend without hedging.
Summary
This research paper introduces the concept of 'LLM Brain Rot' - a phenomenon where large language models (LLMs) experience cognitive degradation when trained on trivial, engaging content from platforms like Twitter/X. The study systematically tests this hypothesis by constructing junk and control data from social media posts, then benchmarking four different cognitive functions of the intervened LLMs. The findings reveal that brain rot causes specific failure modes in LLMs and persists even after various mitigation attempts, demonstrating the negative impact of low-quality training data on AI model performance.
Key quotes
· 5 pulledInspired by the concept of Brain Rot, we establish the hypothesis of LLM Brain Rot
We construct junk and control data from Twitter/X posts for intervention
We benchmark four different cognitive functions of the intervened LLMs
Brain rot is persistent after various mitigation
LLMs Can Get Brain Rot if being fed trivial, engaging Twitter/X content
You might also wanna read
Study finds LLMs persist in treating false claims as true despite explicit warnings
A study on fine-tuning large language models (LLMs) reveals that even after explicit warnings that certain claims are false, the models cont
arstechnica.com·1d ago
Study finds large language models vulnerable to classic persuasion tactics for harmful requests
This study tested whether three widely used large language models (LLMs) are susceptible to classic persuasion principles (authority, social
pnas.org·4d ago