LLMS CAN GET "BRAIN ROT"!

Summary from the TLDR Dev newsletter

Continuous exposure to low-quality, "junk" web data can cause a lasting cognitive decline in LLMs. Through controlled experiments, researchers found that pre-training LLMs on data from X selected for high engagement or sensationalism negatively impacted reasoning, long-context understanding, and ethical behavior. The primary error was identified as "thought-skipping," where models truncated reasoning steps, and standard fine-tuning methods only partially mitigated the damage, showing a persistent change in the models' representation.