AI Models Get Brain Rot, Too

[scipan]

3. This is wild. Wish I knew more about how AI works.

We analyze the reasoning failures in ARC-Challenge to identify different failure modes. We find that the majority failures can be attributed to "thought skipping" (e.g., the model fails to generate intermediate reasoning steps), which significantly increases in models affected by brain rot.

It seems like it's more than just GIGO,... It's something I would have thought was intrinsic to the LLM. Like it's mimicing the low quality tweets' thought processes???

Thanks for this.