Definition
Cognitive bias attacks exploit known biases in either humans or LLM judges to get a result the substance wouldn’t earn — flattering framing, authoritative tone, anchoring, social proof. The LLM-judge version is particularly worrying because the “cognitive biases” of judges become attack surfaces in any evaluation that depends on them.
Episodes covering this
Worth reading next
Papers we haven't done a deep dive on yet, but would recommend on this topic.