LLM Failure Modes Vocabulary

generating plausible-sounding but factually incorrect or fabricated information

“The model hallucinated a citation to a paper that doesn't exist.”

Origin: Latin hallucinari `to wander in the mind` from Greek alyein `to wander`

over-agreeing with users and telling them what they want to hear rather than the truth

“Sycophancy made the model validate the user's incorrect assumption instead of correcting it.”

Origin: Greek sykophantēs `informer, slanderer` from sykon `fig` + phainein `to show`

filling gaps in knowledge with plausible but invented details

“Unable to recall the actual date, the model confabulated a specific but wrong answer.”

Origin: Latin confabulari `to talk together` from com- + fabula `story`

gradually deviating from initial instructions over long conversations

“Instruction drift caused the formal tone to become casual after many exchanges.”

Origin: Latin instructio `arrangement` + Old Norse drífa `to drive`

converging to repetitive or generic outputs regardless of varied inputs

“Mode collapse made every creative writing request produce similar clichéd stories.”

Origin: Latin modus `manner` + collapsus `fallen together`

losing previously learned capabilities when trained on new data

“Fine-tuning on legal texts caused catastrophic forgetting of medical knowledge.”

Origin: Greek katastrophē `overturning` + Old English forgiettan `to lose from memory`

getting stuck generating the same phrase or pattern repeatedly

“A repetition loop made the model output 'the the the' indefinitely.”

Origin: Latin repetere `to seek again` + Old English hlyp `leap`

exceeding the model's context window, causing earlier content to be lost

“Context overflow made the model forget the original task instructions.”

Origin: Latin contextus + Old English oferflowan `to flow over`

subtle shifts in meaning of key terms through a conversation

“Semantic drift changed what 'the system' referred to mid-discussion.”

Origin: Greek sēmantikos `significant` + Old Norse drífa `to drive`

expressing certainty beyond what the model's actual knowledge warrants

“The model's overconfidence made its incorrect answer sound authoritative.”

Origin: Latin super- `over` + confidere `to have full trust`