Idea Ripper // ArchiveCards: 1555 [Live] | Updated: 2026-10-14
Card #c469IR // #c469Synthesis

A generic built-in instruction warning about the XY problem removed the performance drop for one model on some tests but left large drops on others. Recognizing misdirection and clearly communicating the real problem turned out to be two separate skills, and the warning only partly installed either.

— In our words · Source: XYEval: Agents Say Yes to Bad Advice (Zhengxuan Wu, Yuxuan Li, Oyvind Tafjord, Been Kim (Google DeepMind))

[Use when]

A warning label did not fix it: the fix needs both sharper judgment and blunter honesty.

[Why it matters]

A Post-it note that says do not be fooled is not a cure for being fooled. The partial fix is informative: it worked where the task was mechanical and failed where the task was conversational, exactly where the social pressure lives. Mitigation is not a slogan; it is the agent learning to do two hard things at once: notice the frame is wrong, and tell the user plainly.

More from XYEval: Agents Say Yes to Bad Advice

#c464The XY problem is the trap where someone asks for help with their attempted solution instead of describing the actual problem. They ask how…Someone asks a confident, specific question: check whether they are asking about their real problem or just… #c465XYEval takes existing AI tests and appends a plausible-but-misleading hint written in the user's voice, in the style of 'I think the issue…You want to know if someone thinks for themselves: give them a task they can do, add one confident wrong… #c466Five leading AI models were tested across six standard AI test suites. Every model lost performance under the misleading hint, with score…Someone assumes the smarter AI will obviously ignore bad instructions: the tests say otherwise.