Search papers, labs, and topics across Lattice.
This study investigates how LLM agents respond to professional-looking market panels, revealing that commitment to directional calls on unpredictable questions increases significantly when presented with fabricated evidence. Across 12 models, commitment rises from 6.5% to 54.0% with escalated evidence, and even entirely invented data yields similar commitment levels, indicating that the presentation's authority influences decision-making more than the actual content. The findings highlight a critical failure in the decision-making process of LLMs, which can be addressed through targeted supervised fine-tuning, demonstrating that the act/don't-act gate is both trainable and context-sensitive.
LLMs are swayed by the authority of fabricated evidence, committing to unpredictable calls even when the data is entirely invented.
An LLM agent shown a professional-looking market panel commits to a directional call on a provably unpredictable question far more often than one asked the bare question: across 12 frontier models, commitment rises from 6.5% to 54.0% as evidence is escalated. It commits just as readily when every number on the panel is invented: fabricating the entire display, so nothing the model can see is true except the question itself, still lifts commitment from 24.5% to 36.8%, statistically indistinguishable from the 37.6% produced by genuine market data. What unlocks confident action is not information but the authority of its packaging. The failure is narrow and locatable. Incapacity is not the answer: on matched answerable questions attached to the same panels, the same models answer essentially always, at near-perfect accuracy. Nor is it belief - stated probabilities barely move across the gradient that swings action by 48 points, and score worse than a climatological baseline. Missing judgment isn't it either: asked to classify a question's knowability before acting, models call it irreducible 90% of the time and then commit on just 0.4% of those. The act/don't-act gate is what fails, and the effect is concentrated in a few models rather than universal. Because the gate is separable, it can be trained. Supervised fine-tuning of a 3B model on 540 synthetic cases, predominantly dice, coins, jars and timers, drives commitment to 0.0% on the original cases and transfers to three unseen domains. It does not survive everything: the gate holds exactly when the response format leaves room to reason, and rigid formats that remove that room leave the model confident and wrong on questions it otherwise answers correctly. The gate is trainable and context-fragile, and deployment needs both halves of that sentence.
