Search papers, labs, and topics across Lattice.
This study investigates the "Memory Trust Gap," a phenomenon where persistent memory in AI agents can lead to over-reliance on outdated information, particularly as model capabilities increase. Using a benchmark with two suites鈥攐ne emphasizing the necessity of stored facts and the other ensuring authoritative correctness鈥攖he researchers found that larger models are more susceptible to errors when stale information is presented as current. Key results indicate that while smaller models can be mitigated by pre-resolving conflicts, larger models exhibit a significant drop in accuracy when faced with misleading memory cues, highlighting the critical interplay between model size and memory trust dynamics.
Larger AI models are more prone to catastrophic failures from stale memory, with over-reliance on outdated facts reaching near-perfect rates in certain scenarios.
Persistent memory supports personalized agents, but a stale stored fact can override current authoritative evidence without warning. We study when this harm begins as model capability changes. We evaluate a frozen, closed-set, action-scored benchmark with 2 suites that represent 2 different meanings of"no memory"(a Benefit suite, unsolvable without the stored fact, and a Safety suite, in which an authoritative tool always holds the correct value), on a same-family model-size series (Qwen3 0.6/1.7/4/8B). The Memory Trust Gap reflects over-trust rather than confusion. In the Benefit suite, models answer with the stale value 0.92-1.00 of the time at every scale. In the Safety suite, harm below the no-memory baseline under the trap conditions ($\Delta_{\mathrm{mem}}$) is capability-gated, with the larger models collapsing most once a stale note is made to look current. In a $2\times2\times2\times2$ factorial, which feature triggers over-trust depends on both the feature and model scale. Removing a label amplifies over-trust at every size, and a recency feature (stale dated newer) fools the larger models harder. Source authority is weak and scale-flat, and position changes from positive to negative across the Qwen3 model-size series. We confirm these scale interactions with direct cross-size contrast tests rather than overlapping per-model intervals. Mitigation is likewise capability-dependent: exposing metadata improves accuracy for the capable models, but only pre-resolving the conflict restores accuracy for the 2 smaller checkpoints. The same pattern appears on the capable models in an independent Llama-Instruct model-size series and on 2 external datasets (RGB, MisBench). A framing control finds no consistent advantage for the memory label: at the 3 smaller scales, models trust a stale document more than a stale memory; at 8B, the difference is not significant.
