The authors formalise proactive memory recall as a sequence of multiple comparisons: at each turn the assistant compares the current cue against every stored memory and interrupts if the best match exceeds a threshold.
Memory recall accuracy degrades sharply with store size
Chen and Wang prove that false alarms in proactive memory asymptotically follow a power-law decay, and propose a conformal gate that controls them without labels.
Top University
Yichuan Chen · Xin Wang
Zhejiang Sci-Tech University
Research Digest··3 min read
Chen and Wang show that a long-lived assistant which proactively retrieves memories faces a multiple-testing problem: a fixed similarity threshold tuned for small stores produces 87--100% false alarms at size 2048 across benchmarks and embedding models.
The Daily Edition
The front page in your inbox each weekday morning: the lead, the picks, and why each one earned its place.