Beyond Causal Masks: Auditing and Localizing Hidden Future-Token Leakage in Adaptive Sequence Models
Paper recorded by Signals 4 on 2026-10-22 in cs.AI. Abstract reproduced from arXiv; link to the original below.
Published 2026-10-22 on arXiv · recorded by Signals 4 on 2026-10-23
Category: cs.AI · 人工智能 · first seen 2026-10-23
Abstract
Hidden future-token leakage can create false architectural gains: an autoregressive model may appear substantially better under teacher-forced evaluation while its free-running generation deteriorates. We investigate this failure in an adaptive residual-routing decoder whose sequence-level routing summary exposes prefix positions to future tokens despite a correct causal attention mask. An anomalo
Read on arXiv →
Cite this page: Beyond Causal Masks: Auditing and Localizing Hidden Future-Token Leakage in Adaptive Sequence Models: the #197 most recent of 960 cs.AI papers we have recorded (as of 2026-10-22). Source: Signals 4 (Signals API) — https://data.jiangzhang.ca/signals4/t/papers/beyond-causal-masks-auditing-and-localizing-hidden-future-token-leakage-in-adapt.html
Free to quote with attribution to “Signals 4 (Signals API)”. Machine-readable:
papers.json
Get 4 AI signals a day by email — free.
Get 4 AI signals a day by email — free