Back to AI News
arxiv

RESYNC: Residual Latent-State Synchronization for Deadline-Bound Streaming Inference

RESYNC addresses high-rate streaming inference where a large model may miss per-observation deadlines. The large model runs asynchronously while a fast path updates its stale latent state using newer observations and reuses the unchanged prediction head, learning from frozen-model representations without task labels. Across one real and two synthetic streams and five architecture families, RESYNC outperformed compact or distilled alternatives in all twelve dataset-architecture pairs at anchor ages of 4–12 observations, and in eleven pairs at older ages up to 32. In 1 ms arrival replay, it served at least 99.4% of arrivals within the interval with 0.35–0.62 ms median latency on one CPU thread.

AI Brief

It presents a concrete method for meeting strict streaming latency deadlines while retaining a large teacher model’s predictive advantage, with empirical results across multiple architectures and live replay conditions.

arXiv:2611.16523v1 Announce Type: cross Abstract: Predicting every observation in a high-rate stream requires a model to finish before the next arrival. A large model may offer better predictions but miss that deadline. We propose RESYNC: the large model runs asynchronously, while a fast path updates its latest exact but stale latent state using the observations received since that state was computed and applies its unchanged prediction head. The updater learns from the frozen model's representations without task labels. Across one real and two synthetic streams and five architecture families, RESYNC beats the best compact or distilled alternative with a paired 95% interval above zero in all twelve dataset--architecture pairs with a material teacher advantage at anchor ages of 4--12 observations, and in eleven pairs at older tested ages up to 32. The share of the teacher's gain preserved remains complete on the real stream but is lower and generally declines with anchor age on the synthetic streams. In live replay with 1 ms arrivals, the synchronous large models cannot deliver any predictions on time while RESYNC serves at least 99.4% of arrivals within the interval, with 0.35--0.62 ms median latency on one CPU thread: around the same speed as or faster than the compact model, and 7.9--17.7$\times$ faster than synchronous large-model inference. RESYNC targets settings with background compute available but a strict per-observation response deadline.
Be the first to rate this article.

Comments & Questions (0)

Captcha
Please be respectful — let's keep the conversation friendly.

No comments yet

Be the first to comment!