Full talk · edited transcript

X-split ARGUS: An Autoregressive Encoder-Decoder for Ad Ranking

Alexander Ploshkin23 May 2026

Alexander Ploshkin, group lead (early-fusion architectures), recommendation technologies service, Yandex.

The speaker leads an R&D group in the recommendation technologies service: about three years in recommender systems, and five years in computer vision before that. The main claim of the talk: the team developed the next generation of ARGUS for ad ranking, which on top of the previous one gives ×2 profit on the key advertising metric.

Outline: a short RecSys primer; how recommendation technologies are applied in advertising; the path to the X-split ARGUS architecture with its pros and cons; training details; results and expectations.

Continue reading — it's free

Google or a one-time email code. No password.