Disney+ details DRanker transformer ranker, driving 1% engagement and 0.5% DAU lift
_reachsumit · x · 2026-10-14
Disney+ engineers published an arXiv paper on DRanker, the transformer-based multi-task ranking model behind the Disney+ homepage.
- Architecture: encoder-only Transformer over enriched engagement-history sequences, paired with a lightweight causal debiasing tower for popularity bias
- Serving: describes an inference stack that enables transformer-based ranking at scale
- Key finding: systematically evaluated more complex alternatives and found simpler designs consistently won on offline metrics, qualitative behavioral evaluation, and in production
- Impact: deployed globally, DRanker drives over 1% lift in key engagement metrics and 0.5% lift in DAU
The paper also shares practical lessons on data engineering, debiasing, and architectural choices.
More from Research
- PyVBBC 1.5 Released After 3 Years, Bringing Major Upgrades to Bayesian Model Inference — twiecki · 2026-10-14
- AgentHorizon: 1,373-task benchmark exposes unreliable LLM judges for computer-use agents — rohanpaul_ai · 2026-10-14
- ServiceNow and McGill's AgentHorizon benchmark shows LLM judges miss subtle agent failures — rohanpaul_ai · 2026-10-14
- Cambridge's van der Schaar to keynote on AI and oncology at Momentum 2026 — MihaelaVDS · 2026-10-14
- Falcon engineer asks: what does RL add over SFT in a single-step decision process? — antoine_chaffin · 2026-10-14
- Visual SLAM Match Moving: Reconstructing Camera Pose to Composite CG Into Footage — rsasaki0109 · 2026-10-14