Sequential Modality Dropout: A 4-Line Fix for Robust Multi-Modal Recommenders
_reachsumit · x · 2026-08-12
Real-world product catalogs often miss images or text, causing multi-modal recommenders trained on complete data to lose significant accuracy during deployment.
This paper proposes Sequential Modality Dropout (SMD): independently erasing image or text streams with probability $p$ during training. This architecture-agnostic, four-line change significantly improves retention. Under an extreme 95% missing rate, the method retains 61% of HR@10 accuracy, a vast improvement over the 22% baseline.
More from Research
- Reconstructing Articulated Objects from a Single Rest-State Image — snu · 2026-08-12
- DistilVDR: A 524M Compact Visual Document Retriever Distilled from 8B — nanovdr · 2026-08-12
- AdvFD: Boosting Visual Generation via Adversarial Fréchet Distance Loss — Kwai-Kolors · 2026-08-12
- AI Agents Still Can't Conduct Open-Ended Research Due to Cognitive Failures — mikeflache · 2026-08-12
- A Simple Equation to Solve Trainer-Inference Mismatch in LLMs — willccbb · 2026-08-12
- EigenLabs Builds Multi-Agent Environments for Autonomous Scientific Research — gajesh · 2026-08-12