ByteDance Unveils GEAR: End-to-End Generative Ad Retrieval Live on Douyin
_reachsumit · x · 2026-10-01
ByteDance published a paper detailing GEAR, an end-to-end generative ad retrieval framework deployed on Douyin that jointly trains the tokenizer, generator, and reranker.
- Two coupled bottlenecks: representation collapse (the item tokenizer degenerates under distribution shifts) and item collisions (distinct items sharing identical token sequences at scale) — expanding the codebook to fix collisions worsens collapse.
- BasisVQ re-parameterizes the codebook via an orthogonal basis for global gradient sharing and rigid latent-space rotation, stabilizing gradients without ad-hoc heuristics; prefix-aware BasisRQ boosts expressiveness at the same asymptotic complexity.
- A context-conditioned reranking head disambiguates colliding items with minimal compute overhead.
More from Research
- EIG publishes most rigorous analysis yet of AI's real effect on US entrepreneurship — robseamans · 2026-10-01
- AI "mind-reading" tool reconstructs what you see from brain scans — ChuckDBrooks · 2026-10-01
- Training Qwen3-VL-4B into an AR image generator with Cosmos tokenizer — ostrisai · 2026-10-01
- CMU releases CUA-SWE, a benchmark uniting computer-use agents with visual software engineering — CarnegieMellonU · 2026-10-01
- New paper: principal component regression beats every monotone spectral filter, incl. ridge and GD — kfountou · 2026-10-01
- MIT's xvr matches X-rays to 3D scans in seconds with sub-millimeter precision for surgery — MacrinePhD · 2026-10-01