Localization Boosting for Growth Markets: Mitigating Cross-Locale Behavioral Bias in Learning-to-Rank
About
Adobe Express is expanding internationally, but the US has a disproportionately large content supply and interaction volume. Learning-to-rank (LTR) models trained primarily on behavioral feedback inherit this imbalance: templates popular in US are over-served in non-US locales. This cross-locale exposure bias suppresses local content discoverability and degrades ranking quality in growth locales. We show that click-only training suppresses semantically informative localization features. Adding vision-language model (VLM) graded relevance labels as auxiliary supervision alongside clicks improves semantic alignment but does not preserve local content visibility. We propose a multi-objective framework combining behavioral supervision, VLM-derived relevance signals, and locale-aware boosting. Across five locales, the resulting model improves relevance while restoring stable localization, demonstrating the importance of disentangling exposure from semantic supervision.
Related benchmarks
| Task | Dataset | Result | Rank | |
|---|---|---|---|---|
| Learning to Rank | DE (Germany) region query set (test) | Precision@2081 | 2 | |
| Learning to Rank | GB (Great Britain) region query set (test) | Precision@2088 | 2 | |
| Learning to Rank | JP (Japan) region query set (test) | Precision@2084 | 2 | |
| Learning to Rank | FR (France) region query set (test) | Precision@2082 | 2 | |
| Ranking | Adobe Express template interaction logs (US) | Ranking Quality0.68 | 2 | |
| Ranking | Adobe Express template interaction logs (DE) | Ranking Quality67.3 | 2 | |
| Ranking | Adobe Express template interaction logs FR | Ranking Quality66.4 | 2 | |
| Ranking | Adobe Express template interaction logs (GB) | Ranking Quality0.682 | 2 | |
| Ranking | Adobe Express template interaction logs (JP) | Ranking Quality0.714 | 2 | |
| Learning to Rank | US (United States) region query set (test) | Prec@2067 | 2 |