GenFacet: end-to-end generative faceted search¶
Fidelity: core mechanism. Runs the defining mechanism verifiable on public data. Private data, production models, and serving remain outside scope.
Paper information¶
| Field | Details |
|---|---|
| Paper | arXiv 2603.19665 |
| Organization | JD.com |
| First public release | 2026-03-20 (arXiv v1) |
| Authors' code | No public author code was found on 2026-09-05. |
| Adapter / method | genfacet |
| Local reproduction code | src/auto_research/reproductions/genfacet/ |
Original paper¶
Motivation and method¶
Unify e-commerce faceted search as contextual facet generation and intent-driven query rewriting. Teacher distillation and multi-task GRPO preference alignment optimize downstream search satisfaction directly, linking facet selection to retrieval.
flowchart LR
A["Public behavior and candidates"] --> B["GenFacet core mechanism"]
B --> C["Matched-budget method output"]
A --> D["Common direct baseline"]
C --> E["Full-catalog ranking and structural diagnostics"]
D --> EKey figure from the paper¶
Figure 1: Overview of the paper's method and key components. From the original paper; copyright remains with the authors.
Core equations¶
Paper results¶
JD.com's online A/B test reports facet CTR +42.0% and user conversion +2.0%. This is paper evidence, not a local public-data outcome.
Local reproduction¶
Local comparison: Under the same full-catalog protocol, baseline NDCG@10 is 0.05401 and GenFacet NDCG@10 is 0.05171, -4.26% relative. The local proxy does not extrapolate to online lift.
See the three-seed artifact for complete results, means, standard deviations, and 95% confidence intervals.
Reproduction boundaries¶
The local run uses a public MovieLens-1M subset and auditable proxies to validate the core computation. It does not reproduce private logs, production models, online traffic assignment, or serving. No claim is made to reproduce absolute paper metrics or online gains.