MLCC: multilayer compressed cross networks¶
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 2602.12041 |
| Organization | Bilibili |
| First public release | 2026-02-12 (arXiv v1) |
| Authors' code | https://github.com/shishishu/MLCC |
| Adapter / method | mlcc |
| Local reproduction code | src/auto_research/reproductions/mlcc/ |
Original paper¶
Motivation and method¶
Each layer compresses high-dimensional features, applies dynamic explicit feature crosses in the compact space, and projects them back into the backbone. MC-MLCC uses parallel subspaces to increase expressivity while controlling parameter and FLOP growth.
flowchart LR
A["Public behavior and candidates"] --> B["MLCC 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: Overall workflow, stages, and data flow. From the original paper; copyright remains with the authors.
Core equations¶
Paper results¶
The paper reports up to +0.52% AUC, over 26× lower FLOPs, and a 32% increase in Advertiser Value in production experiments. This is paper evidence, not a local public-data outcome.
Local reproduction¶
Local comparison: Under the same full-catalog MovieLens protocol, baseline NDCG@10 is 0.05401 and treatment NDCG@10 is 0.05530, +2.40% relative. The local proxy and production task differ, so there is no extrapolation 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 fixed public MovieLens 100K subset and auditable proxies to check the core computation only. It does not reproduce private logs, the production foundation model, online traffic assignment, or serving. There is no claim to reproduce absolute paper metrics or online gains.