简体中文 · English
DIRECTOR: Parallel recommendation using transport-optimized dynamic indices¶
Core mechanism reproduction. The defining public-data mechanism runs locally; production models, private logs and serving are out of scope. A generic baseline is not a complete industrial system.
Paper information¶
| Field | Details |
|---|---|
| Paper | arXiv 2607.26418 |
| Institution | University of Science and Technology of China |
| First public release | 2026-07-29 (arXiv v1) |
| Author code | No official author code in the paper when checked on 2026-09-01. |
| Local adapter or method key | director |
| Local reproduction code | src/auto_research/reproductions/director/ |
Original paper summary¶
Background and main changes¶
Generate a dynamic-index distribution for each slot and constrain all slots jointly with Sinkhorn optimal transport. Global matching rather than independent argmax reduces conflicts and produces the entire slate in parallel.
flowchart LR
A["Public behavior and candidates"] --> B["DIRECTOR core mechanism"]
B --> C["Matched-budget output"]
A --> D["Paper-defined direct baseline"]
C --> E["Shared metrics and structural diagnostics"]
D --> E
Key figure from the paper¶
Figure 1: Core architecture, modules and connections. From the paper; copyright belongs to the authors. Click for the source.
Core equations¶
Offline and online results¶
- Two disjoint 10% buckets ran for 7 days: valid view +0.519% (95% CI [0.45%, 0.59%]), comments +0.695%, likes +0.330%, CPU -66.7%.
- These are paper online results, not conclusions from public local data.
Local reproduction¶
Comparison: Independent dynamic indices per position versus adding Sinkhorn transport and hard global matching. Seed 42 NDCG@10:
0.05334versus0.03857baseline. This cannot be extrapolated to production valid view.
Three seeds, means, standard deviations and 95% intervals: metrics/public-seeds42-44.json.
Scope¶
Uses public MovieLens features or auditable proxy tasks derived from them. It cannot reproduce private logs, online traffic allocation, production model scale or serving. Treat the local result as mechanism verification, not an online lift or directly comparable paper score.
