ZoRRO: personalized news recommendation with no trained parameters¶

简体中文

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 2607.10910
Organization	Technical University of Denmark
First public release	2026-07-12 (arXiv v1)
Authors' code	johanneskruse/zorro
Adapter / method	`zorro`
Local reproduction code	`src/auto_research/reproductions/zorro/`

Original paper¶

Motivation and method¶

Without model training, combine time decay, semantic similarity, and category relations among historical articles for high-throughput news-candidate ranking.

flowchart LR
  A["Public behavior and candidates"] --> B["ZoRRO core mechanism"]
  B --> C["Matched-budget method output"]
  A --> D["Paper-defined direct baseline"]
  C --> E["Common metrics and structural diagnostics"]
  D --> E

Key figure from the paper¶

Figure 2: Overview of the paper's method and key components. From the original paper; copyright remains with the authors.

Core equations¶

\[ s(i|H)=\\lambda_h\\sum_{j\\in H}w_j\\cos(e_j,e_i)+\\lambda_c\\sum_{j\\in H}w_j\\mathbf1[c_j=c_i]. \]

Paper results¶

In a six-day real-traffic experiment at Ekstra Bladet, CTR is 4.19%, close to NRMS at 4.33% and above the popularity baseline at 2.96%. 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 ZoRRO NDCG@10 is 0.05201, -3.70% relative, a negative public-proxy result. The local proxy does not extrapolate to online lift.

See the three-seed artifact for complete results, means, standard deviations, and 95% confidence intervals.

auto-research reproduce --paper zorro --dataset-dir data --seed 42

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.