跳转至

POEM: partial-order-enhanced real-time sequence modeling

简体中文

Fidelity: core mechanism. Runs the defining mechanism verifiable on public data. Private data, production models, and serving are outside scope.

Paper information

Field Details
Paper arXiv 2606.29946
Organization Kuaishou
First public release 2026-06-29 (arXiv v1)
Authors' code The paper did not provide official or author code when checked on 2026-09-02.
Adapter / method poem
Local reproduction code src/auto_research/reproductions/poem/

Original paper

Motivation and method

Convert relative ranks from several ranking objectives on the current request into partial-order signals. Construct training sequences that better reflect current interests than chronology alone.

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

Key figure from the paper

POEM: partial-order-enhanced real-time sequence modeling — Figure 1

Figure 1: Overall workflow, key stages, and data flow. From the original paper; copyright remains with the authors.

Core equation

\[ i\\succ j\\iff \\sum_m\\mathbf1[r_m(i)<r_m(j)]>\\sum_m\\mathbf1[r_m(j)<r_m(i)]. \]

Paper results

In a seven-day Kuaishou A/B test with 5% traffic per arm, time per user increased +0.249% on the main app and +0.213% on the Lite app. These are paper-reported online measurements only, not local public-data results.

Local reproduction

Local comparison: Under the same full-catalog protocol, NDCG@10 is 0.05401 for the baseline and 0.05009 for POEM, -7.25% relative. This is a negative result on a public proxy task and does not extrapolate to production lift.

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

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

Reproduction boundaries

The local run uses a public MovieLens-1M subset and auditable proxy objectives. It does not reproduce private company logs, online traffic assignments, production model scale, or serving. This is core-mechanism validation, not reproduction of the paper's absolute results or online gain.