Guess Where You Go: Generative Next Point-of-Interest Recommendation in Amap¶

Fidelity: core mechanism. Paper results, local public-data measurements, and omissions are separate.

Paper information¶

Field	Details
Paper	arXiv 2607.26073
Organization	Amap / Alibaba
First public release	2026-07-13 (arXiv v1)
Authors' code	Public implementation
Adapter / method	`guess-where-you-go`
Local reproduction code	`src/auto_research/reproductions/guess_where_you_go/`

Original paper¶

Motivation and method¶

Encode the next point of interest as a semantic-ID sequence. Spatiotemporal features, curriculum pretraining and SFT, and EAKTO optimize longer-term feedback.

flowchart LR
 A["Spatiotemporal history and SIDs"] --> B["CPT and SFT"] --> C["EAKTO for next POI"]

Key figure from the paper¶

Figure 1: Core architecture, modules, and connections. From the original paper; copyright remains with the authors.

Core equations¶

\[ J=J_{SFT}+\lambda\mathbb E[A(s,a)\log\pi_\theta(a|s)] \]

Paper results¶

The main online result is P-CTR +5.83% in a one-month A/B test. Offline tables and online metrics use private or paper-specific protocols. The private paper protocol is not directly comparable with local MovieLens measurements.

Local reproduction¶

Local comparison: The baseline is the shared transition + content scorer. The treatment adds only guess-where-you-go on the same users, items, full-catalog candidates, and seed.

MovieLens-100K, 220 users and 360 items, seed 42: NDCG@10 0.0540 → 0.0591 (+9.38%); Hit@10 0.1091 → 0.1045. Validation selects mixture weights only; the test set is not used for tuning.

auto-research reproduce --paper guess-where-you-go --dataset-dir data --seed 42

See the fixed metrics.

Reproduction boundaries¶

The core mechanism runs on MovieLens-100K with the same split, fixed full-catalog candidates, and seed. Private features, the production foundation model, and online traffic are unavailable. The implementation has independent model state and scoring, rather than mapping a paper name to a shared weighted formula. See the adapter's omitted_core_components for omissions. Local changes and paper A/B gains are reported separately.