yufan/amazon2023-user-interactions

Amazon Reviews 2023 — User Interactions (5-core, leave-one-out, sequential)

User–item interaction data for five Amazon Reviews 2023 categories, processed into ready-to-use sequential / generative recommendation splits with the de-facto standard recipe (5-core filtering → chronological ordering → leave-one-out split).

Every record keeps the timestamp, and the splits are byte-for-byte reproducible from the official Amazon Reviews 2023 release; the statistics also match, exactly, numbers reported by multiple peer-reviewed papers (see § Validation).

Item content features (title, images, price, brand, …) are in the companion dataset yufan/amazon2023-item-metadata, joined by parent_asin.

All files are JSON Lines (.jsonl).

Subsets (configs) & splits

Configs are named <processing>_<Category> and grouped by processing method; each has train / validation / test splits (single train for interactions_*). 5 categories × 4 processing variants = 20 configs.

Processing variant (config prefix)	What it is
seq_maxlen50_*	leave-one-out sequential split, history ≤ 50 (ID-based / SASRec-style)
seq_maxlen20_*	leave-one-out sequential split, history ≤ 20 (semantic-ID / generative)
last_out_*	leave-one-out direct split (no history column; MF/BPR-style)
interactions_*	5-core interaction table (no split applied; single train)

Categories: Video_Games, Industrial_and_Scientific, Beauty_and_Personal_Care, Musical_Instruments, Books.

from datasets import load_dataset

# sequential split (history <= 50), Video Games
ds = load_dataset("yufan/amazon2023-user-interactions", "seq_maxlen50_Video_Games")
train, valid, test = ds["train"], ds["validation"], ds["test"]

# item content features (companion dataset), same category
meta = load_dataset("yufan/amazon2023-item-metadata", "Video_Games")["train"]

Which config do I use?

If your model is …	Use
Sequential / ID-based (SASRec, GRU4Rec)	seq_maxlen50_<Category>
Generative / Semantic-ID (TIGER)	seq_maxlen20_<Category>
Non-sequential (MF, BPR, LightGCN)	last_out_<Category>
Custom (re-split yourself)	interactions_<Category>

Repository layout

interactions/<Category>.jsonl
last_out/<Category>.{train,valid,test}.jsonl
seq_maxlen20/<Category>.{train,valid,test}.jsonl
seq_maxlen50/<Category>.{train,valid,test}.jsonl
stats.json   stats.md

Record schemas

interactions_* and last_out_* — one interaction per line:

{"user_id": "AEVPP…N7WFQ", "parent_asin": "B09JY72CNG", "rating": 4.0, "timestamp": 1630594913298}

seq_maxlen20_* / seq_maxlen50_* — same, plus the chronological history (list of parent_asin, most-recent-last, capped at the max length):

{"user_id": "AEVPP…N7WFQ", "parent_asin": "B09JY72CNG", "rating": 4.0,
 "timestamp": 1630594913298, "history": ["B08R5B7YS4", "B0863MT183", "…"]}

parent_asin is the Amazon product (parent) id — the standard item id for the Amazon Reviews 2023 benchmark, and the join key to the item-metadata dataset.

---

Data-processing protocol

Source: the official Amazon Reviews 2023 (McAuley-Lab/Amazon-Reviews-2023, website).

1. Input: official de-duplicated raw interactions (benchmark/.../rating_only).
2. 5-core filtering: iteratively drop users and items with < 5 interactions until every remaining user and item has ≥ 5.
3. Chronological ordering: per user, sort ascending by timestamp; keep the earliest occurrence of each (user, item) pair.
4. Leave-one-out split: per user, last → test, second-to-last → validation, the rest → train.
5. Sequential history: attach each target's prior history, truncated to the most recent 20 or 50 items.

The 5-core and leave-one-out logic is borrowed verbatim from the official benchmark scripts (kcore_filtering.py, last_out_split.py), so the splits reproduce the official 5core/last_out exactly. This recipe was popularised by TIGER (Rajput et al., NeurIPS 2023, arXiv:2305.05065).

---

Validation / double-check

Level 1 — exact match to the official benchmark. For every category, our re-derived 5-core interaction count and number of test users are identical to the official McAuley-Lab/Amazon-Reviews-2023 5core/rating_only and 5core/last_out files (asserted in the build script).

Level 2 — exact match to numbers reported in peer-reviewed papers. Under the same conditions (Amazon Reviews 2023 · same category · 5-core · leave-one-out), our (users / items / interactions) match to the digit the dataset statistics reported by ~15 independent papers:

Category	Ours — users / items / interactions	Reported identically by (same 5-core + LOO)
Musical_Instruments	57,439 / 24,587 / 511,836	9 papers — UTGRec, MTGRec, LARES, CCFRec, Pctx, LLaDA-Rec, HSTU-BLaIR, Augment-or-Not, MLPs
Video_Games	94,762 / 25,612 / 814,586	4 — GrIT, Not-Just-What-But-When, MLPs, HSTU-BLaIR
Industrial_and_Scientific	50,985 / 25,848 / 412,947	4 — GrIT, MLPs, Augment-or-Not, Token-Weighted
Beauty_and_Personal_Care	729,576 / 207,649 / 6,624,441	2 — AlphaFree, Closing-the-Gap
Books	776,370 / 495,063 / 9,488,297	2 — Not-Just-What-But-When, Hi-SAM (≈, rounded)

≈ 20 reported data points across ~15 independent papers, all matching exactly (a few differ by 1 from rounding, e.g. HSTU-BLaIR's 814,585 vs 814,586).

Cross-validation references (papers reporting identical statistics)

UTGRec (2504.04405) · MTGRec / Pre-training Generative Rec. (2504.04400) · LARES (2505.16865) · CCFRec / Bridging Textual-Collaborative (2503.12183) · Pctx (2510.21276) · LLaDA-Rec (2511.06254) · HSTU-BLaIR (2504.10545) · Augment-or-Not (2505.23053) · MLPs (2605.12617) · GrIT (2602.19728) · Not-Just-What-But-When (2507.23209) · Token-Weighted (2601.17787) · AlphaFree (2603.02653) · Closing-the-Gap (2508.14910) · Hi-SAM (2602.11799)

Near-matches confirm the same base. A few papers use the same 5-core + LOO but report slightly smaller counts because they add one extra filter — e.g. ReSID drops items lacking structured side-info (≈2–6 % smaller across every category: VG 94,515; Musical 57,359; Books 775,503), Multimodal-GR drops items without images (Beauty 724,796), SPARC keeps only rating ≥ 4 (Books 459,133). Being only marginally smaller is itself evidence that the underlying 5-core + LOO base is identical.

Larger differences only arise when the conditions differ (excluded by the precondition above): the old 2014 Beauty (22,363 users; BSARec, RecCocktail, Understanding-GR, Beyond-Unimodal), sub-sampling (Efficient-Responsible 2,289; Heterogeneous 10k/domain), or omitting 5-core / temporal-truncation (R²ec, Reinforced-PO) — these are the papers' own design choices, not discrepancies in this dataset.

---

Statistics (after 5-core)

Category	#Users	#Items	#Interactions	Avg len	Median	Sparsity
Video_Games	94,762	25,612	814,586	8.5961	6	0.99966437
Industrial_and_Scientific	50,985	25,848	412,947	8.0994	6	0.99968665
Beauty_and_Personal_Care	729,576	207,649	6,624,441	9.0799	7	0.99995627
Musical_Instruments	57,439	24,587	511,836	8.9109	7	0.99963757
Books	776,370	495,063	9,488,297	12.2214	7	0.99997531

Leave-one-out split row counts

Category	direct train	valid	test	seq20 train	seq50 train
Video_Games	625,062	94,762	94,762	530,300	530,300
Industrial_and_Scientific	310,977	50,985	50,985	259,992	259,992
Beauty_and_Personal_Care	5,165,289	729,576	729,576	4,435,713	4,435,713
Musical_Instruments	396,958	57,439	57,439	339,519	339,519
Books	7,935,557	776,370	776,370	7,159,187	7,159,187

(valid = test = #users by construction. Sequential train uses the rolling next-item scheme; the 20/50 variants share the same row count, differing only in history length.)

---

Intended uses & limitations

Next-item sequential recommendation, generative / semantic-ID recommendation, content/multimodal recommendation (with the companion item-metadata dataset), cold-start studies. Implicit feedback (all retained reviews are positives; use rating for explicit labels). No leakage by construction. The two largest categories (Books, Beauty) are processed from the official pre-filtered 5-core data (verified equivalent).

License & citation

Derived from Amazon Reviews 2023 (McAuley Lab); for research use. Please cite:

Yupeng Hou, Jiacheng Li, Zhankui He, An Yan, Xiusi Chen, Julian McAuley. Bridging Language and Items for Retrieval and Recommendation. 2024. arXiv:2403.03952 · https://amazon-reviews-2023.github.io/

Shashank Rajput et al. Recommender Systems with Generative Retrieval (TIGER). NeurIPS 2023. arXiv:2305.05065