SASRec: causal self-attention for next-item recommendation

What changed

Adapted self-attention with causal masking, learned item/position embeddings and residual feed-forward blocks to predict next items from user action histories.

Scientific scope

Adapted self-attention with causal masking, learned item/position embeddings and residual feed-forward blocks to predict next items from user action histories. UC San Diego academic work, not retroactively a company invention. Uses a finite truncated/padded history and inherited attention machinery. Attention weights and benchmark gains do not prove causally identified preferences or superiority for every sparsity level; model and evaluation choices matter.

Evidence

The linked original sources describe the stated contribution and its evidence limits. WRS has not independently rerun the experiments or certified a proof. UC San Diego academic work, not retroactively a company invention. Uses a finite truncated/padded history and inherited attention machinery. Attention weights and benchmark gains do not prove causally identified preferences or superiority for every sparsity level; model and evaluation choices matter.

Contributors

• Source-listed authors and technical contributors · SCIENCE: Source-listed authors or explicitly credited technical contributors for the cited versions. Family entries retain source-specific attribution; a combined list is not one original-paper byline or equal individual work. Names alone do not establish resolved identities.

• Self-Attentive Sequential Recommendation

Provisional WRS assessment, edition 2026-10-16.1. Independent expert calibration is pending. This grade does not certify correctness or assign ownership.

Full scientific record · Browse contribution grades · Request a correction