EASformer: Environment-Action-State Alignment for Industrial Generative Ranking

Key points

• Modern industrial ranking systems increasingly adopt generative recommendation (GR) paradigms, in which user behavior sequences and non-sequential (NS) features are tokenized and jointly modeled by Transformer-style backbones.1
• However, sequence and NS representations often carry asymmetric semantics: historical representations may combine environment context with observed-action semantics, whereas candidate-side NS representations contain no observed action but incorporate richer supplementary context, such as user-item interaction features and recent behavior lists.1
• EASformer constructs an explicitly aligned context environment space shared by historical and non-sequence tokens, while disentangling user actions from historical token and candidate-specific context form non-sequence tokens.1
• Based on this aligned space, EASformer enables complementary historical action retrieval and state-conditioned action prediction.1

Sources (1)

• [1]EASformer: Environment-Action-State Alignment for Industrial Generative Ranking
  arXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 20, 06:08 AM

  To fill this gap, we propose EASformer, an Environment-Action-State disentangled architecture for industrial app ranking.

  Modern industrial ranking systems increasingly adopt generative recommendation (GR) paradigms, in which user behavior sequences and non-sequential (NS) features are tokenized and jointly modeled by Transformer-style backbones.

Extractive summary: sentences quoted from the sources.