# Align3GR: Unified Multi-Level Alignment for LLM-based Generative Recommendation

arXiv: https://arxiv.org/abs/2511.11255 · HTML: https://mlsanity.com/paper/2511.11255/

## Key facts

- Cited by 5 papers in the map (data through 14 Oct 2026).
- Placed in LLM reasoning for recommendation, one of 17 directions in LLM reasoning.
- First posted on arXiv 14 Nov 2025.

## Placed in

- LLM reasoning / [LLM reasoning for recommendation](https://mlsanity.com/lr/llm-reasoning-for-recommendation/)

## Built on by (most cited first, 5 of 5)

- [End-to-End Semantic ID Generation for Generative Advertisement Recommendation](https://mlsanity.com/paper/2602.10445/) (2026)
- [PRISM: Purified Representation and Integrated Semantic Modeling for Generative Sequential Recommendation](https://mlsanity.com/paper/2601.16556/) (2026)
- [RGAlign-Rec: Ranking-Guided Alignment for Latent Query Reasoning in Recommendation Systems](https://mlsanity.com/paper/2602.12968/) (2026)
- [Awakening Dormant Users: Generative Recommendation with Counterfactual Functional Role Reasoning](https://mlsanity.com/paper/2602.13134/) (2026)
- [Deep Interest Mining for Intent-Enriched Semantic IDs in Multimodal Generative Recommendation](https://mlsanity.com/paper/2604.20861/) (2026)

## Abstract (arXiv)

Large Language Models (LLMs) demonstrate significant advantages in leveraging structured world knowledge and multi-step reasoning capabilities. However, fundamental challenges arise when transforming LLMs into real-world recommender systems due to semantic and behavioral misalignment. To bridge this gap, we propose Align3GR, a novel framework that unifies token-level, behavior modeling-level, and preference-level alignment. Our approach introduces: Dual tokenization fusing user-item semantic and collaborative signals. Enhanced behavior modeling with bidirectional semantic alignment. Progressive DPO strategy combining self-play (SP-DPO) and real-world feedback (RF-DPO) for dynamic preference adaptation. Experiments show Align3GR outperforms the SOTA baseline by +17.8% in Recall@10 and +20.2% in NDCG@10 on the public dataset, with significant gains in online A/B tests and full-scale deployment on an industrial large-scale recommendation platform.