Junbo Qi, Yi Zhang, Hanchu Ni, Che Liu, Zhimin Yao, Ruilin Yang, Xiancong Ren, Liangjian Wen, Wei Ge, Yuya Ieiri, Osamu Yoshie, Yong Dai, Xiaozhu Ju. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2026.

Recommender systems aim to leverage user interaction signals to recommend items that users are likely to be interested in. Motivated by the success of Large Language Model (LLMs), generative recommendation (GR) has recently gained increasing attention, typically following a two-stage paradigm: supervised fine-tuning followed by preference alignment. However, aligning generative recommenders with users' personalized preferences remains challenging, as user feedback is inherently heterogeneous and uncertain. Different types of user interaction signals reflect varying levels of intent and should therefore be modeled differently. In this work, we propose Adaptive Mix Preference Optimization (AMPO), an adaptive alignment framework that mixes likelihood and preference objectives with self-calibrated, sample-wise confidence adjustment. AMPO introduces an adaptive target margin that leverages the model's own probability ratio to modulate optimization strength: confident pairs receive full margins that reinforce correct rankings, while uncertain pairs receive reduced margins that prevent overfitting to ambiguous signals. Additionally, AMPO incorporates negative log-likelihood regularization on preferred items to counteract likelihood displacement, a phenomenon where contrastive objectives cause preferred and non-preferred probabilities to collapse simultaneously. Such a design eliminates the need for a reference model, yielding up to 2.5x speedup and 35% memory reduction. Extensive experiments on public benchmarks and a large-scale industrial dataset demonstrate consistent improvements in ranking metrics. Online A/B tests on a major e-commerce platform further confirm statistically significant gains in click-through and conversion rates. The code is available at https://github.com/jumbo-q/ampo.

Novelty recommendation sustains long-term user engagement by exposing users to content that is both relevant and meaningfully different from their recent consumption. In large-scale e-commerce, this requires composing coherent yet non-redundant recommendation lists, a task fundamentally constrained by the relevance-diversity trade-off. Large language models (LLMs) offer a unified generative paradigm for inferring user intent and producing semantically coherent candidates, yet industrial deployment faces two critical challenges: (i) scarce supervision for modeling novelty transitions and diversity-aware list construction, and (ii) reward granularity mismatch, where standard RL assigns coarse sequence-level rewards that fail to capture item-level redundancy and complementarity. We present BALANCE, a hierarchical reasoning-and-generation framework that decomposes novelty recommendation into three structured stages: generating a Novelty Tag for exploration direction, refining an Interest Topic for intent specification, and constructing a Recommendation List for facet coverage. We address data scarcity through a self-reflection pipeline that synthesizes high-quality supervision by integrating real behavior logs with structured rationales. We resolve granularity mismatch through Sequence-Item Policy Optimization (SIPO), which jointly optimizes sequence- and item-level objectives via granularity-aware advantage fusion. Extensive offline experiments and online A/B test on the JD.com recommender system, validate the performance of our method, highlighting its superior novelty and diversity without compromising relevance.

Generative Retrieval (GR) offers a promising paradigm for recommendation through next-token prediction (NTP). However, scaling it to large-scale industrial systems introduces three challenges: (i) within a single request, the identical model inputs may produce inconsistent outputs due to the pagination request mechanism; (ii) the prohibitive cost of encoding long user behavior sequences with multi-token item representations based on semantic IDs, and (iii) aligning the generative policy with nuanced user preference signals. We present GenRec, a preference-oriented generative framework deployed on the JD App https://www.jd.com that addresses above challenges within a single decoder-only architecture. For training objective, we propose Page-wise NTP task, which supervises over an entire interaction page rather than each interacted item individually, providing denser gradient signal and resolving the one-to-many ambiguity of point-wise training. On the prefilling side, an asymmetric linear Token Merger compresses multi-token Semantic IDs in the prompt while preserving full-resolution decoding, reducing input length by ~2× with negligible accuracy loss. To further align outputs with user satisfaction, we introduce GRPO-SR, a reinforcement learning method that pairs Group Relative Policy Optimization with NLL regularization for training stability, and employs Hybrid Rewards combining a dense reward model with a relevance gate to mitigate reward hacking. In month-long online A/B tests serving production traffic, GenRec achieves 9.5% improvement in click count and 8.7% in transaction count over the existing pipeline.