Sequential recommendation models dynamic user preferences but is constrained by the long-tail distribution of real-world interactions. While LLM-derived text embeddings provide semantic priors to alleviate data sparsity, existing adaptation methods are hindered by popularity bias during semantic projection and semantic prior erosion during sequence training. In this paper, we propose Adversarial Debiasing and Asymmetric Alignment (ADAA), a framework designed to alleviate these representation bottlenecks. Specifically, ADAA incorporates an adversarial debiasing mechanism to decouple interaction frequency factors from the projected embeddings, yielding debiased semantic representations. Furthermore, it employs an asymmetric anchoring strategy during the sequence training phase. This strategy restricts the optimization space of sparse items to preserve their injected semantic priors, while permitting head items to flexibly fit sequential transition patterns. Through a parameter-level injection mechanism, ADAA integrates offline-aligned semantics without introducing additional parameters to the online architecture. Extensive experiments on three real-world datasets demonstrate that ADAA improves long-tail recommendation accuracy while preserving the online inference efficiency of the original sequential backbone.

Recently, Large Language Models (LLMs) have demonstrated remarkable capabilities in natural language understanding, reasoning, and generation, prompting the recommendation community to leverage these powerful models to address fundamental challenges in traditional recommender systems, including limited comprehension of complex user intents, insufficient interaction capabilities, and inadequate recommendation interpretability. This survey presents a comprehensive synthesis of this rapidly evolving field. We consolidate existing studies into three paradigms: (i) recommender-oriented methods, which directly enhance core recommendation mechanisms; (ii) interaction-oriented methods, which conduct multi-turn conversations to elicit preferences and deliver interpretable explanations; and (iii) simulation-oriented methods, that model user-item interactions through multi-agent frameworks. Then, we dissect a four-module agent architecture: profile, memory, planning, and action. Then we review representative designs, public datasets, and evaluation protocols. Finally, we give the open challenges that impede real-world deployment, including cost-efficient inference, robust evaluation, and security.

Recently slot filling has witnessed great development thanks to deep learning and the availability of large-scale annotated data. However, it poses a critical challenge to handle a novel domain whose samples are never seen during training. The recognition performance might be greatly degraded due to severe domain shifts. Most prior works deal with this problem in a two-pass pipeline manner based on metric learning. In practice, these dominant pipeline models may be limited in computational efficiency and generalization capacity because of non-parallel inference and context-free discrete label embeddings. To this end, we re-examine the typical metric-based methods, and propose a new adaptive end-to-end metric learning scheme for the challenging zero-shot slot filling. Considering simplicity, efficiency and generalizability, we present a cascade-style joint learning framework coupled with context-aware soft label representations and slot-level contrastive representation learning to mitigate the data and label shift problems effectively. Extensive experiments on public benchmarks demonstrate the superiority of the proposed approach over a series of competitive baselines.