LLM-based recommendation agents personalize what they know through evolving per-user semantic memory, yet how they perform reasoning relies on a universal, static system prompt shared across all users. When a recommendation fails, the agent only updates its memory, but never examines or revises the reasoning that leads to the failure. To close this gap, we propose SAGER (Self-Evolving Agent for Personalized Recommendation), the first framework to equip each user with a dedicated, continuously evolving policy skill. SAGER incorporates three tailored mechanisms for recommendation scenarios: (i) a two-representation skill architecture constrained by a Cognitive Injection Boundary, decoupling a rich evolution substrate from a minimal inference-time injection; (ii) a contrastive differential chain-of-thought engine that diagnoses ranking errors against unchosen near-misses under a refine-not-replace principle; and (iii) skill-augmented listwise reasoning that mitigates the Score-Tie Bottleneck, enabling the slim skill to serve as an active tie-breaker. Evaluated on four public benchmarks, SAGER achieves state-of-the-art accuracy. Furthermore, plugging its policy skill layer into three structurally distinct host agents delivers consistent 9%–12% gains on the H@1 metric. These results indicate that policy skill evolution provides architecture-agnostic improvements that are largely complementary to memory accumulation.

Conversational news recommendation requires grounding each suggestion in a rapidly evolving article corpus while addressing implicit user intents that lack explicit retrievable keywords. To characterize this scenario, we identify 6 intent types from production dialogues: five are implicit and pose fundamental challenges to standard RAG pipelines, forming a critical retrieve-first bottleneck. To address these issues, we introduce intent-driven Semantic ID (SID) generation under a Generate-then-Match paradigm. With two-stage training that consists of multi-task SID alignment and GPT-4 Chain-of-Thought distillation, an LLM maps diverse intents to hierarchical SID prefixes, which are then fuzzy-matched to the current news pool to guarantee fully grounded recommendations. Profile-Aware Dual-Signal Reasoning (PADR) further enables cold-start users to obtain valid recommendations using only profiles. On a mainstream Chinese news platform, our 7B model achieves 0% hallucination and 12.4% L1 match in the 152K open-generation SID space (4 × random baseline). It matches GPT-4+Hybrid RAG on L1 while surpassing it on finer-grained metrics (L2 2 ×, Category +1.2pp) at ∼ 100 × lower cost. Cold-start users, where existing baselines score 0%, achieve 18.0% L1 (6 × random), the highest among all user groups.

In-context learning of large-language models (LLMs) has achieved remarkable success in the field of natural language processing, while extensive case studies reveal that the single-step chain-of-thought prompting approach faces challenges such as attention diffusion and inadequate performance in complex tasks like text-to-SQL. To improve the contextual learning capabilities of LLMs in text-to-SQL, a workflow paradigm method is proposed, aiming to enhance the attention and problem-solving scope of LLMs through decomposition. Specifically, the information determination module for eliminating redundant information and the brand-new prompt structure based on problem classification greatly enhance the model’s attention. Additionally, the inclusion of self-correction and active learning modules greatly expands the problem-solving scope of LLMs, hence improving the upper limit of LLM-based approaches. Extensive experiments conducted on three datasets demonstrate that our approach outperforms other methods by a significant margin. About 2-3 percentage point improvements compared to the existing baseline on the Spider Dev, Spider-Realistic, and Bird Dev datasets and new SOTA results on the Spider Test dataset are achieved. Our code is available on GitHub: https://github.com/FlyingFeather/DEA-SQL.

Unsupervised Domain Adaptation (UDA) aims to transfer the knowledge of source domain to the unlabeled target domain. Existing methods typically require to learn to adapt the target model by exploiting the source data and sharing the network architecture across domains. However, this pipeline makes the source data risky and is inflexible for deploying the target model. This paper tackles a novel setting where only a trained source model is available and different network architectures can be adapted for target domain in terms of deployment environments. We propose a generic framework named Cross-domain Knowledge Distillation (CdKD) without needing any source data. CdKD matches the joint distributions between a trained source model and a set of target data during distilling the knowledge from the source model to the target domain. As a type of important knowledge in the source domain, for the first time, the gradient information is exploited to boost the transfer performance. Experiments on cross-domain text classification demonstrate that CdKD achieves superior performance, which verifies the effectiveness in this novel setting.