LLM-based rerankers in recommender systems are inherently constrained by limited context windows: they can only reorder a small retrieved candidate set, so any relevant item missed by upstream retrieval cannot be recovered at ranking time. To mitigate this dependency, we propose RoutePO, which repositions the LLM from an item ranker to a recall strategy router for multi-channel recall. Conditioned on a user’s profile/context (interaction history and optional demographics), the LLM outputs personalized fusion weights over heterogeneous retrieval experts, enabling dynamic candidate aggregation without exposing long candidate lists from the entire item pool to the prompt. We train the router via a synergistic two-stage alignment process: Coarse Routing Alignment (SFT) warm-starts the policy with a quality-aware oracle prior over retrieval experts, and Fine-grained Metric Optimization (RL) directly optimizes recall-stage ranking metrics (e.g., NDCG) by sampling simplex-valued routing weights and rewarding fused-list quality. To support continuous exploration over the simplex under GRPO-style optimization, we introduce SofT-GRPO based on Gumbel–Softmax reparameterization with nucleus restriction and KL regularization. Finally, we instantiate fusion with both weighted score merging and a weight-aware interleaving strategy (Snake Fusion) that preserves within-channel rankings. Experiments on three real-world datasets demonstrate consistent gains over strong individual recallers and multi-channel fusion baselines, including up to 15.88% relative improvement in NDCG@50 on Food. Our implementation is available on https://github.com/zhengzaiyi/RoutePO.

Reasoning distillation from powerful teacher models to smaller students faces the Gap Curse: as teachers grow more sophisticated, their complex distributions increasingly diverge from what students can approximate, causing performance degradation. Existing mitigation strategies either filter out challenging examples through data selection or introduce weaker intermediate assistant models, inherently compromising supervision coverage or quality. We propose Teacher Alignment, which directly adapts the teacher toward the student’s distribution without discarding data or degrading reasoning quality. However, naive alignment through standard knowledge distillation triggers catastrophic collapse of the teacher’s reasoning capabilities. To address this, we reformulate teacher alignment as reinforcement learning and introduce TeacherGRPO, built on Group Relative Policy Optimization with two key innovations: (i) Curriculum Selective Alignment applies dual token- and distribution-level curricula to focus rewards on high-signal reasoning gaps while filtering noise from trivial tokens and uncertain tail distributions, and (ii) Importance-Adaptive Length Regularization selectively penalizes verbose redundancy while preserving pedagogically critical reasoning steps. The aligned teacher then distills knowledge to students via standard pipelines. Extensive experiments show TeacherGRPO significantly outperforms baselines across diverse reasoning benchmarks and distillation methods. Our code is available at https://anonymous.4open.science/r/TeacherGRPO-290F/README.md.

Large Language Models (LLMs) have achieved remarkable success in natural language processing by encoding extensive knowledge, but their utility relies on timely updates as human knowledge keeps evolving. In this paper, we investigate the problem of LLM knowledge updates, which requires simultaneously unlearning unwanted information and learning new knowledge. Existing approaches that tackle unlearning and learning separately encounter task conflicts and knowledge management issues when applied to comprehensive knowledge updates.In this paper, we validate our findings with theoretical analysis and empirical evidence, and propose LOKA, a conflict-aware framework for Large language mOdel Knowledge updAtes. During training, LOKA introduces an adaptive knowledge memory approach in which updated knowledge is allocated across multiple memory units. During inference, LOKA retrieves the most relevant memory unit from the knowledge memory and integrates it with the original LLM to apply updated knowledge, while a learning-based router controls the activation of the knowledge memory to improve knowledge utilization. Extensive experiments demonstrate the efficacy of LOKA in achieving accurate, flexible, and conflict-aware knowledge updates.

Retrieval-Augmented Generation (RAG) is introduced to enhance Large Language Models (LLMs) by integrating external knowledge. However, conventional RAG approaches treat retrieved documents as independent units, often overlooking their interdependencies. Hybrid-RAG, a recently proposed paradigm that combines textual documents and graph-structured relational information for RAG, mitigates this limitation by collecting entity documents during graph traversal. However, existing methods only retrieve related documents from local neighbors or subgraphs in the knowledge base, which often miss relevant information located further away from a global view. To overcome the above challenges, we propose CoRAG that dynamically chooses whether to retrieve information through direct textual search or explore graph structures in the knowledge base. Our architecture blends different retrieval results, ensuring the potentially correct answer is chosen based on the query context. The textual retrieval components also enable global retrieval by scoring non-neighboring entity documents based on semantic relevance, bypassing the locality constraints of graph traversal. Experiments on semi-structured (relational and textual) knowledge base QA benchmarks demonstrate the outstanding performance of CoRAG.

In recent years, Graph Neural Networks (GNNs) have become successful in molecular property prediction tasks such as toxicity analysis. However, due to the black-box nature of GNNs, their outputs can be concerning in high-stakes decision-making scenarios, e.g., drug discovery. Facing such an issue, Graph Counterfactual Explanation (GCE) has emerged as a promising approach to improve GNN transparency. However, current GCE methods usually fail to take domain-specific knowledge into consideration, which can result in outputs that are not easily comprehensible by humans. To address this challenge, we propose a novel GCE method, LLM-GCE, to unleash the power of large language models (LLMs) in explaining GNNs for molecular property prediction. Specifically, we utilize an autoencoder to generate the counterfactual graph topology from a set of counterfactual text pairs (CTPs) based on an input graph. Meanwhile, we also incorporate a CTP dynamic feedback module to mitigate LLM hallucination, which provides intermediate feedback derived from the generated counterfactuals as an attempt to give more faithful guidance. Extensive experiments demonstrate the superior performance of LLM-GCE. Our code is released on https://github.com/YinhanHe123/new_LLM4GNNExplanation.