Writing a literature review requires a deep understanding of the relationships among cited papers: how they build on, challenge, or offer alternative perspectives to one another. We present Graph-Reasoning Aided Survey Planning (GRASP), a framework combining LLM planning for related work generation with graph algorithms to extract key relationships among cited papers. Our two-layer graph structure consists of a Graph of Thoughts and an Argument-Counteragument Planning Network, representing the cited papers at different levels of granularity, and we apply topology-aware pruning via a Steiner tree to identify the core inter-paper relationships captured in our graph. Our citation analysis-based evaluation shows that GRASP generates RWS that closely match human-written targets in terms of the discourse roles, intents, and grouping of citations.

Unlike conventional static graphs, dynamic graphs evolve over time as their links change. Recent benchmarks, such as the Dynamic Text-attributed Graph Benchmark, further enable the study of dynamic graph learning with rich textual information on nodes and edges. The success of graph instruction tuning on static text-attributed graphs motivates us to extend this approach to dynamic graphs. However, previous works suggest directly using LLMs as predictors on dynamic graphs often underperforms traditional dynamic graph models. To address this issue, we propose a novel two-stage training framework consisting of Dynamic Text-attributed Graph Instruction Tuning (DTGIT) followed by standard supervised training. Unlike prior approaches that employ instruction tuning to enhance LLMs’ understanding of graph structures and directly use LLMs as predictors for downstream tasks, our proposed instruction tuning framework serves as a warm-up stage before standard supervised training. Specifically, during the instruction tuning stage, the embeddings produced by the dynamic graph model are aligned with temporal semantics in the natural language space, while the LLM itself is not directly used as a predictor for downstream tasks. In the second stage, the dynamic graph model is initialized with the knowledge acquired during instruction tuning and further fine-tuned in a supervised manner for downstream tasks. Extensive experiments demonstrate that existing dynamic graph models (e.g., DyGFormer) can achieve consistent performance improvement on the downstream tasks, particularly on link prediction, when trained with our framework.

Conversational agents, such as ChatGPT and Doubao, have become essential daily assistants for billions of users. To further enhance engagement, these systems are evolving from passive responders to proactive companions. However, existing efforts focus on activation within ongoing dialogues, while overlooking a key real-world bottleneck. In the conversation initiation stage, users may have a vague need but no explicit query intent, creating a first-message barrier where the conversation holds before it begins. To overcome this, we introduce Conversation Starter Generation: generating personalized starters to guide users into conversation. However, unlike in-conversation stages where immediate context guides the response, initiation must operate in a cold-start moment without explicit user intent. To pioneer in this direction, we present IceBreaker that frames human ice-breaking as a two-step handshake: (i) evoke resonance via Resonance-Aware Interest Distillation from session summaries to capture trigger interests, and (ii) stimulate interaction via Interaction-Oriented Starter Generation, optimized with personalized preference alignment and a self-reinforced loop to maximize engagement. Online A/B tests on one of the world’s largest conversational agent products show that IceBreaker improves user active days by +1.84‰ and click-through rate by +94.25‰, and has been deployed in production.

Customizing LLMs for a specific task involves separating high-quality responses from lower-quality ones. This skill can be developed using supervised fine-tuning with extensive human preference data. However, obtaining a large volume of expert-annotated data is costly for most tasks. In this paper, we explore a novel method to optimize LLMs using ranking metrics. This method trains the model to prioritize the best responses from a pool of candidates created for a particular task. Rather than a traditional full ordering, we advocate for a partial ordering, as achieving consensus on the perfect order of candidate responses can be challenging. Our partial ordering is more robust, less sensitive to noise, and can be achieved with limited human annotations or through heuristic methods. We test our system’s improved response generation ability using benchmark datasets, including textual entailment and multi-document question answering. We conduct ablation studies to understand crucial factors, such as how to gather candidate responses for a specific task, determine their most suitable order, and balance supervised fine-tuning with ranking metrics. Our approach, named RESCUE, offers a promising avenue for enhancing the response generation and task accuracy of LLMs.