Publications

1. arXiv

   MOSAIC: An Omni-Modal Benchmark for Long-Term Memory and Cross-Modal Reasoning in Conversational Agents

   Hong Ting Tsang, Boyi Zhang, Haoyu Huang, Jiaxin Bai, Yufei Li, Yisen Gao, Zhongwei Xie, Hongyu Luo, Jeff Z. Pan, and Yangqiu Song

   arXiv, 2026

   Abs DOI arXiv Bib PDF Code Data Website

   Omni-modal assistants must remember not only what users say, but also what they show and convey. Existing long-term memory benchmarks are mostly text-and-image or report a single end-to-end score, so they cannot locate whether evidence is lost when it is represented, retrieved, delivered, or read. We introduce MOSAIC, a benchmark of 923 questions over persona-grounded histories of 11-24 sessions spanning text, image, video, and audio. MOSAIC constructs targeted spatial, temporal, and prosodic signals rather than decorative media, and pairs them with a matched diagnostic protocol that freezes the question, reader, prompt, and budget while intervening on representation, retrieval, and delivery. Matched decomposition shows that representation loss is modality-dependent, residual reading dominates even under gold evidence, and abstention remains weak on these controls. MOSAIC thus turns omni-modal memory evaluation from a leaderboard into a diagnosis of write, retrieve, deliver, and read failures.

   @article{mosaic,
     title = {MOSAIC: An Omni-Modal Benchmark for Long-Term Memory and Cross-Modal Reasoning in Conversational Agents},
     author = {Tsang, Hong Ting and Zhang, Boyi and Huang, Haoyu and Bai, Jiaxin and Li, Yufei and Gao, Yisen and Xie, Zhongwei and Luo, Hongyu and Pan, Jeff Z. and Song, Yangqiu},
     year = {2026},
     journal = {arXiv},
     doi = {10.48550/arXiv.2610.29249},
   }

2. ACL

   AutoGraph-R1: End-to-End Reinforcement Learning for Knowledge Graph Construction

   Hong Ting Tsang, Jiaxin Bai, Haoyu Huang, Qiao Xiao, Tianshi Zheng, Baixuan Xu, Shujie Liu, and Yangqiu Song

   In ACL 2026 (Main), 2026

   Abs DOI arXiv Bib PDF Code

   Building effective knowledge graphs (KGs) for Retrieval-Augmented Generation (RAG) is pivotal for advancing question answering (QA) systems. However, its effectiveness is hindered by a fundamental disconnect: the knowledge graph (KG) construction process is decoupled from its downstream application, yielding suboptimal graph structures. To bridge this gap, we introduce AutoGraph-R1, the first framework to directly optimize KG construction for task performance using Reinforcement Learning (RL). AutoGraph-R1 trains an LLM constructor by framing graph generation as a policy learning problem, where the reward is derived from the graph’s functional utility in a RAG pipeline. We design two novel, task-aware reward functions, one for graphs as knowledge carriers and another as knowledge indices. Across multiple QA benchmarks, AutoGraph-R1 consistently enables graph RAG methods to achieve significant performance gains over using task-agnostic baseline graphs. Our work shows it is possible to close the loop between construction and application, shifting the paradigm from building intrinsically ’good’ graphs to building demonstrably ’useful’ ones.

   @inproceedings{autograph-r1,
     title = {AutoGraph-R1: End-to-End Reinforcement Learning for Knowledge Graph Construction},
     author = {Tsang, Hong Ting and Bai, Jiaxin and Huang, Haoyu and Xiao, Qiao and Zheng, Tianshi and Xu, Baixuan and Liu, Shujie and Song, Yangqiu},
     year = {2026},
     booktitle = {ACL 2026 (Main)},
     doi = {10.48550/arXiv.2510.15339},
   }

3. ACL

   AutoSchemaKG: Autonomous Knowledge Graph Construction through Dynamic Schema Induction from Web-Scale Corpora

   Jiaxin Bai, Wei Fan, Qi Hu, Qing Zong, Chunyang Li, Hong Ting Tsang, Hongyu Luo, Yauwai Yim, Haoyu Huang, Xiao Zhou, Feng Qin, Tianshi Zheng, Xi Peng, Xin Yao, Huiwen Yang, Leijie Wu, Yi Ji, Gong Zhang, Renhai Chen, and Yangqiu Song

   In ACL 2026 (Main, Oral)— codebase maintainer and core contributor , 2026

   Abs DOI arXiv Bib PDF Code

   We present AutoSchemaKG, a framework for fully autonomous knowledge graph construction that eliminates the need for predefined schemas. Our system leverages large language models to simultaneously extract knowledge triples and induce comprehensive schemas directly from text, modeling both entities and events while employing conceptualization to organize instances into semantic categories. Processing over 50 million documents, we construct ATLAS (Automated Triple Linking And Schema induction), a family of knowledge graphs with 900+ million nodes and 5.9 billion edges. This approach outperforms state-of-the-art baselines on multi-hop QA tasks and enhances LLM factuality. Notably, our schema induction achieves 92% semantic alignment with human-crafted schemas with zero manual intervention, demonstrating that billion-scale knowledge graphs with dynamically induced schemas can effectively complement parametric knowledge in large language models.

   @inproceedings{autoschema,
     title = {AutoSchemaKG: Autonomous Knowledge Graph Construction through Dynamic Schema Induction from Web-Scale Corpora},
     author = {Bai, Jiaxin and Fan, Wei and Hu, Qi and Zong, Qing and Li, Chunyang and Tsang, Hong Ting and Luo, Hongyu and Yim, Yauwai and Huang, Haoyu and Zhou, Xiao and Qin, Feng and Zheng, Tianshi and Peng, Xi and Yao, Xin and Yang, Huiwen and Wu, Leijie and Ji, Yi and Zhang, Gong and Chen, Renhai and Song, Yangqiu},
     year = {2026},
     booktitle = {ACL 2026 (Main, Oral)},
     doi = {10.48550/arXiv.2505.23628},
   }

2026

1. ICLR

   AtlasKV: Augmenting LLMs with Billion-Scale Knowledge Graphs in 20GB VRAM

   Haoyu Huang, Hong Ting Tsang, Jiaxin Bai, Xi Peng, Gong Zhang, and Yangqiu Song

   In ICLR 2026, 2026

   Abs Bib HTML PDF

   Retrieval-augmented generation (RAG) has shown some success in augmenting large language models (LLMs) with external knowledge. However, as a non-parametric knowledge integration paradigm for LLMs, RAG methods heavily rely on external retrieval modules and the retrieved textual context prior. Especially for very large scale knowledge augmentation, they would introduce substantial inference latency due to expensive searches and much longer relevant context. In this paper, we propose a parametric knowledge integration method, called AtlasKV, a scalable, effective, and general way to augment LLMs with billion-scale knowledge graphs (KGs) (e.g. 1B triples) using very little GPU memory cost (e.g. less than 20GB VRAM). In AtlasKV, we introduce KG2KV and HiKVP to integrate KG triples into LLMs at scale with sub-linear time and memory complexity. It maintains strong knowledge grounding and generalization performance using the LLMs’ inherent attention mechanism, and requires no external retrievers, long context priors, or retraining when adapting to new knowledge.

   @inproceedings{atlaskv,
     title = {AtlasKV: Augmenting LLMs with Billion-Scale Knowledge Graphs in 20GB VRAM},
     author = {Huang, Haoyu and Tsang, Hong Ting and Bai, Jiaxin and Peng, Xi and Zhang, Gong and Song, Yangqiu},
     year = {2026},
     booktitle = {ICLR 2026},
   }

2. EMNLP

   The Cognitive Bandwidth Bottleneck: Shifting Long-Horizon Agents from Planning with Actions to Planning with Schemas

   Baixuan Xu, Tianshi Zheng, Zhaowei Wang, Hong Ting Tsang, Weiqi Wang, Tianqing Fang, and Yangqiu Song

   In EMNLP 2026, 2026

   Abs DOI arXiv Bib HTML PDF

   Enabling LLMs to effectively operate long-horizon task which requires long-term planning and multiple interactions is essential for open-world autonomy. Conventional methods adopt planning with actions where a executable action list would be provided as reference. However, this action representation choice would be impractical when the environment action space is combinatorial exploded (e.g., open-ended real world). This naturally leads to a question: As environmental action space scales, what is the optimal action representation for long-horizon agents? In this paper, we systematically study the effectiveness of two different action representations. The first one is conventional planning with actions (PwA) which is predominantly adopted for its effectiveness on existing benchmarks. The other one is planning with schemas (PwS) which instantiate an action schema into action lists (e.g., "move [OBJ] to [OBJ]" -> "move apple to desk") to ensure concise action space and reliable scalability. This alternative is motivated by its alignment with human cognition and its compliance with environment-imposed action format restriction. We propose cognitive bandwidth perspective as a conceptual framework to qualitatively understand the differences between these two action representations and empirically observe a representation-choice inflection point between ALFWorld (~35 actions) and SciWorld (~500 actions), which serve as evidence of the need for scalable representations. We further conduct controlled experiments to study how the location of this inflection point interacts with different model capacities: stronger planning proficiency shifts the inflection rightward, whereas better schema instantiation shifts it leftward. Finally, noting the suboptimal performance of PwS agents, we provide an actionable guide for building more capable PwS agents for better scalable autonomy.

   @inproceedings{cognitivebottleneck,
     title = {The Cognitive Bandwidth Bottleneck: Shifting Long-Horizon Agents from Planning with Actions to Planning with Schemas},
     author = {Xu, Baixuan and Zheng, Tianshi and Wang, Zhaowei and Tsang, Hong Ting and Wang, Weiqi and Fang, Tianqing and Song, Yangqiu},
     year = {2026},
     booktitle = {EMNLP 2026},
     doi = {10.18653/2026.emnlp-main.755}
   }

3. ACL

   ContextLens: Modeling Imperfect Privacy and Safety Context for Legal Compliance

   Haoran Li, Yulin Chen, Huihao Jing, Wenbin Hu, Tsz Ho Li, Chanhou Lou, Hong Ting Tsang, Sirui Han, and Yangqiu Song

   In ACL 2026 (Main), 2026

   Abs DOI arXiv Bib PDF

   Individuals’ concerns about data privacy and AI safety are highly contextualized and extend beyond sensitive patterns. Addressing these issues requires reasoning about the context to identify and mitigate potential risks. We propose ContextLens, a semi-rule-based framework that leverages LLMs to ground the input context in the legal domain and explicitly identify both known and unknown factors for legal compliance. Experiments on GDPR and EU AI Act compliance benchmarks show that ContextLens significantly improves LLMs’ compliance assessment and can identify ambiguous and missing factors.

   @inproceedings{contextlens,
     title = {ContextLens: Modeling Imperfect Privacy and Safety Context for Legal Compliance},
     author = {Li, Haoran and Chen, Yulin and Jing, Huihao and Hu, Wenbin and Li, Tsz Ho and Lou, Chanhou and Tsang, Hong Ting and Han, Sirui and Song, Yangqiu},
     year = {2026},
     booktitle = {ACL 2026 (Main)},
     doi = {10.48550/arXiv.2604.12308}
   }

4. arXiv

   DeepRefine: Agentic Knowledge Refinement via Reinforcement Learning

   Haoyu Huang, Jiaxin Bai, Shujie Liu, Yang Wei, Huihao Jing, Hong Ting Tsang, Yisen Gao, Zhongwei Xie, Yufei Li, and Yangqiu Song

   arXiv, 2026

   Abs DOI arXiv Bib PDF

   External knowledge enables large language model (LLM) agents to ground their actions and decisions beyond intrinsic parametric memory in open-ended, knowledge-intensive downstream tasks. Yet the quality of the underlying knowledge bases is systematically limited by incompleteness, incorrectness, or redundancy, manifested as missing evidence or cross-document links, low-confidence or imprecise claims, and ambiguous or coreference resolution issues. Such defects compound under iterative use, degrading retrieval fidelity and downstream task performance. We present DeepRefine, a reinforcement learning framework for agentic knowledge refinement that evolves the quality of any pre-constructed structured knowledge bases, e.g., knowledge graphs or LLM-Wikis, with user queries to make it more suitable for the downstream tasks. DeepRefine performs multi-turn interactions with the knowledge base and conducts abductive diagnosis over the interaction history, localizes likely defects, and executes targeted refinement actions for incremental knowledge base updates. To further optimize refinement policies of DeepRefine without golden refinement trajectories, we introduce a Gain-Beyond-Draft (GBD) reward and train the reasoning process end-to-end via reinforcement learning. Extensive experiments demonstrate consistent downstream gains over strong baselines.

   @article{deeprefine,
     title = {DeepRefine: Agentic Knowledge Refinement via Reinforcement Learning},
     author = {Huang, Haoyu and Bai, Jiaxin and Liu, Shujie and Wei, Yang and Jing, Huihao and Tsang, Hong Ting and Gao, Yisen and Xie, Zhongwei and Li, Yufei and Song, Yangqiu},
     year = {2026},
     journal = {arXiv},
     doi = {10.48550/arXiv.2605.10488}
   }

5. arXiv

   Can Language Models Imagine Without Seeing? Ekphrasis: Measuring Visual Creative Ideation in Text-Only LLMs

   Hongyu Luo, He Wang, Huihao Jing, Hong Ting Tsang, Yuxuan Liu, Wuganjing Song, Yauwai Yim, Chunyang Li, and Yangqiu Song

   arXiv, 2026

   Abs DOI arXiv Bib PDF

   Current evaluations do not isolate whether text-only language models can originate visual concepts before image generation. Fluent visual prose can hide visual-plan failures: an answer may appear creative while repeating familiar visual clichés or failing to specify a renderable scene. We define Visual Creative Ideation (VCI) as the ability to produce textual visual plans that are useful, expressive, and population-novel, and introduce Ekphrasis, a 400-task benchmark spanning Abstraction, Combination, Transformation, and Adaptation. Ekphrasis scores anonymized pairwise comparisons with dimension-specific checklists, aggregates preferences with Bradley-Terry models, and uses Typed Idea Graphs to convert task-specific population clichés into novelty references. Across 14 language models, VCI separates usefulness, expressiveness, and novelty rather than reducing to fluency: strong models achieve similar overall scores through different profiles, and useful plans can remain visually clichéd. A cross-modal grounding study further shows that text-level VCI ordering largely survives faithful rendering and blind image-level preference judgment, supporting Ekphrasis as a measure of visual ideation beyond prose quality.

   @article{ekphrasis,
     title = {Can Language Models Imagine Without Seeing? Ekphrasis: Measuring Visual Creative Ideation in Text-Only LLMs},
     author = {Luo, Hongyu and Wang, He and Jing, Huihao and Tsang, Hong Ting and Liu, Yuxuan and Song, Wuganjing and Yim, Yauwai and Li, Chunyang and Song, Yangqiu},
     year = {2026},
     journal = {arXiv},
     doi = {10.48550/arXiv.2608.06967}
   }

6. arXiv

   Towards In-Parameter Memory Augmentation for Large Language Models

   Haoyu Huang, Zhongwei Xie, Jiaxin Bai, Yisen Gao, Hong Ting Tsang, Wuganjing Song, Huihao Jing, Yufei Li, and Yangqiu Song

   arXiv, 2026

   Abs DOI arXiv Bib PDF

   Recently Large Language Models (LLMs) and LLM-based agents increasingly need to incorporate knowledge acquired after pretraining, e.g., domain facts, user preferences, documents, and interaction experience. In-context learning (ICL) and ICL-based agent harness remain flexible, but they consume context capacity and incur repeated discretized encoding cost that grows with context length. In-parameter memory offers a complementary substrate: reusable memory information is represented in model parameters, adapters, or other parameter-like objects that are composed into the forward pass at inference time. This survey focuses on methods that augment LLMs with such parametric memory at deployment: a memory-bearing parameter object is plugged into the forward pass during inference, whether it is acquired before or during deployment. We organize the landscape with two orthogonal axes: Parameter Placement, which includes Embedding, Attention, FFN layers, or Hybrid when two or more layers are used; and Parameter Acquisition Time, which distinguishes methods whose memory object is acquired during deployment (online) from those acquired before it (offline). We clarify boundaries, conduct comparisons, and discuss open directions in interference, safety, co-design with ICL, and recursive self-improvement.

   @article{inparam-memory,
     title = {Towards In-Parameter Memory Augmentation for Large Language Models},
     author = {Huang, Haoyu and Xie, Zhongwei and Bai, Jiaxin and Gao, Yisen and Tsang, Hong Ting and Song, Wuganjing and Jing, Huihao and Li, Yufei and Song, Yangqiu},
     year = {2026},
     journal = {arXiv},
     doi = {10.48550/arXiv.2610.08630}
   }

7. arXiv

   KGPFN: Unlocking the Potential of Knowledge Graph Foundation Model via In-Context Learning

   Yisen Gao, Jiaxin Bai, Haoyu Huang, Zhongwei Xie, Yufei Li, Hong Ting Tsang, Sirui Han, and Yangqiu Song

   arXiv, 2026

   Abs DOI arXiv Bib PDF

   Knowledge graph (KG) foundation models aim to generalize to graphs with unseen entities and relations by learning transferable relational structure. Most existing methods, however, focus on relation-level universality, leaving in-context learning, the other pillar of foundation models, largely unexplored for KG reasoning. Context in KGs is structured and heterogeneous: accurate prediction requires conditioning both on the local neighborhood of the query entities and on global context that summarizes how the query relation behaves across many instances. We propose KGPFN, a KG foundation model built on a Prior-Data Fitted Network (PFN) that combines transferable relational representations with inference-time in-context learning over structured context. KGPFN learns relation representations by message passing on relation graphs and extracts multi-scale local context from the intermediate head representations of a multi-layer NBFNet. It then builds relation-specific global context from positive and negative examples of the query relation, together with their local structural representations, and aggregates this context with feature-level and sample-level attention. Through multi-graph pretraining, KGPFN learns to combine structural representations with labeled contextual evidence without inference-time parameter updates. On 57 knowledge graphs, KGPFN achieves the best average MRR both without and with fine-tuning, and context sensitivity analyses highlight the value of negative context examples. Our code is available at this https URL.

   @article{kgpfn,
     title = {KGPFN: Unlocking the Potential of Knowledge Graph Foundation Model via In-Context Learning},
     author = {Gao, Yisen and Bai, Jiaxin and Huang, Haoyu and Xie, Zhongwei and Li, Yufei and Tsang, Hong Ting and Han, Sirui and Song, Yangqiu},
     year = {2026},
     journal = {arXiv},
     doi = {10.48550/arXiv.2605.14907}
   }

8. ICLR

   NewtonBench: Benchmarking Generalizable Scientific Law Discovery in LLM Agents

   Tianshi Zheng, Kelvin Kiu-Wai Tam, Newt Hue-Nam K. Nguyen, Baixuan Xu, Zhaowei Wang, Jiayang Cheng, Hong Ting Tsang, Weiqi Wang, Jiaxin Bai, Tianqing Fang, Yangqiu Song, Ginny Y. Wong, and Simon See

   In ICLR 2026, 2026

   Abs DOI arXiv Bib PDF Code

   Large language models are emerging as powerful tools for scientific law discovery, a foundational challenge in AI-driven science. However, existing benchmarks for this task suffer from a fundamental methodological trilemma, forcing a trade-off between scientific relevance, scalability, and resistance to memorization. Furthermore, they oversimplify discovery as static function fitting, failing to capture the authentic scientific process of uncovering embedded laws through the interactive exploration of complex model systems. To address these critical gaps, we introduce NewtonBench, a benchmark comprising 324 scientific law discovery tasks across 12 physics domains. Our design mitigates the evaluation trilemma by using metaphysical shifts - systematic alterations of canonical laws - to generate a vast suite of problems that are scalable, scientifically relevant, and memorization-resistant. Moreover, we elevate the evaluation from static function fitting to interactive model discovery, requiring agents to experimentally probe simulated complex systems to uncover hidden principles. Our extensive experiment reveals a clear but fragile capability for discovery in frontier LLMs: this ability degrades precipitously with increasing system complexity and exhibits extreme sensitivity to observational noise. Notably, we uncover a paradoxical effect of tool assistance: providing a code interpreter can hinder more capable models by inducing a premature shift from exploration to exploitation, causing them to satisfice on suboptimal solutions. These results demonstrate that robust, generalizable discovery in complex, interactive environments remains the core challenge. By providing a scalable, robust, and scientifically authentic testbed, NewtonBench offers a crucial tool for measuring true progress and guiding the development of next-generation AI agents capable of genuine scientific discovery.

   @inproceedings{newtonbench,
     title = {NewtonBench: Benchmarking Generalizable Scientific Law Discovery in LLM Agents},
     author = {Zheng, Tianshi and Tam, Kelvin Kiu-Wai and Nguyen, Newt Hue-Nam K. and Xu, Baixuan and Wang, Zhaowei and Cheng, Jiayang and Tsang, Hong Ting and Wang, Weiqi and Bai, Jiaxin and Fang, Tianqing and Song, Yangqiu and Wong, Ginny Y. and See, Simon},
     year = {2026},
     booktitle = {ICLR 2026},
     doi = {10.48550/arXiv.2510.07172},
   }

9. NeurIPS

   NGDB-Zoo: Towards Efficient and Scalable Neural Graph Databases Training

   Zhongwei Xie, Jiaxin Bai, Shujie Liu, Haoyu Huang, Yufei Li, Yisen Gao, Hong Ting Tsang, and Yangqiu Song

   In NeurIPS 2026, 2026

   Abs DOI arXiv Bib PDF Code

   NGDB-Zoo is a framework for efficient and scalable training of neural graph databases, addressing key challenges in performance and scalability.

   @inproceedings{ngdb_zoo,
     title = {NGDB-Zoo: Towards Efficient and Scalable Neural Graph Databases Training},
     author = {Xie, Zhongwei and Bai, Jiaxin and Liu, Shujie and Huang, Haoyu and Li, Yufei and Gao, Yisen and Tsang, Hong Ting and Song, Yangqiu},
     year = {2026},
     booktitle = {NeurIPS 2026},
     doi = {10.48550/arXiv.2602.21597},
   }

10. arXiv

    PatchWorld: Gradient-Free Optimization of Executable World Models for Agent Environments

    Jiaxin Bai, Yue Guo, Yifei Dong, Jiaxuan Xiong, Tianshi Zheng, Yixia Li, Tianqing Fang, Yufei Li, Yisen Gao, Haoyu Huang, Zhongwei Xie, Hong Ting Tsang, Zihao Wang, Lihui Liu, Jeff Z. Pan, and Yangqiu Song

    arXiv, 2026

    Abs DOI arXiv Bib PDF

    World models for interactive text agents must typically be learned from observation-action trajectories alone. Specifically, the environment returns text observations after each action, but does not expose a ground-truth latent state nor an inspectable transition model.A research gap remains in how to induce executable code as a world model in this black-box setting for prediction and agent decision making. We introduce PatchWorld, a gradient-free framework that turns offline trajectories into executable Python world models through counterexample-guided code this http URL of predicting the next observation with a black-box model, PatchWorld induces symbolic belief-state programs whose action updates can be inspected, replayed, and locally patched. Across seven AgentGym environments, PatchWorld-Simple achieves the highest code-based decision-making score among evaluated methods (76.4% macro success in live one-step lookahead), matching or exceeding LLM-based lookahead while invoking no LLM calls inside the world-model prediction module itself. We further find that a human-specified residual-memory bias improves surface observation fidelity but weakens agent decision-making utility. This reveals a tradeoff in executable world models, since improving observation fidelity can come at the expense of action-discriminative dynamics, and vice versa. Code is available at this https URL.

    @article{patchworld,
      title = {PatchWorld: Gradient-Free Optimization of Executable World Models for Agent Environments},
      author = {Bai, Jiaxin and Guo, Yue and Dong, Yifei and Xiong, Jiaxuan and Zheng, Tianshi and Li, Yixia and Fang, Tianqing and Li, Yufei and Gao, Yisen and Huang, Haoyu and Xie, Zhongwei and Tsang, Hong Ting and Wang, Zihao and Liu, Lihui and Pan, Jeff Z. and Song, Yangqiu},
      year = {2026},
      journal = {arXiv},
      doi = {10.48550/arXiv.2605.30880}
    }

11. EMNLP

    NGDBench: Towards Neural Graph Data Management

    Yufei Li, Yisen Gao, Jiaxuan Xiong, Jiaxin Bai, Shijie Zhong, Haoyu Huang, Zhongwei Xie, Hong Ting Tsang, and Yangqiu Song

    In Findings of the Association for Computational Linguistics: EMNLP 2026, 2026

    Abs DOI arXiv Bib HTML PDF Code

    Data critical to real-world decision-making is increasingly found within organizations. Such data is heterogeneous, constantly evolving, and only imperfectly captured. However, current data management systems remain largely passive, retrieving what is explicitly stored while offering limited support for uncovering implicit structure or reasoning under noise, incompleteness, and continuous updates. We argue that next-generation data management requires neural capabilities, which can uncover complex latent relationships, distinguish reliable signals from noise, and remain consistent as the underlying data state evolves. To support this direction, we introduce NGDBench, a benchmark across five domains that unifies structured and unstructured sources. NGDBench adopts a graph view because graphs provide a flexible abstraction for modeling complex systems, capturing latent relationships, and subsuming structured formats such as relational tables. Each instance pairs a clean latent graph with a realistically perturbed observed graph. NGDBench supports full Cypher queries and dynamic data management operations. Evaluations of state-of-the-art Text-to-Cypher by LLMs and GraphRAG pipelines reveal that current neural query methods remain sensitive to noise and struggle with dynamic state tracking, highlighting the need for resilient, inference-capable data management.

    @inproceedings{neural_graph_data_management,
      title = {NGDBench: Towards Neural Graph Data Management},
      author = {Li, Yufei and Gao, Yisen and Xiong, Jiaxuan and Bai, Jiaxin and Zhong, Shijie and Huang, Haoyu and Xie, Zhongwei and Tsang, Hong Ting and Song, Yangqiu},
      year = {2026},
      booktitle = {Findings of the Association for Computational Linguistics: EMNLP 2026},
      doi = {10.18653/2026.findings-emnlp.1546},
    }

2025

1. arXiv

   Transformers for Complex Query Answering over Knowledge Hypergraphs

   Hong Ting Tsang, Zihao Wang, and Yangqiu Song

   arXiv, 2025

   Abs DOI arXiv Bib PDF

   Complex Query Answering (CQA) has been extensively studied in recent years. In order to model data that is closer to real-world distribution, knowledge graphs with different modalities have been introduced. Triple KGs, as the classic KGs composed of entities and relations of arity 2, have limited representation of real-world facts. Real-world data is more sophisticated. While hyper-relational graphs have been introduced, there are limitations in representing relationships of varying arity that contain entities with equal contributions. To address this gap, we sampled new CQA datasets: JF17k-HCQA and M-FB15k-HCQA. Each dataset contains various query types that include logical operations such as projection, negation, conjunction, and disjunction. In order to answer knowledge hypergraph (KHG) existential first-order queries, we propose a two-stage transformer model, the Logical Knowledge Hypergraph Transformer (LKHGT), which consists of a Projection Encoder for atomic projection and a Logical Encoder for complex logical operations. Both encoders are equipped with Type Aware Bias (TAB) for capturing token interactions. Experimental results on CQA datasets show that LKHGT is a state-of-the-art CQA method over KHG and is able to generalize to out-of-distribution query types.

   @article{cqa_knowledge_hyper,
     title = {Transformers for Complex Query Answering over Knowledge Hypergraphs},
     author = {Tsang, Hong Ting and Wang, Zihao and Song, Yangqiu},
     year = {2025},
     journal = {arXiv},
     doi = {10.48550/arXiv.2504.16537}
   }

2. EMNLP

   From Automation to Autonomy: A Survey on Large Language Models in Scientific Discovery.

   Tianshi Zheng, Zheye Deng, Hong Ting Tsang, Weiqi Wang, Jiaxin Bai, Zihao Wang, and Yangqiu Song

   In EMNLP, 2025

   Abs DOI arXiv Bib PDF Code

   Large Language Models (LLMs) are catalyzing a paradigm shift in scientific discovery, evolving from task-specific automation tools into increasingly autonomous agents and fundamentally redefining research processes and human-AI collaboration. This survey systematically charts this burgeoning field, placing a central focus on the changing roles and escalating capabilities of LLMs in science. Through the lens of the scientific method, we introduce a foundational three-level taxonomy-Tool, Analyst, and Scientist-to delineate their escalating autonomy and evolving responsibilities within the research lifecycle. We further identify pivotal challenges and future research trajectories such as robotic automation, self-improvement, and ethical governance. Overall, this survey provides a conceptual architecture and strategic foresight to navigate and shape the future of AI-driven scientific discovery, fostering both rapid innovation and responsible advancement. Github Repository: https://github.com/HKUST-KnowComp/Awesome-LLM-Scientific-Discovery.

   @inproceedings{from_automation_to_autonomy,
     title = {From Automation to Autonomy: A Survey on Large Language Models in Scientific Discovery.},
     author = {Zheng, Tianshi and Deng, Zheye and Tsang, Hong Ting and Wang, Weiqi and Bai, Jiaxin and Wang, Zihao and Song, Yangqiu},
     year = {2025},
     booktitle = {EMNLP},
     doi = {10.48550/arXiv.2505.13259},
   }

3. ICMLC

   TERAG: Token-Efficient Graph-Based Retrieval-Augmented Generation

   Qiao Xiao, Hong Ting Tsang, and Jiaxin Bai

   In ICMLC, 2025

   Abs DOI arXiv Bib PDF

   Graph-based Retrieval-augmented generation (RAG) has become a widely studied approach for improving the reasoning, accuracy, and factuality of Large Language Models. However, many existing graph-based RAG systems overlook the high cost associated with LLM token usage during graph construction, hindering large-scale adoption. To address this, we propose TERAG, a simple yet effective framework designed to build informative graphs at a significantly lower cost. Inspired by HippoRAG, we incorporate Personalized PageRank (PPR) during the retrieval phase, and we achieve at least 80% of the accuracy of widely used graph-based RAG methods while consuming only 3%-11% of the output tokens.

   @inproceedings{terag,
     title = {TERAG: Token-Efficient Graph-Based Retrieval-Augmented Generation},
     author = {Xiao, Qiao and Tsang, Hong Ting and Bai, Jiaxin},
     year = {2025},
     booktitle = {ICMLC},
     doi = {10.48550/arXiv.2509.18667}
   }

4. DEB

   Top Ten Challenges Towards Agentic Neural Graph Databases

   Jiaxin Bai, Zihao Wang, Yukun Zhou, Hang Yin, Weizhi Fei, Qi Hu, Zheye Deng, Jiayang Cheng, Tianshi Zheng, Hong Ting Tsang, Yisen Gao, Zhongwei Xie, Yufei Li, Lixin Fan, Binhang Yuan, Wei Wang, Lei Chen, Xiaofang Zhou, and Yangqiu Song

   IEEE Data Engineering Bulletin, 2025

   Abs DOI arXiv Bib PDF

   Graph databases (GDBs) like Neo4j and TigerGraph excel at handling interconnected data but lack advanced inference capabilities. Neural Graph Databases (NGDBs) address this by integrating Graph Neural Networks (GNNs) for predictive analysis and reasoning over incomplete or noisy data. However, NGDBs rely on predefined queries and lack autonomy and adaptability. This paper introduces Agentic Neural Graph Databases (Agentic NGDBs), which extend NGDBs with three core functionalities: autonomous query construction, neural query execution, and continuous learning. We identify ten key challenges in realizing Agentic NGDBs: semantic unit representation, abductive reasoning, scalable query execution, and integration with foundation models like large language models (LLMs). By addressing these challenges, Agentic NGDBs can enable intelligent, self-improving systems for modern data-driven applications, paving the way for adaptable and autonomous data management solutions.

   @article{top10_challenge_ngdb,
     title = {Top Ten Challenges Towards Agentic Neural Graph Databases},
     author = {Bai, Jiaxin and Wang, Zihao and Zhou, Yukun and Yin, Hang and Fei, Weizhi and Hu, Qi and Deng, Zheye and Cheng, Jiayang and Zheng, Tianshi and Tsang, Hong Ting and Gao, Yisen and Xie, Zhongwei and Li, Yufei and Fan, Lixin and Yuan, Binhang and Wang, Wei and Chen, Lei and Zhou, Xiaofang and Song, Yangqiu},
     year = {2025},
     journal = {IEEE Data Engineering Bulletin},
     doi = {10.48550/arXiv.2501.14224}
   }