Aug 2026· Proceedings of the 32nd ACM SIGKDD Conference on Knowledge Discovery and Data Mining V.2· pp. 732-742· 1 citation· 13 references
Abstract
Assessing scientific novelty is inherently complex, requiring evaluation of both a paper's internal structure and its contribution within the broader research landscape. Existing large language model (LLM) approaches often rely on surface-level similarity or citation retrieval, overlooking the integration between a paper's content and its contextual grounding in related literature. To address this gap, we introduce SciNova, a benchmark containing 3,063 papers from both ICLR and NeurIPS, with full content, bibliographies, and peer review scores for novelty prediction. Building on this benchmark, we propose GraphMind, a model that jointly processes micro- and macro-level structures for novelty prediction and rationale generation. It represents each paper as a hierarchical graph that captures its claims, methods, and experiments (micro-level), while its related paper graphs encode both cited and semantically similar works (macro-level). Experiments show that by incorporating this bi-level graph, GraphMind significantly outperforms baseline LLMs in both novelty score prediction and rationale generation.
This paper proposes Concept-guided Counterfactual Subgraph Retrieval (CCSGR), a dataset-grounded formulation that retrieves from a large graph subgraphs that are structurally and semantically similar to a query but induce different predictions under the same model, yielding domain-valid and verifiable counterfactual explanations.
Hsi-Wen Chen, Jian Pei, De-Nian Yang et al.· Proceedings of the 32nd ACM...· 0 citations
This paper introduces a novel task, graph textual summarization, which aims to generate natural language descriptions that capture both the semantic content and structural characteristics of graph data based on large language models (LLMs). Unlike traditional summarization tasks for text, images, or videos, summarizing graph data presents unique challenges due to its abstract and non-intuitive nature. We develop a comprehensive benchmark construction framework and construct three diverse benchmark datasets, including text-attributed graphs (TAGs), knowledge graphs (KGs), and heterogeneous attributed graphs (HAGs). To facilitate empirical studies, we deliver three representative benchmark datasets: SurveyGraphSum, consisting of 7835 citation graphs from survey papers across 10 major domains and 56 subdomains; BookGraphSum, containing 6175 knowledge graphs extracted from 186 books; and E-comGraphSum, with 912 heterogeneous graphs from real-world e-commerce networks, each representing a fraud group with 14 node types, 19 relation types, and 9 numerical attributes related to user behavior and interactions. In addition, we propose a unified evaluation framework that assesses both linguistic quality and factual consistency, providing a comprehensive approach to evaluating generated summaries for content fidelity and accuracy. By formalizing the task, releasing benchmark datasets, and proposing an evaluation framework, we aim to advance research in graph textual summarization and promote methods that produce accurate, informative, and human-readable summaries for complex graphs.
Xiaoxuan Gou, Weiguo Zheng, Hanqing Guo et al.· Proceedings of the 32nd ACM...· 0 citations
Scientific papers may relate by problem, method, result, or contribution, but document-level retrievers collapse these into a single similarity score without saying why they are related. Citation- and similarity-based retrieval alone also confines search to the neighbourhood of what is already known, whereas generative retrieval generates document identifiers directly, enabling the exploratory retrieval that scientific discovery depends on. We connect papers in a graph whose edges are typed by these four facets, derived from facet items and citation signals, and distil it into a generative retriever whose identifiers are the papers'own facet text. Two graph properties do not survive naive distillation. First, because every training pair is an edge, naive enumeration indexes just 84% of the corpus. Coverage-aware distillation makes every paper learnable through a reverse-neighbour fallback, a minimum-coverage threshold, and edge-importance weighting. Second, constrained decoding guarantees that every generated identifier is a valid paper, but not that the graph connects it to the query. Graph-weighted reciprocal rank fusion scales each candidate's rank term by its query-candidate edge weight, dropping unsupported ones. On LitWeave, our constructed corpus of 11,359 NLP papers, Graft recovers 91% of its graph teacher's Recall@20 with no nearest-neighbour index or encoder at inference, and outperforms the graph teacher on query papers outside the corpus. It reproduces the graph's own facet labels at 0.922 precision, so every returned paper arrives labelled with the facet that surfaced it rather than an opaque score.
Italo Luis da Silva, Hanqi Yan, Yujing Wang et al.· 0 citations
Counterfactual explanations for graph-structured data seek to determine minimal and realistic modifications required in an input graph to alter a model's prediction to a predefined output. Although counterfactual explainers that support modifying the graph by both adding and removing edges have recently emerged, there is still a lack of general and efficient methods, especially when considering the quality of the generated explanations. Moreover, the problem remains far from solved, as existing methods exhibit different strengths and weaknesses, often trading off between explanation size, coverage and quality. For this reason, it is important to identify where each method performs well and where it falls short, so as to guide future research in the field. Thus, our study compares six state-of-the-art (SOTA) models on a diverse set of real-world and synthetic datasets, covering both binary and multi-class graph and node classification tasks, and evaluates their performance using diverse quantitative and qualitative metrics.
Maria Myrto Villia, Filippos Gouidis, T. Patkos et al.· 0 citations
Surveys remain the primary way researchers grasp the lineage of methods within an AI subfield, but they scale poorly against the current rate of publication. Existing taxonomy-induction methods are largely leaf-bound and time-agnostic; they tend to force transitional papers into mature leaves and can create topological inversions between ancestors and descendants. We propose EvoTree, a staged framework that decouples conceptual backbone learning from temporal refinement: a graph-aware encoder with distribution-based hierarchical clustering yields a stable taxonomy backbone; temporal fine-tuning then re-attaches marginal papers to internal nodes under monotonic-path constraints; a final LLM pass labels concepts without altering the topology. We release the first annotated benchmark for this task across 11 AI subfields. EvoTree attains the highest NMI and citation-direction accuracy among all baselines and the best concept purity on the annotated benchmark, and is the only method with non-trivial marginal-paper detection on the annotated set.
Similarity-guided Structural Matching Learning for Graph Dataset Condensation (SSGDC) is proposed, which efficiently reduces repository size while maintaining both task performance and structural information.
Yiyang Zhang, Yutong Ye, Yingbo Zhou et al.· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.
