Tag
This paper defines a hierarchy of faithfulness criteria for knowledge base completion models and evaluates current embedding models, finding they are not logically faithful.
This paper evaluates small language models beyond answer accuracy in knowledge graph question answering by isolating graph navigation capabilities, showing significant differences in path fidelity and the need for broader evaluation metrics.
This paper investigates graph-guided evidence navigation to improve long-novel reasoning with a local 9B language model, providing exploratory evidence from experiments on detective novels without statistically significant gains over baselines.
SCoP introduces a constraint-centric framework for temporal knowledge graph question answering that externalizes temporal decisions to filter evidence and improve answer accuracy on benchmark datasets like MultiTQ and TimelineCronQ-R.
The paper presents KG-Chronos-2, a method that enhances a frozen time-series foundation model with knowledge graph augmentation for improved surrogate forecasting of water-surface elevation in HEC-RAS, demonstrating significant RMSE reduction compared to baseline approaches.
This paper introduces G-RELIC, a graph-based approach that combines LLMs with graphs to improve consistency in relexicalization of clinical documents, enhancing relational integrity and temporal coherence while maintaining privacy benchmarks.
This paper proposes VisKG-LM, a method that compiles knowledge graphs into visual memory for efficient multiple-choice question answering, achieving performance gains over baselines by decoupling graph encoding from language reasoning.
This paper benchmarks Knowledge Graph-based augmentation against Retrieval Augmented Generation for culturally specific question answering, demonstrating significant error reduction on Latin American culture data using the LatamQA dataset.
CoG is a cognitive-inspired, training-free framework for adaptive knowledge exploration in retrieval-augmented generation, achieving state-of-the-art performance on multi-hop QA benchmarks through plan-explore-reflect cycles and bidirectional graph-text synergy.
GraphProfiler introduces an auditable LLM-based profiler that uses source-linked personal knowledge graphs to infer sensitive attributes from user-generated content, providing evidence citations for predictions.
The article explains why knowledge graphs are superior to flat lists for memory storage in AI systems, as they effectively handle multiple references to the same entity and prevent fragmented or contradictory search results.
The paper proposes reifying input graphs into a fixed vocabulary, allowing vanilla GNNs to achieve zero-shot link prediction across unseen knowledge graphs and relational databases, matching dedicated foundation models like ULTRA.
The paper proposes Semantic Bayesian World Models (SBWMs) to bridge the gap between knowledge graphs' crisp assertions and foundation models' probabilistic reasoning, enabling a shared, evolving web of beliefs for unified reasoning under uncertainty.
HyGRAIL is a cost-aware framework that combines GNN triage with LLM review for scientific hypothesis discovery on knowledge graphs, achieving improved F1 scores and reduced LLM calls on the MatKG benchmark.
The article discusses challenges and asks for community experiences regarding the breakdown of AI agent memory systems after months in production use.
The paper introduces AVA, a framework to evaluate NLP embeddings' ability to capture ontological reasoning, finding significant limitations and challenging the assumption that strong NLP performance translates to Semantic Web competence.
This paper evaluates a graph-based memory framework for long-term LLM agents, finding that it does not outperform flat vector retrieval on recall metrics, but a selective forgetting module effectively reduces storage with minimal performance loss.
This paper proposes LitEm, a neural regression model that enables transductive knowledge graph embedding models to predict numerical attributes, achieving strong benchmark results and introducing a co-training framework for improved performance.
The article debates whether AI models need dedicated long-term memory systems like RAG or if scaling context windows is sufficient, presenting arguments for both approaches and seeking community input.
HyperExtract is an LLM-powered framework that converts unstructured documents into structured knowledge graphs, hypergraphs, and lists, simplifying knowledge extraction and management.