{"slug": "towards-a-learnable-retriever-for-textual-graph-question-answering", "title": "Towards a Learnable Retriever for Textual Graph Question Answering", "summary": "Researchers posting arXiv paper 2610.33511v1 propose TaLenT, an end-to-end framework for textual graph question answering that pairs a learnable LLM-based retriever with a fine-tuned LLM-based generator in a single retrieve-then-generate pipeline. TaLenT supervises the retriever with gold answer signals, stratifying graph nodes into gold answers, answer-related nodes, and irrelevant noise and applying a weighted pseudo-label loss to prioritize gold answers and preserve answer-related nodes while discarding noise. Experiments across three GraphQA datasets show TaLenT consistently improves downstream performance.", "body_md": "arXiv:2610.33511v1 Announce Type: new \nAbstract: Textual graph question answering (GraphQA) and graph-based retrieval-augmented generation (GraphRAG) have gained increasing attention as a way to ground large language models (LLMs) on structured knowledge. Despite extensive efforts, existing methods still face two key gaps. The first is an architectural gap: existing methods typically fine-tune an LLM-based generator, but do not also make the retriever itself a learnable LLM for graph evidence selection within one end-to-end retrieve-then-generate pipeline. The second is a retrieval-supervision gap: existing retrievers rarely use gold answer signals to supervise evidence selection. As a result, they may fail to recall evidence nodes required for reasoning, resulting in insufficient coverage of the gold answer set; meanwhile, naive attempts to boost recall often introduce excessive irrelevant information that compromises compactness and overwhelms the LLM. To address these gaps, we propose TaLenT, an LLM-based end-to-end retriever-generator framework for textual GraphQA. To bridge the architectural gap, TaLenT pairs a learnable LLM-based retriever for evidence selection with a fine-tuned LLM-based generator within one unified retrieve-then-generate pipeline. To bridge the retrieval-supervision gap, we leverage gold answer signals as retriever supervision. Specifically, we stratify graph nodes into gold answers, answer-related nodes, and irrelevant noise. We then adopt a weighted pseudo-label loss to prioritize gold answers and preserve answer-related nodes for coverage, while discarding irrelevant noise for compactness. Experimental results across three GraphQA datasets demonstrate that TaLenT consistently improves downstream performance.", "url": "https://wpnews.pro/news/towards-a-learnable-retriever-for-textual-graph-question-answering", "canonical_source": "https://www.machinebrief.com/news/towards-a-learnable-retriever-for-textual-graph-question-ans-rojc", "published_at": "2026-10-27 04:00:00+00:00", "updated_at": "2026-10-27 04:42:12.866036+00:00", "lang": "en", "topics": ["artificial-intelligence", "large-language-models", "ai-research", "natural-language-processing"], "entities": ["TaLenT", "arXiv", "GraphQA", "GraphRAG"], "also_reported_by": [], "alternates": {"html": "https://wpnews.pro/news/towards-a-learnable-retriever-for-textual-graph-question-answering", "markdown": "https://wpnews.pro/news/towards-a-learnable-retriever-for-textual-graph-question-answering.md", "text": "https://wpnews.pro/news/towards-a-learnable-retriever-for-textual-graph-question-answering.txt", "jsonld": "https://wpnews.pro/news/towards-a-learnable-retriever-for-textual-graph-question-answering.jsonld"}}