{"paper":{"title":"Why Neighborhoods Matter: Traversal Context and Provenance in Agentic GraphRAG","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"In Agentic GraphRAG, accurate answers depend on both cited evidence and the uncited traversal context from the agent's graph exploration.","cross_cats":["cs.IR"],"primary_cat":"cs.AI","authors_text":"Maximilian von Zastrow, Riccardo Terrenzi, Serkan Ayvaz","submitted_at":"2026-05-14T17:25:20Z","abstract_excerpt":"Retrieval-Augmented Generation can improve factuality by grounding answers in external evidence, but Agentic GraphRAG complicates what it means for citations to be faithful. In these systems, an agent explores a knowledge graph before producing an answer and a small set of citations. We frame citation faithfulness as a trajectory-level problem: final citations should not only support the answer, but also account for the graph traversal, structure, and visited-but-uncited entities that may influence it. Through controlled ablation experiments, we compare the effects of isolating, removing, and "},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"Our results show that cited evidence is often necessary, as removing it substantially changes answers and reduces accuracy. However, citations are not sufficient, because accurate answers can also depend on uncited traversal context and surrounding graph structure.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"The controlled ablation experiments (isolating, removing, and masking cited and uncited graph entities) accurately isolate the causal influence of traversal context without artifacts from the specific graphs, agents, or masking procedures used.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"In Agentic GraphRAG, cited evidence is necessary but not sufficient for accurate answers, as uncited traversal context and graph structure also affect results, requiring evaluation of the full retrieval trajectory.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"In Agentic GraphRAG, accurate answers depend on both cited evidence and the uncited traversal context from the agent's graph exploration.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"0c3c024737a3af9eb50e57d50174713f5be0de19ba41155d8da874e2e71b6e5f"},"source":{"id":"2605.15109","kind":"arxiv","version":1},"verdict":{"id":"32747af5-d3de-4fa3-a029-4c9c71ab3f09","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-15T03:13:33.419336Z","strongest_claim":"Our results show that cited evidence is often necessary, as removing it substantially changes answers and reduces accuracy. However, citations are not sufficient, because accurate answers can also depend on uncited traversal context and surrounding graph structure.","one_line_summary":"In Agentic GraphRAG, cited evidence is necessary but not sufficient for accurate answers, as uncited traversal context and graph structure also affect results, requiring evaluation of the full retrieval trajectory.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"The controlled ablation experiments (isolating, removing, and masking cited and uncited graph entities) accurately isolate the causal influence of traversal context without artifacts from the specific graphs, agents, or masking procedures used.","pith_extraction_headline":"In Agentic GraphRAG, accurate answers depend on both cited evidence and the uncited traversal context from the agent's graph exploration."},"references":{"count":12,"sample":[{"doi":"","year":2025,"title":"L. Huang, W. Yu, W. Ma, W. Zhong, Z. Feng, H. Wang, Q. Chen, W. Peng, X. Feng, B. Qin, et al., A survey on hallucination in large language models: Principles, taxonomy, challenges, and open questions,","work_id":"e26d5a9d-9c9a-4052-8415-8abf26f0e142","ref_index":1,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"10.1038/s41586-026-10549-w","year":2026,"title":"A. T. Kalai, O. Nachum, S. S. Vempala, E. Zhang, Evaluating large language models for accuracy incentivizes hallucinations, Nature (2026). URL: https://doi.org/10.1038/s41586-026-10549-w. doi:10.1038/","work_id":"205b16fc-61ee-4bd7-b63d-5e77f703aaac","ref_index":2,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2020,"title":"P. Lewis, E. Perez, A. Piktus, F. Petroni, V. Karpukhin, N. Goyal, H. Küttler, M. Lewis, W.-t. Yih, T. Rocktäschel, S. Riedel, D. Kiela, Retrieval-augmented generation for knowledge-intensive nlp task","work_id":"7e0ba150-a864-4c5a-bdbd-5d858f8ade4b","ref_index":3,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2025,"title":"B. Peng, Y. Zhu, Y. Liu, X. Bo, H. Shi, C. Hong, Y. Zhang, S. Tang, Graph retrieval-augmented generation: A survey, ACM Transactions on Information Systems 44 (2025) 1–52","work_id":"b70e85ce-7c19-4f0d-8702-f26d715bce76","ref_index":4,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"10.1007/s10462-025-11422-4","year":2025,"title":"Agentic AI: A Comprehensive Survey of Architectures, Applications, and Future Directions.Artificial Intelligence Review, 59(11)","work_id":"9c4ef2e0-6d02-4bfe-92d1-8cd8827e3577","ref_index":5,"cited_arxiv_id":"","is_internal_anchor":false}],"resolved_work":12,"snapshot_sha256":"81ed9fd29a2975faf7d06b806c60990c7eb50cac582e1997b215725ffb48046e","internal_anchors":1},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"}