{"id":{"repo_id":"msu","oai_identifier":"oai:d.lib.msu.edu:etd_53112"},"canonical_url":"https://search.ndltd.org/etd/msu/oai:d.lib.msu.edu:etd_53112","repository":{"repo_id":"msu","name":"Michigan State University","base_url":"https://d.lib.msu.edu/oai"},"display":{"title":"SPATIAL REASONING OVER NARRATIVES","abstract":"Thesis (M.S.)--Michigan State University. Computer Science - Master of Science, 2026","abstract_html":"Thesis (M.S.)--Michigan State University. Computer Science - Master of Science, 2026","abstract_has_math":false,"creators":["Shakir, SK M M Samir"],"institution":null,"degree_name":null,"degree_level":null,"degree_discipline":null,"degree_department":null,"school":null,"contributors":["Kordjamshidi, Parisa","Ross, Arun","Cui, Zijun"],"advisors":[],"committee_chairs":[],"committee_members":[],"year":2026,"date_issued":"2026","date_published":"2026","updated_at":"2026-11-18T18:59:37Z","subjects":["Computer science","Artificial intelligence"],"languages":["eng"],"rights":["In Copyright"],"rights_urls":[],"identifier_entries":[{"key":"dc:identifier","label":"Identifier","values":["etd:53112","local:Shakir_grad.msu_0128N_22124"],"render_values":[{"text":"etd:53112","href":null,"code":true},{"text":"local:Shakir_grad.msu_0128N_22124","href":null,"code":true}]}],"person_identifier_entries":[]},"links":{"outbound_url":"https://doi.org/doi:10.25335/km1a-9c05","outbound_label":"DOI","outbound_source":"dc:identifier","record_url":"https://doi.org/doi:10.25335/km1a-9c05","record_label":"Repository record","record_kind":"doi","record_source":"dc:identifier","download_url":null,"download_label":null,"download_source":null,"access_status":null,"embargo_until":null,"access_note":null},"source_record":{"url":"https://d.lib.msu.edu/oai?verb=GetRecord&metadataPrefix=oai_dc&identifier=oai%3Ad.lib.msu.edu%3Aetd_53112","prefix":"oai_dc"},"metadata_groups":[{"id":"people","label":"People","entries":[{"key":"dc:contributor","label":"Contributor","values":["Kordjamshidi, Parisa","Ross, Arun","Cui, Zijun"]},{"key":"dc:creator","label":"Author","values":["Shakir, SK M M Samir"]}]},{"id":"academic_context","label":"Academic Context","entries":[{"key":"dc:date","label":"Dc Date","values":["2026"]},{"key":"dc:relation","label":"Dc Relation","values":["Electronic Theses & Dissertations"]},{"key":"dc:type","label":"Dc Type","values":["text","Electronic dissertations"]}]},{"id":"subjects_keywords","label":"Subjects and Keywords","entries":[{"key":"dc:subject","label":"Dc Subject","values":["Computer science","Artificial intelligence"]}]},{"id":"language_rights","label":"Language and Rights","entries":[{"key":"dc:language","label":"Dc Language","values":["eng"]},{"key":"dc:rights","label":"Dc Rights","values":["In Copyright"]}]},{"id":"identifiers","label":"Identifiers","entries":[{"key":"dc:identifier","label":"Identifier","values":["etd:53112","local:Shakir_grad.msu_0128N_22124","https://doi.org/doi:10.25335/km1a-9c05"]}]},{"id":"additional","label":"Additional Metadata","entries":[{"key":"dc:description","label":"Description","values":["Thesis (M.S.)--Michigan State University. Computer Science - Master of Science, 2026","Spatial reasoning over text is challenging because it often requires integrating global context and multi-hop reasoning while applying spatial logic. For example, from a text description of a dining table, determining the location of a specific object, such as a cup of coffee, is a form of spatial reasoning. In principle, such problems can be solved using and combining spatial rules such as “If X is to the left of Y, then Y is to the right of X.\" Although LLMs may have knowledge of such reasoning rules internally encoded in their parameters, we hypothesize that making these rules explicit in the context can improve their performance on such problems. However, there are two key challenges to this procedure: one is ensuring coverage of all applicable rules, and the second is the large number of irrelevant rules for a specific problem. In this thesis, we investigate methods for adding such rules to the context while trying to address these two core challenges. Our approach includes hand-curated rule sets, together with techniques to select and instantiate rules for each problem. We also study a method where the LLM itself generates the rules, avoiding dependence on a fixed rule set while helping address both the coverage and irrelevance issues. We evaluate these ideas on two datasets. One is TurnaboutLLM, a benchmark derived from the detective game Ace Attorney, where identifying contradictions requires inferring spatial facts from diverse and realistic narrative contexts, making rule coverage especially difficult. The second is SpaRTUN, a dataset for complex spatial reasoning in a toy environment. The more rigid structure reduces the number of spatial rules needed, shifting the challenge to how those rules are applied. Overall, our findings show the advantage of explicit rules in the context, while we provide an extensive analysis of these datasets and pinpoint various challenges in deriving and applying such rules in spatial reasoning problems.","Description based on online resource. Title from PDF t.p. (Michigan State University Fedora Repository, viewed ).","Includes bibliographical references."]},{"key":"dc:format","label":"Dc Format","values":["application/pdf","1 online resource (65 p.) : ill."]},{"key":"dc:title","label":"Title","values":["SPATIAL REASONING OVER NARRATIVES"]}]}],"canonical_facts":{"dc:contributor":["Kordjamshidi, Parisa","Ross, Arun","Cui, Zijun"],"dc:creator":["Shakir, SK M M Samir"],"dc:date":["2026"],"dc:description":["Thesis (M.S.)--Michigan State University. Computer Science - Master of Science, 2026","Spatial reasoning over text is challenging because it often requires integrating global context and multi-hop reasoning while applying spatial logic. For example, from a text description of a dining table, determining the location of a specific object, such as a cup of coffee, is a form of spatial reasoning. In principle, such problems can be solved using and combining spatial rules such as “If X is to the left of Y, then Y is to the right of X.\" Although LLMs may have knowledge of such reasoning rules internally encoded in their parameters, we hypothesize that making these rules explicit in the context can improve their performance on such problems. However, there are two key challenges to this procedure: one is ensuring coverage of all applicable rules, and the second is the large number of irrelevant rules for a specific problem. In this thesis, we investigate methods for adding such rules to the context while trying to address these two core challenges. Our approach includes hand-curated rule sets, together with techniques to select and instantiate rules for each problem. We also study a method where the LLM itself generates the rules, avoiding dependence on a fixed rule set while helping address both the coverage and irrelevance issues. We evaluate these ideas on two datasets. One is TurnaboutLLM, a benchmark derived from the detective game Ace Attorney, where identifying contradictions requires inferring spatial facts from diverse and realistic narrative contexts, making rule coverage especially difficult. The second is SpaRTUN, a dataset for complex spatial reasoning in a toy environment. The more rigid structure reduces the number of spatial rules needed, shifting the challenge to how those rules are applied. Overall, our findings show the advantage of explicit rules in the context, while we provide an extensive analysis of these datasets and pinpoint various challenges in deriving and applying such rules in spatial reasoning problems.","Description based on online resource. Title from PDF t.p. (Michigan State University Fedora Repository, viewed ).","Includes bibliographical references."],"dc:format":["application/pdf","1 online resource (65 p.) : ill."],"dc:identifier":["etd:53112","local:Shakir_grad.msu_0128N_22124","https://doi.org/doi:10.25335/km1a-9c05"],"dc:language":["eng"],"dc:relation":["Electronic Theses & Dissertations"],"dc:rights":["In Copyright"],"dc:subject":["Computer science","Artificial intelligence"],"dc:title":["SPATIAL REASONING OVER NARRATIVES"],"dc:type":["text","Electronic dissertations"]},"updated_at":"2026-11-18T18:59:37Z"}