Research

Paper

AI LLM March 18, 2026

Text-to-Stage: Spatial Layouts from Long-form Narratives

Authors

Jefferson Hernandez, Swarnadeep Saha, Chenxi Whitehouse, Sanjeel Parekh, Calvin Murdock, Yuliang Li, W. Owen Brimijoin, Vamsi Krishna Ithapu, Ishwarya Ananthabhotla

Abstract

In this work, we probe the ability of a language model to demonstrate spatial reasoning from unstructured text, mimicking human capabilities and automating a process that benefits many downstream media applications. Concretely, we study the narrative-to-play task: inferring stage-play layouts (scenes, speaker positions, movements, and room types) from text that lacks explicit spatial, positional, or relational cues. We then introduce a dramaturgy-inspired deterministic evaluation suite and, finally, a training and inference recipe that combines rejection SFT using Best-of-N sampling with RL from verifiable rewards via GRPO. Experiments on a text-only corpus of classical English literature demonstrate improvements over vanilla models across multiple metrics (character attribution, spatial plausibility, and movement economy), as well as alignment with an LLM-as-a-judge and subjective human preferences.

Metadata

arXiv ID: 2603.17832
Provider: ARXIV
Primary Category: cs.CL
Published: 2026-03-18
Fetched: 2026-03-19 06:01

Related papers

Raw Data (Debug)
{
  "raw_xml": "<entry>\n    <id>http://arxiv.org/abs/2603.17832v1</id>\n    <title>Text-to-Stage: Spatial Layouts from Long-form Narratives</title>\n    <updated>2026-03-18T15:26:27Z</updated>\n    <link href='https://arxiv.org/abs/2603.17832v1' rel='alternate' type='text/html'/>\n    <link href='https://arxiv.org/pdf/2603.17832v1' rel='related' title='pdf' type='application/pdf'/>\n    <summary>In this work, we probe the ability of a language model to demonstrate spatial reasoning from unstructured text, mimicking human capabilities and automating a process that benefits many downstream media applications. Concretely, we study the narrative-to-play task: inferring stage-play layouts (scenes, speaker positions, movements, and room types) from text that lacks explicit spatial, positional, or relational cues. We then introduce a dramaturgy-inspired deterministic evaluation suite and, finally, a training and inference recipe that combines rejection SFT using Best-of-N sampling with RL from verifiable rewards via GRPO. Experiments on a text-only corpus of classical English literature demonstrate improvements over vanilla models across multiple metrics (character attribution, spatial plausibility, and movement economy), as well as alignment with an LLM-as-a-judge and subjective human preferences.</summary>\n    <category scheme='http://arxiv.org/schemas/atom' term='cs.CL'/>\n    <category scheme='http://arxiv.org/schemas/atom' term='cs.AI'/>\n    <category scheme='http://arxiv.org/schemas/atom' term='cs.LG'/>\n    <published>2026-03-18T15:26:27Z</published>\n    <arxiv:primary_category term='cs.CL'/>\n    <author>\n      <name>Jefferson Hernandez</name>\n    </author>\n    <author>\n      <name>Swarnadeep Saha</name>\n    </author>\n    <author>\n      <name>Chenxi Whitehouse</name>\n    </author>\n    <author>\n      <name>Sanjeel Parekh</name>\n    </author>\n    <author>\n      <name>Calvin Murdock</name>\n    </author>\n    <author>\n      <name>Yuliang Li</name>\n    </author>\n    <author>\n      <name>W. Owen Brimijoin</name>\n    </author>\n    <author>\n      <name>Vamsi Krishna Ithapu</name>\n    </author>\n    <author>\n      <name>Ishwarya Ananthabhotla</name>\n    </author>\n  </entry>"
}