dorsal/arxiv
View SchemaDeriving Character Logic from Storyline as Codified Decision Trees
| Authors | Letian Peng, Kun Zhou, Longfei Yun, Yupeng Hou, Jingbo Shang |
|---|---|
| Categories | |
| ArXiv ID | 2601.10080vv1 |
| URL | https://arxiv.org/abs/2601.10080 |
| License | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ |
Abstract
Role-playing (RP) agents rely on behavioral profiles to act consistently across diverse narrative contexts, yet existing profiles are largely unstructured, non-executable, and weakly validated, leading to brittle agent behavior. We propose Codified Decision Trees (CDT), a data-driven framework that induces an executable and interpretable decision structure from large-scale narrative data. CDT represents behavioral profiles as a tree of conditional rules, where internal nodes correspond to validated scene conditions and leaves encode grounded behavioral statements, enabling deterministic retrieval of context-appropriate rules at execution time. The tree is learned by iteratively inducing candidate scene-action rules, validating them against data, and refining them through hierarchical specialization, yielding profiles that support transparent inspection and principled updates. Across multiple benchmarks, CDT substantially outperforms human-written profiles and prior profile induction methods on $85$ characters across $16$ artifacts, indicating that codified and validated behavioral representations lead to more reliable agent grounding.
{
"annotation_id": "d5ea4344-f8e9-4fe5-9473-dcecb09efc99",
"date_created": "2026-02-17T05:53:23.787000Z",
"date_modified": "2026-02-17T05:53:23.787000Z",
"file_hash": "a714720991146a089b204011ca9b57f384aae4053bff68c18834744c9be358e0",
"private": false,
"record": {
"abstract": "Role-playing (RP) agents rely on behavioral profiles to act consistently across diverse narrative contexts, yet existing profiles are largely unstructured, non-executable, and weakly validated, leading to brittle agent behavior. We propose Codified Decision Trees (CDT), a data-driven framework that induces an executable and interpretable decision structure from large-scale narrative data. CDT represents behavioral profiles as a tree of conditional rules, where internal nodes correspond to validated scene conditions and leaves encode grounded behavioral statements, enabling deterministic retrieval of context-appropriate rules at execution time. The tree is learned by iteratively inducing candidate scene-action rules, validating them against data, and refining them through hierarchical specialization, yielding profiles that support transparent inspection and principled updates. Across multiple benchmarks, CDT substantially outperforms human-written profiles and prior profile induction methods on $85$ characters across $16$ artifacts, indicating that codified and validated behavioral representations lead to more reliable agent grounding.",
"arxiv_id": "2601.10080",
"authors": [
"Letian Peng",
"Kun Zhou",
"Longfei Yun",
"Yupeng Hou",
"Jingbo Shang"
],
"categories": [
"cs.CL"
],
"license": "http://arxiv.org/licenses/nonexclusive-distrib/1.0/",
"title": "Deriving Character Logic from Storyline as Codified Decision Trees",
"url": "https://arxiv.org/abs/2601.10080",
"version": "v1"
},
"schema_id": "dorsal/arxiv",
"source": {
"execution_id": "f36558d1-6b83-4a78-b922-8802617dcbba",
"id": "arXiv Dataset",
"type": "Model",
"variant": "snapshot-2026-01-17",
"version": "0.1.0"
},
"user_id": 1000002
}