Deriving Character Logic from Storyline as Codified Decision Trees
Letian Peng, Kun Zhou, Longfei Yun, Yupeng Hou, Jingbo Shang
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 datadriven framework that induces an executable and interpretable decision structure from largescale 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. 1 1 Codes and datasets used in experiments are available at https://github.com/KomeijiForce/Codified_Decision_Tree Node 1-1 (ROOT) -Positive character -Direct expression Node 2-2 -Easily ignited -Active participation Node 2-1 -Popipa member -Vo.&Gt. position Node 2-3 -Poor at study -Always space out 2-2: An exciting plan comes? 2-1: Live or rehearsal? 2-3: In a class scene? 3-1: Others (e.g., Arisa) resist? Node 3-1 -Encourage other members Node 3-2 -Generally don't know the answer 3-2: Questioned by teacher? × (Crashes into the classroom) Hi, Kasumi! I've come up with a super good idea! Let's hold a live on the moon! That's gotta be super happy, right? Yeah ! Hold it, hold it, hold it! Why does moon = happy in your head?! And more importantly, there's no air on the moon! You'd suffocate before even playing a song! Tsurumaki Kokoro/弦巻こころ Ichigaya Arisa/市ヶ谷有咲 Ehh ? But that sounds amazing! Popipa's live on moon, shining under stars, kirakira! Don't worry, Arisa, we'll figure something out! It'll be fun! ♪ Toyama Kasumi/戸山香澄 Codified Decision Tree (CDT) for Toyama Kasumi/戸山香澄
Ask about this paper
Your agent reads all of it.
Lune indexed this paper to the last equation, along with the top-tier papers that cite it. Ask a question and the answer quotes them.
Your agent calls
Luneget_paper_fulltext
Free to start. No credit card required.
Terminal
Install the CLIlune papers fulltext dd993bc9-25d5-4c0a-a7bb-b6cd77d055d1Cited by top-tier papers1
Ask how each one uses itBuilds on14
- LoRA: Low-Rank Adaptation of Large Language ModelsEdward J. Hu, Yelong Shen, Phillip Wallis, Zeyuan Allen-Zhu et al.ICLR 2022 · 18,833 citations
- SimCSE: Simple Contrastive Learning of Sentence EmbeddingsTianyu Gao, Xingcheng Yao, Danqi ChenEMNLP 2021 · 2,496 citations
- PAL: Program-aided Language ModelsLuyu Gao, Aman Madaan, Shuyan Zhou, Uri Alon et al.ICML 2023 · 700 citations
- DeBERTaV3: Improving DeBERTa using ELECTRA-Style Pre-Training with Gradient-Disentangled Embedding SharingPengcheng He, Jianfeng Gao, Weizhu ChenICLR 2023 · 394 citations
- Character-LLM: A Trainable Agent for Role-PlayingYunfan Shao, Linyang Li, Junqi Dai, Xipeng QiuEMNLP 2023 · 97 citations
Related papers
- Codifying Character Logic in Role-PlayingLetian Peng, Jingbo ShangNeurIPS 2025 · 7 citations
- DREAM: LLM-based Dynamic Role-playing via Event-Aware Memory GraphZhihao Xiao, Mengting Li, Xintao Wang, Linfeng Li et al.KDD 2026 · 1 citation
- Codified Finite-state Machines for Role-playingLetian Peng, Yupeng Hou, Kun Zhou, Jingbo ShangICLR 2026 · 1 citation
- Generative Models for Automatic Medical Decision Rule Extraction from TextYuxin He, Buzhou Tang, Xiaoling WangEMNLP 2024 · 2 citations
- NL2CA: Auto-formalizing Cognitive Decision-Making from Natural Language Using an Unsupervised CriticNL2LTL FrameworkZihao Deng, Yijia Li, Renrui Zhang, Peijun YeAAAI 2026
