Can LLM Agents Stick to the Script? Modeling Commitment in Interactive Narratives

The rapid advancement of Large Language Models (LLMs) is revolutionizing AI for Games by enabling open-ended and fluid interactive storytelling. However, existing research has largely overlooked the critical challenge of maintaining logical consistency and narrative integrity against unconstrained user interventions. To address this, we formulate this challenge as Narrative Commitment Preservation (NCP), and take interactive narrative as our testbed. We introduce NCP-Bench, a benchmark of 100 narrative environments derived from movie synopses. Each environment includes a structured narrative specification (trajectory, commitments, and initial facts) that we can automatically check throughout the interaction between the player agent and the narrator agent. Experiments across state-of-the-art LLMs reveal that high linguistic quality does not guarantee commitment preservation; even strong models frequently generate logically conflicting content under adversarial interventions, with the best-performing model (GPT-5.2) achieving only 40\% survival rate after 20 turns and fact conflict rates ranging from 40\% to 68\% across models.

When artificial intelligence acts as a storyteller in video games, it can write beautiful, immersive dialogue — but it frequently forgets what it just said. Imagine a game master who tells you a door is locked, then casually lets you walk through it in the next moment without any explanation. We built a new testing ground with 100 interactive stories drawn from movies, where an AI must guide a player through a plot while we deliberately try to trick it into making logical mistakes. Our system automatically spots when the AI contradicts itself or breaks the rules of its own world. We discovered that even the most advanced AI models fail roughly 60% of the time after just 20 back-and-forth exchanges. This work provides a practical way to measure whether an AI storyteller is not only eloquent, but also logically trustworthy.