Choice-75: A Dataset on Decision Branching in Script Learning

TL;DR

Choice-75 introduces a novel dataset for decision branching in script learning, challenging AI systems to handle complex, decision-based narratives with diverse scenarios, highlighting current LLM limitations.

Contribution

The paper presents the first benchmark dataset focusing on decision branching in script learning, expanding beyond linear event sequences to include diverse, decision-driven scenarios.

Findings

LLMs perform decently on overall tasks

Significant challenges remain in complex scenarios

Choice-75 enables evaluation of decision-making in narrative understanding

Abstract

Script learning studies how stereotypical events unfold, enabling machines to reason about narratives with implicit information. Previous works mostly consider a script as a linear sequence of events while ignoring the potential branches that arise due to people's circumstantial choices. We hence propose Choice-75, the first benchmark that challenges intelligent systems to make decisions given descriptive scenarios, containing 75 scripts and more than 600 scenarios. We also present preliminary results with current large language models (LLM). Although they demonstrate overall decent performance, there is still notable headroom in hard scenarios.