ACPBench Hard: Unrestrained Reasoning about Action, Change, and Planning

TL;DR

ACPBench Hard is introduced, a generative version of ACPBench, with open-ended questions which the model needs to answer, indicating that even the current frontier language models have a long way to go before they can reliably reason about planning.

Abstract

The ACPBench dataset provides atomic reasoning tasks required for efficient planning. The dataset is aimed at distilling the complex plan generation task into separate atomic reasoning tasks in their easiest possible form, boolean or multiple-choice questions, where the model has to choose the right answer from the provided options. While the aim of ACPBench is to test the simplest form of reasoning about action and change, when tasked with planning, a model does not typically have options to choose from and thus the reasoning required for planning dictates an open-ended, generative form for these tasks. To that end, we introduce ACPBench Hard, a generative version of ACPBench, with open-ended questions which the model needs to answer. Models that perform well on these tasks could in principle be integrated into a planner or be used directly as a policy. We discuss the complexity of these tasks as well as the complexity of validating the correctness of their answers and present validation algorithms for each task. Equipped with these validators, we test the performance of a variety of models on our tasks and find that for most of these tasks the performance of even the largest models is still subpar. Our experiments show that no model outperforms another in these tasks and with a few exceptions all tested language models score below 65%, indicating that even the current frontier language models have a long way to go before they can reliably reason about planning. In fact, even the so-called reasoning models struggle with solving these reasoning tasks. ACPBench Hard collection is available at the following link: https://ibm.github.io/ACPBench

Figures (8)

• Figure 1: Task-wise accuracy of large size language models.
• Figure 2: Example of a question for the progression task in ACPBench Hard. The question is composed out of the context, which contains the domain and the problem description, as well as the current state description, and the inputs, the actual question.
• Figure 3: Grammar used for parsing the model response.
• Figure 4: Domain-wise accuracy of GPT-4o.
• Figure 5: Comparison of prediction error for GPT-4o on different question formats.