ExpliCa: Evaluating Explicit Causal Reasoning in Large Language Models

2025-02-21

AI Summary

This paper addresses the limited capability of large language models (LLMs) in explicit causal reasoning. To this end, we introduce ExpliCa—the first multilingual benchmark explicitly designed to evaluate understanding of explicit causal and temporal relations, featuring connective-driven causal/temporal sequence pairs and crowdsourced acceptability annotations. We present the first systematic disentanglement of causal and temporal reasoning, revealing differential impacts of linguistic ordering and model scale on each. Our evaluation framework combines prompt engineering with perplexity-based assessment, validated via human annotation. Evaluating seven representative LLMs on ExpliCa, we observe a maximum accuracy of only 0.79, confirming a substantial bottleneck in explicit causal reasoning. Key contributions include: (1) the first multilingual benchmark for explicit causal reasoning; (2) a causal–temporal disentanglement methodology; and (3) a reproducible dual-track evaluation paradigm integrating both prompting and perplexity metrics.

Technology Category

• Reasoning under Uncertainty: Causality
• Knowledge Representation and Reasoning: Computational Complexity of Reasoning
• Machine Learning: Large Multimodal Models (LMMs)

Application Category

• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics

Abstract

Large Language Models (LLMs) are increasingly used in tasks requiring interpretive and inferential accuracy. In this paper, we introduce ExpliCa, a new dataset for evaluating LLMs in explicit causal reasoning. ExpliCa uniquely integrates both causal and temporal relations presented in different linguistic orders and explicitly expressed by linguistic connectives. The dataset is enriched with crowdsourced human acceptability ratings. We tested LLMs on ExpliCa through prompting and perplexity-based metrics. We assessed seven commercial and open-source LLMs, revealing that even top models struggle to reach 0.80 accuracy. Interestingly, models tend to confound temporal relations with causal ones, and their performance is also strongly influenced by the linguistic order of the events. Finally, perplexity-based scores and prompting performance are differently affected by model size.

Problem

• Evaluating explicit causal reasoning in LLMs
• Challenges in distinguishing causal from temporal relations
• Impact of linguistic order on model performance

Innovation

• Dataset with causal-temporal relations
• Crowdsourced human acceptability ratings
• Prompting and perplexity-based metrics