TRAM: Benchmarking Temporal Reasoning for Large Language Models

TL;DR

TRAM is a comprehensive benchmark consisting of ten datasets designed to evaluate the temporal reasoning abilities of large language models across various temporal aspects, highlighting current limitations and guiding future improvements.

Contribution

This paper introduces TRAM, the first standardized, multi-faceted benchmark for evaluating temporal reasoning in large language models, covering diverse temporal tasks.

Findings

Current LLMs lag behind human performance in temporal reasoning

GPT-4 and Llama2 show limited capabilities in zero-shot and few-shot settings

Baseline models perform significantly worse than state-of-the-art models

Abstract

Reasoning about time is essential for understanding the nuances of events described in natural language. Previous research on this topic has been limited in scope, characterized by a lack of standardized benchmarks that would allow for consistent evaluations across different studies. In this paper, we introduce TRAM, a temporal reasoning benchmark composed of ten datasets, encompassing various temporal aspects of events such as order, arithmetic, frequency, and duration, designed to facilitate a comprehensive evaluation of the TeR capabilities of large language models (LLMs). We evaluate popular LLMs like GPT-4 and Llama2 in zero-shot and few-shot scenarios, and establish baselines with BERT-based and domain-specific models. Our findings indicate that the best-performing model lags significantly behind human performance. It is our aspiration that TRAM will spur further progress in enhancing the TeR capabilities of LLMs.