A Dual-Task Paradigm to Investigate Sentence Comprehension Strategies in Language Models
For researchers studying human-like language processing in AI, this work provides a new behavioral paradigm to probe how resource limitations affect comprehension strategies, though the findings are incremental as they extend known human-like behaviors to additional models and tasks.
The paper proposes a dual-task paradigm combining arithmetic and sentence comprehension to test whether cognitive resource constraints shift language models toward plausibility-based comprehension. Results show that GPT-4o, o3-mini, and o4-mini exhibit a larger accuracy gap between plausible and implausible sentences under dual-task conditions, mirroring human rational inference.
Language models (LMs) behave more like humans when their cognitive resources are restricted, particularly in predicting sentence processing costs such as reading times. However, it remains unclear whether such constraints similarly affect sentence comprehension strategies. Besides, existing methods do not directly target the balance between memory storage and sentence processing, which is central to human working memory. To address this issue, we propose a dual-task paradigm that combines an arithmetic computation task with a sentence comprehension task, such as "The 2 cocktail + blended 3 =..." Our experiments show that under dual-task conditions, GPT-4o, o3-mini, and o4-mini shift toward plausibility-based comprehension, mirroring humans' rational inference. Specifically, these models show a greater accuracy gap between plausible sentences (e.g., "The cocktail was blended by the bartender") and implausible sentences (e.g., "The bartender was blended by the cocktail") in the dual-task condition compared to the single-task conditions. These findings suggest that constraints on the balance between memory and processing resources promote rational inference in LMs. More broadly, they support the view that human-like sentence comprehension fundamentally arises from the allocation of limited cognitive resources.