2 papers
cs.CL2024
Self-Attention Limits Working Memory Capacity of Transformer-Based Models
Dongyu Gong, Hantao Zhang
Recent work on Transformer-based large language models (LLMs) has revealed striking limits in their working memory capacity, similar to what has been found in human behavioral stud…
cs.AI2024
Working Memory Capacity of ChatGPT: An Empirical Study
Dongyu Gong, Xingchen Wan, Dingmin Wang
Working memory is a critical aspect of both human intelligence and artificial intelligence, serving as a workspace for the temporary storage and manipulation of information. In thi…