Resource-rational-memory-encoding

This is the repository for the ACL 2026 paper "Memory efficiency and resource-rational encoding in sentence processing."

Our models are adapted and trained from nanoGPT (Karpathy, 2023), and can be downloaded from this google drive folder. Code for Experiment 1 and 2 in the paper is based on these models directly trained from nanoGPT.

We've also uploaded our models to huggingface. Code for Experiment 3 in the paper is based on these huggingface models:

• $\lambda$ = 0.0 (model)
• $\lambda$ = 0.001 (model)
• $\lambda$ = 0.01 (model)
• $\lambda$ = 0.1 (model)
• $\lambda$ = 1.0 (model)