N-Back Performance
0.887Human-Model Similarity (1 - NWD)TASKPR
Evaluation Results
| Method | Links | |
|---|---|---|
| TASKPRModel Backbone=Claude Opus 4.62026.05 | 0.887 | |
| TASKPRModel Backbone=Llama 3 8B2026.05 | 0.855 | |
| COMPACTORModel Backbone=Llama 3.3 70B2026.05 | 0.524 | |
| TASKPRModel Backbone=Llama 3.3 70B2026.05 | 0.451 | |
| MEMPRModel Backbone=Llama 3.3 70B2026.05 | 0.379 | |
| HUMPRModel Backbone=Llama 3.3 70B2026.05 | 0.301 | |
| MEMPRModel Backbone=Llama 3 8B2026.05 | 0.077 | |
| COMPACTORModel Backbone=Llama 3 8B2026.05 | 0.028 | |
| MEMPRModel Backbone=Claude Opus 4.62026.05 | 0.016 | |
| HUMPRModel Backbone=Claude Opus 4.62026.05 | 0.015 | |
| COMPACTORModel Backbone=Claude Opus 4.62026.05 | -0.013 | |
| HUMPRModel Backbone=Llama 3 8B2026.05 | -0.072 |