Comparison of average error between unlearned positions and unlearned compositions

Clustered bar chart showing the comparison of average error between unlearned positions and unlearned compositions. 4 groups of training compositions of decreasing variety are shown on the x axis, y axis show average error from 0.00 to 0.07. As the variety within training groups decreases, the average error increases.

Comparison of average error between cases of unlearned object positions (blue) and unlearned compositions of words (orange) among groups with different numbers of compositions in training – Group A is trained on 40 compositions of 5 colors and 8 verb clauses, like ‘grasp yellow’, ‘move red to left’, and ‘put blue on green’. As the number of compositions in training decreases, error increases, especially when encountering unlearned compositions. This suggests that a greater variety in training is crucial for learning to generalize to different combinations of words, including unlearned combinations.

Date:
23 December 2024
Credit:
Vijayaraghavan et al., 2025
Share on: