1Cademy - A language model is tasked with finding an answer span within a text. After processing the text, it produces the following log-probabilities for each token being the start or the end of the answer. To find the best span, one must find the start index ($j_1$) and end index ($j_2$) that maximize the sum of their log-probabilities, with the constraint that the start index cannot be after the end index ($j_1 \le j_2$). | Index | Token | Start Log-Prob | End Log-Prob | |-------|-----------|----------------|--------------| | 1 | The | -5.1 | -8.1 | | 2 | first | -4.2 | -7.2 | | 3 | person | -4.5 | -6.5 | | 4 | was | -5.5 | -5.5 | | 5 | Neil | -0.9 | -3.1 | | 6 | Armstrong | -2.1 | -0.5 | Given the table above, what is the predicted answer span?

Learn Before

Span Prediction Inference Formula

Multiple Choice

A language model is tasked with finding an answer span within a text. After processing the text, it produces the following log-probabilities for each token being the start or the end of the answer. To find the best span, one must find the start index ( $j_1$ ) and end index ( $j_2$ ) that maximize the sum of their log-probabilities, with the constraint that the start index cannot be after the end index ( $j_1 \le j_2$ ).

Index	Token	Start Log-Prob	End Log-Prob
1	The	-5.1	-8.1
2	first	-4.2	-7.2
3	person	-4.5	-6.5
4	was	-5.5	-5.5
5	Neil	-0.9	-3.1
6	Armstrong	-2.1	-0.5

Given the table above, what is the predicted answer span?

Updated 2025-09-29

Contributors are:

Who are from:

Learn Before

Related