token_probs 3rd to the last confidence result always giving low confidence level

[topa-22]

Hello can I ask why is it the 3rd to the last result of token_probs always giving me a low confidence score? regardless of any phrase in the word/token. BTW it pronounced perfectly by an American native speaker

"text_tokens": [
"",
" Sm",
"ok",
"y",
" fires",
" lack",
" flame",
" and",
" heat",
".",
""
],
"token_probs": [
0.9760881662368774,
0.9851165413856506,
0.6812853217124939,
0.7833024263381958,
0.8377408385276794,
0.8858919739723206,
0.8837830424308777,
0.8425670862197876,
0.10593103617429733,
0.9830518364906311,
0.0
],