arxiv macros and figure environment seem mis-handled
For example,
- In the first example of arxiv validation set,
Then $(a)\ensuremath{\mathbb{R}}ightarrow (b)\ensuremath{\mathbb{R}}ightarrow (c)$.
- In second example of arxiv validation set, (unclosed \caption, lost caption)
\begin{figure}
\caption{\label{f1}
\label{f1}
\end{figure}
- In third example of arxiv validation set, (unclosed \caption)
\begin{figure}
\caption{Proposed hierarchical control architecture. $\mathcal{S}
\label{fig:hierarchical_outinner}
\end{figure}
- In fourth example of arxiv validation set,
\deltaltaand\lambdambdaappear
The macros look like you were too zealous in replace macros with their definitions. For example \deltalta can arise from replace \de in \delta with \delta without checking for word boundary.
Hi, we sincerely appreciate your careful feedback.
Upon checking, for the unclosed \caption issue, it is because of the nested parentheses in \catption, which we hadn't considered before; for the macros replacement issue, it is due to not checking for word boundary.
We have fixed the above issues, the new version is coming soon! Stay tuned!
Thanks again!
i found also a large amount of spacing before EOL, as well as EOL's at the end of the texts. I cleaned up on my side, but since u are on a re-release process maybe it can have room for it :D
btw, first models out of it are out.. i can tell the performance is very good :)
hi, thanks for your feedback. Do you mean those spaces, some unnatural new lines in a paragraph?
btw, first models out of it are out.. i can tell the performance is very good :)
Wow! We are eagerly anticipating the arrival of your models or specific results."
they are on the hub, card is citing accordingly. look ur dataset card :D
give it a shot:
fblgit/UNA-POLAR-10.7B-InstructMath-v2
Hi, our latest version (v0.2) of MathPile is out. Welcome to check it.
@SinclairWang
i noticed a very large amount of excessive \n and excessive spacing, did u guys fixed those on v0.2 ?
Yeah, the issue improper line wrapping in paragraphs is fixed (in arXiv). Any feedback is welcome!