In previous versions of ChatGPT, tasks such as document formatting, applying headings, adding page breaks, and generating Table of Contents (TOC) worked flawlessly. However, in recent versions, several issues have arisen that affect these functionalities.
Observed Issues:
- Inconsistent Heading Detection:
-
Problem: The system failed to consistently recognize and apply heading styles (Heading 1, Heading 2, etc.), which are crucial for generating a reliable TOC.
-
Impact: TOC generation was incomplete, with many volumes missing entries or containing empty sections. Pre-existing documents with different formatting styles exacerbated the issue.
Page Break Application Failures: -
Problem: Attempts to add page breaks between “Volume #” sections were unsuccessful. In some cases, page breaks were added incorrectly, causing unintended layout issues and disrupting the document structure.
-
Impact: Documents were rendered less readable, requiring manual intervention to revert changes.
Automatic TOC Generation Failures: -
Problem: The TOC generation tool did not detect headings correctly, even when headings were applied manually. In many cases, only partial or empty TOC entries were produced.
-
Impact: Despite multiple attempts to adjust and reformat headings, the generated TOC remained unreliable.
-
What Worked Previously:
-
In earlier versions, heading detection and TOC generation worked reliably, even with varying document structures.
-
Paragraph spacing and graphic preservation continue to work well.
Conclusion:
These issues, especially around headings and TOC generation, seem to have emerged with recent updates and should be addressed to restore previously reliable functionality. I hope this feedback will contribute to improving document-related features in future versions.