Conversation
![gemini-code-assist[bot]](https://avatars.githubusercontent.com/in/956858?s=60&v=4){kind=small}
There was a problem hiding this comment.
Code Review
This pull request updates the LLMTextQualityV5 class by refining the criteria for identifying mathematical and code-related errors. It introduces two distinct failure modes for formulas—broken LaTeX syntax and stripped mathematical content—and provides detailed corruption patterns for code blocks, such as lost indentation and broken identifiers. The review feedback correctly identifies that several examples intended to show 'broken' syntax or 'unclosed' tags were actually syntactically valid or correctly closed, and points out an inconsistency in the code block example regarding the presence of code fences.
| **(A) Broken LaTeX syntax** — delimiters or environments are present but malformed: | ||
| - Delimiters unmatched: $ without closing $ (LaTeX context, not dollar signs) | ||
| - Environments unclosed: \\begin{{align}} without \\end{{align}} | ||
| - Syntax broken: \\frac{{a}}{{b missing closing }} |
There was a problem hiding this comment.
The example provided for "Syntax broken" (\\frac{{a}}{{b missing closing }}) results in a syntactically valid LaTeX expression (\frac{a}{b missing closing }) when the template is formatted. To effectively demonstrate broken syntax to the LLM, the closing brace should be omitted so that the expression remains unclosed.
| - Syntax broken: \\frac{{a}}{{b missing closing }} | |
| - Syntax broken: \\frac{{a}}{{b missing closing |
| - Delimiters unmatched: $ without closing $ (LaTeX context, not dollar signs) | ||
| - Environments unclosed: \\begin{{align}} without \\end{{align}} | ||
| - Syntax broken: \\frac{{a}}{{b missing closing }} | ||
| - HTML tags unclosed: <sub>text without </sub> |
There was a problem hiding this comment.
The example for "HTML tags unclosed" (<sub>text without </sub>) actually shows a correctly closed tag. To properly demonstrate an unclosed tag to the model, the closing </sub> should be removed.
| - HTML tags unclosed: <sub>text without </sub> | |
| - HTML tags unclosed: <sub>text without |
| for i in range( 1,len(sys .argv), 2): | ||
| pts .append([int(sys .argv[i]), int(sys .argv[i +1])]) | ||
| ``` | ||
| Correct version would have a code fence, proper indentation, and no spaces inside `sys.argv`. |
There was a problem hiding this comment.
The explanation states that a "Correct version would have a code fence", but the "BAD" example provided in the preceding lines (85-90) already includes a code fence. This inconsistency might confuse the LLM. The explanation should be updated to focus on the issues actually demonstrated in the example, such as the lack of indentation and the corrupted identifiers.
| Correct version would have a code fence, proper indentation, and no spaces inside `sys.argv`. | |
| Correct version would have proper indentation and no spaces inside identifiers. |
2 participants
No description provided.