Adjust large prompt for better assessment

[Yuvraj-Sarathe]

Adjust the large prompt for better assessment of the profile.

Right now it is decent enough but is not critical enough.

However, it should also not be rude.

Right now it is extremely leninent on coders! Make it more critical

Make sure the AI detects the actual problems and gives a good score (NOT 7 AT ALL COSTS)!

[Diwakar-odds]

/claim

🚀 Issue Assignment Request & Implementation Proposal

Problem Statement:
The AI prompt for assessing coder profiles is too lenient and hands out high scores (like 7) too easily, failing to provide critical feedback.

Proposed Approach:
I will refine the LLM system prompt to enforce stricter grading criteria. I will instruct the model to explicitly look for edge cases, performance issues, and architectural flaws, ensuring it docks points appropriately without being rude. I'll add specific guidelines to avoid defaulting to a 7.

Expected Outcome:
More accurate, critical, and helpful assessments that reflect true coding ability.

Estimated Time Needed:
24 hours.

[Yuvraj-Sarathe]

@Diwakar-odds I would like you to also add your before and after result changing the prompt results from the same ai for the same account.

[karrisanthoshigayatri]

Hi @Yuvraj-Sarathe, this looks like an interesting prompt engineering task. If the issue becomes available, I'd be happy to work on it. Approach: I'll refine the assessment prompt to apply stricter yet fair evaluation criteria, reduce score inflation, provide more actionable feedback, and compare before-and-after results using the same profile to validate the improvements.

[Yuvraj-Sarathe]

The issue is open for all once again due to inactivity and bad PR.

[Yuvraj-Sarathe]

Hi @Yuvraj-Sarathe, this looks like an interesting prompt engineering task. If the issue becomes available, I'd be happy to work on it. Approach: I'll refine the assessment prompt to apply stricter yet fair evaluation criteria, reduce score inflation, provide more actionable feedback, and compare before-and-after results using the same profile to validate the improvements.

are you still interested in this issue?