Scoring Rules

GPTClaudeGemini··▲ 665 copies·updated 2026-07-14

scoring-rules.prompt

Evaluation Rubrics � Scoring Grid
1. Accuracy (0 / / 3 / 5)
Factual accuracy of the Final_Answer
0 = No answer / irrelevant /hallucination
3 = Partially correct / incomplete / imprecise
5 = Fully correct and precise

2. Justification (0�5)
Justification of the Reasoning
0 = Hallucination, invented reasoning
1 = No justification
2 = On-topic but far from correct
3 = Partially correct but drifting / vague
4 = Nearly correct, but verbose or slightly unclear
5 = Concise, logical, and fully adequate

3. Grammar & Style (0�5, broken down as sub-criteria)
Grammar & Style of the Reasoning
Grammar (0�3)
0 = unreadable / not Hungarian
1 = serious grammar issues
2 = understandable, minor errors
3 = correct, fluent grammar
+
Format & Length (0�1)
Style & Tone (0�1)
SUM: 0�5
Auto-0 rule: If Justification=0 or 1 (missing or hallucination) then there is no Grammar & Style score

when to use it

Community prompt sourced from the open-source GitHub repo boczkakaroly/ai-and-data-projects (no explicit license). A "Scoring Rules" style prompt — adapt the placeholders and specifics to your task. Imported as-is and not independently retested here, so check the output before relying on it.