Paper Grading Agent with Personalized Feedback
This agent analyzes student papers in depth to produce detailed grading with criteria-based scoring, personalized feedback, and improvement recommendations. It adapts to the academic level, subject, and evaluation criteria defined by the teacher to provide constructive and supportive feedback.
For who
Secondary and higher education teachers looking to save time on grading while offering rich and individualized feedback to each student.
Input
La copie de l'élève (texte complet ou photo transcrite), le sujet ou la consigne, la matière, le niveau scolaire, le barème de notation et les compétences évaluées.
steps (4)
Evaluation Framework Analysis
promptBuilds the evaluation framework adapted to the subject and level before grading.
Detailed Paper Grading
promptGrades the paper criterion by criterion using the evaluation framework.
Personalized Student Feedback
promptWrites constructive, supportive feedback tailored to the student's strengths and weaknesses.
Teacher Summary and Recommendations
promptProduces a summary card for the teacher with skill diagnosis and remediation suggestions.
Output
Une grille d'évaluation critériée, une correction détaillée avec note, un feedback personnalisé pour l'élève avec exercice de remédiation, et une fiche de synthèse enseignant avec profil de compétences et recommandations pédagogiques.
Example
Input
Subject: French — Level: 10th grade — Type: Literary commentary — Topic: Comment on the extract from The Imaginary Invalid by Moliere (Act III, scene 10) — Scale: /20 — Student paper: 'In this scene Moliere criticizes doctors. He uses irony to mock them. Toinette disguises herself as a doctor and says nonsense. Argan believes her which shows he is foolish. Moliere wants to show that doctors of his time are charlatans...' (350-word paper)
Output
Score: 09/20 — Feedback: 'You have correctly identified Moliere's satirical purpose and the role of Toinette's disguise, which is a good starting point. However, your commentary remains descriptive (you retell the scene) instead of analyzing the literary devices. For example, when you write that Argan is foolish, you should show how Moliere constructs this comedy through the dialogue...' — Remediation exercise: rewrite a paragraph following the Device, Quotation, Analysis, Interpretation method.
Customization
| Parameter | Description | Default |
|---|---|---|
| Niveau de détail du feedback | Ajuste la longueur et la profondeur du feedback élève selon le temps disponible | détaillé (feedback complet avec exercice de remédiation) |
| Ton du feedback | Adapte le registre entre encourageant (élèves fragiles), équilibré (standard) ou exigeant (élèves avancés) | équilibré |
| Format de la note | Système de notation : sur 20, sur 10, par compétences (A/B/C/D), ou sans note | /20 |
Technical Notes
For optimal results, provide the complete student paper rather than a summary. If the paper is handwritten, transcribe it faithfully preserving spelling and syntax errors — the agent will analyze them.
The agent can process papers from all literary subjects and humanities. For scientific subjects (mathematics, physics), formulas must be entered in readable notation (e.g., x^2 + 3x = 0). Diagrams and figures cannot be analyzed in text mode.
To grade multiple papers for the same assignment, rerun the agent for each paper while keeping the same reference framework (step 1). You can then consolidate the summary sheets to obtain a whole-class overview.
Related Prompts
Stable Diffusion Prompt for Translating Text
Stable Diffusion, known for its AI image generation capabilities, can also be used creatively to produce visuals containing translated text. While it's not a dedicated translation tool like DeepL or Google Translate, Stable Diffusion excels at creating images that incorporate text in different languages—signs, posters, book covers, user interfaces, or multilingual marketing materials. By combining precise prompts with typographic control techniques (ControlNet, img2img), you can generate visuals where the source text is replaced with its translation in the target language, while preserving the graphic style, layout, and atmosphere of the original image. This approach is particularly useful for designers, international marketing teams, and content creators who need to quickly localize visual assets without using editing software. In this guide, we provide an optimized prompt to leverage Stable Diffusion for visual translation, with variations suited to your expertise level and tips for achieving professional results.
Master's Thesis Defense Script
Thesis Defense
Article Repurposing for Social Posts
Maximizing the lifespan of existing content
DALL-E Prompt to Correct Text
DALL-E, OpenAI's image generation model, offers a little-known feature: the ability to correct text directly embedded in an image. Whether you've generated an image containing a typo, a misspelled word on a poster, or illegible text on a marketing visual, DALL-E allows you to intervene using its editing tool (inpainting). By selecting the area containing the erroneous text and providing a precise prompt, you can ask the model to regenerate only that portion with the corrected text. This approach avoids recreating the entire image and preserves the original style, colors, and composition. However, it's important to note that DALL-E remains an image generator and not a word processor: typographic rendering can vary, and long or complex texts sometimes require several attempts. To maximize your chances of success, use structured prompts, specify the desired font, and limit the amount of text to correct in a single pass. This guide provides you with the optimal prompts for achieving reliable text corrections in your DALL-E generated images.