> ## Documentation Index
> Fetch the complete documentation index at: https://docs.ryzeapp.org/llms.txt
> Use this file to discover all available pages before exploring further.

# Papers & mistake memory

> How an uploaded past paper becomes mistake memory and a revision signal.

Upload a PDF or image of a past paper. Ryze extracts the questions, and the student self-marks each result.

## How it works

1. **Upload** — a paper up to 20 MB (PDF, JPG, PNG, WebP) is stored privately against the student's account.
2. **Extraction** — Ryze reads the paper and extracts structured questions, including parent/sub-part structure.
3. **Marking** — the student marks each question Correct / Partial / Wrong, optionally with exact earned marks and a self-chosen mistake category. A self-chosen category needs no AI call — it's the student's own assessment.
4. **Mistake insight (optional)** — the student can opt in to an AI-generated insight across their missed questions, batched rather than one call per question.
5. **Performance view** — Ryze shows the latest result per distinct question, known marks lost, a "Fix First" ranking, and patterns across papers.

> [!NOTE]
>
> Unknown partial marks are never guessed — they're simply excluded from scoring. Mastery uses a neutral starting point plus a minimum amount of evidence; it's a preparation signal, not a predicted exam score, and time spent alone never substitutes for measured mastery.

## Session linkage

A first result saved during an active study session appears in that session's recap automatically. Editing an earlier result corrects its original record — it doesn't count as new practice.


## Related topics

- [Introduction](https://docs.ryzeapp.org/index.md)
- [Study](https://docs.ryzeapp.org/product/study.md)
- [Research](https://docs.ryzeapp.org/research/overview.md)


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.