gold-vs-trans/swedish_skolprov

---
configs:
- config_name: all
data_files:
- split: train
path:
- "hp_2024-10-20_VERB.csv"
- "kunskapsprov_lakare_2024-11-26-mcq-140.csv"
- "kunskapsprov_tandlakare_TDL-2411.csv"
- "kunskapsprov_audionom_2023-10-17.csv"
- "kunskapsprov_apotekare_2024-01-23_del_a.csv"
- "matematik_och_fysikprovet.csv"
default: true
- config_name: högskoleprovet
data_files: "hp_2024-10-20_VERB.csv"
- config_name: kunskapsprov_läkare
data_files: "kunskapsprov_lakare_2024-11-26-mcq-140.csv"
- config_name: kunskapsprov_tandläkare
data_files: "kunskapsprov_tandlakare_TDL-2411.csv"
- config_name: kunskapsprov_audionom
data_files: "kunskapsprov_audionom_2023-10-17.csv"
- config_name: kunskapsprov_apotekare
data_files: "kunskapsprov_apotekare_2024-01-23_del_a.csv"
- config_name: matematik_och_fysikprovet
data_files: "matematik_och_fysikprovet.csv"
---

# Swedish skolprov documentation

This repository contains data from six Swedish knowledge tests. The data is in the form of multiple-choice questions and answers. The data is stored in CSV files.

The following tests are in the data:
- högskoleprovet (Swedish Scholastic Aptitude Test)
- kunskapsprov läkare (medical doctor test)
- kunskapsprov tandläkare (dentist test)
- kunskapsprov audionom
- kunskapsprov apotekare
- matematik och fysikprovet (math and highschool test)

```python
from datasets import load_dataset

ds = load_dataset("Ekgren/swedish_skolprov", "all")
```

## Description of the data

| field name | description |
|--------------------|------------------------------------------------------------------------------------------------|
| test_id | id or name of the test |
| section | name of section of the test |
| subsection | name of subsection of the test |
| question_id | numerical id for question related to test |
| question_resource | if it exists reading material related to question |
| question | question |
| option_a | answer alternative to question |
| option_b | answer alternative to question |
| option_c | answer alternative to question |
| option_d | answer alternative to question |
| option_e | answer alternative to question |
| system_prompt | system prompt for answering questions |
| prompt | pre-formatted prompt with background, question_resource, question and answer alternatives |
| answer | the correct answer to the question in the format of `a,b,c,d,e` or `A,B,C,D,E` |
| question_points | the points assigned to a question |
| uid | a unique id constructed from some combination of test_id, section, subsection and question_id |

## Small leaderboard
In the table below are the results on the tests for a small selection of models. The tests evaluate the ability to understand and answer questions in Swedish as well as specific knowledge with respect to subdomains.

| rank | model_name | **total** | hp | matematik_och_fysikprovet | läkare | tandläkare | audionom | apotekare | open_weights |
|------|-----------------------------------------|-----------|---------|---------------------------------|-------------|-----------------|---------------|----------------|--------------|
| 1. | openai/o3-mini | **0.85** | 1.9 | 0.77 | 0.89 | 0.80 | 0.82 | 0.96 | ❌ |
| 2. | anthropic/claude-3.7-sonnet | **0.84** | 2.0 | 0.62 | 0.88 | 0.87 | 0.85 | 1.00 | ❌ |
| 3. | openai/gpt-4o | **0.82** | 2.0 | 0.51 | 0.89 | 0.87 | 0.89 | 0.98 | ❌ |
| 4. | google/gemini-2.0-flash-001 | **0.82** | 2.0 | 0.63 | 0.87 | 0.81 | 0.81 | 0.96 | ❌ |
| 5. | meta-llama/llama-3.1-405b-instruct | **0.81** | 2.0 | 0.55 | 0.91 | 0.76 | 0.82 | 0.96 | ✅ |
| 6. | deepseek/deepseek-V3 | **0.80** | 2.0 | 0.60 | 0.83 | 0.71 | 0.89 | 0.94 | ✅ |
| 7. | x-ai/grok-2-1212 | **0.80** | 2.0 | 0.50 | 0.88 | 0.80 | 0.82 | 0.98 | ❌ |
| 8. | meta-llama/llama-3.3-70b-instruct | **0.78** | 1.9 | 0.53 | 0.82 | 0.77 | 0.82 | 0.94 | ✅ |
| 9. | qwen/qwen-max | **0.77** | 1.7 | 0.60 | 0.84 | 0.66 | 0.78 | 0.92 | ❌ |
| 10. | cohere/command-a | **0.72** | 1.8 | 0.48 | 0.78 | 0.70 | 0.72 | 0.89 | ✅ |
| 11. | mistralai/mistral-small-3.1-24b-instruct| **0.71** | 1.6 | 0.50 | 0.78 | 0.61 | 0.76 | 0.91 | ✅ |
| 12. | google/gemma-3-27b-it | **0.67** | 1.8 | 0.50 | 0.64 | 0.69 | 0.72 | 0.70 | ✅ |
| 13. | allenai/olmo-2-0325-32b-instruct | **0.57** | 1.3 | 0.36 | 0.63 | 0.50 | 0.53 | 0.77 | ✅ |
| 14. | meta-llama/llama-3.2-3b-instruct | **0.36** | 0.7 | 0.19 | 0.46 | 0.41 | 0.31 | 0.38 | ✅ |

## The tests
### Högskoleprovet
- https://www.studera.nu/hogskoleprov/om/forbereda/tidigare/

### Matematik- och fysikprovet
- https://www.kth.se/mattefysik/matematik-och-fysikprovet/matematik-och-fysikprovet-1.1233988

### Kunskapsprov läkare
- https://www.umu.se/utbildning/sok/kunskapsprov/kunskapsprov-for-lakare/
- https://www.umu.se/utbildning/sok/kunskapsprov/kunskapsprov-for-lakare/teoretiskt-delprov/

### Kunskapsprov tandläkare
- https://ki.se/dentmed/utbildning/kunskapsprov-for-tandlakare-med-utbildning-fran-lander-utanfor-euees

### Kunskapsprov farmaceuter
- https://www.uu.se/utbildning/anmalan-och-antagning/utlandsk-utbildning/arbeta-som-farmaceut-i-sverige/kunskapsprov
- https://www.uu.se/utbildning/anmalan-och-antagning/utlandsk-utbildning/arbeta-som-farmaceut-i-sverige/kunskapsprov/forberedelser-till-det-teoretiska-kunskapsprovet

### Kunskapsprov audionomer
- https://www.gu.se/studera/hitta-utbildning/kunskapsprov-audionomer