src/about.py

from dataclasses import dataclass
from enum import Enum

@dataclass
class Task:
    benchmark: str
    metric: str
    col_name: str
    type: str
    source: str


# Select your tasks here
# ---------------------------------------------------
class Tasks(Enum):
    # task_key in the json file, metric_key in the json file, name to display in the leaderboard 

    # base
    task0 = Task("arc_easy", "accuracy", "ARC-Easy", "base", "https://github.com/UKGovernmentBEIS/inspect_evals/tree/main/src/inspect_evals/arc")
    task1 = Task("arc_challenge", "accuracy", "ARC-Challenge", "base", "https://github.com/UKGovernmentBEIS/inspect_evals/tree/main/src/inspect_evals/arc")
    task2 = Task("drop", "mean", "DROP", "base", "https://github.com/UKGovernmentBEIS/inspect_evals/tree/main/src/inspect_evals/drop")
    task3 = Task("winogrande", "accuracy", "WinoGrande", "base", "https://github.com/UKGovernmentBEIS/inspect_evals/tree/main/src/inspect_evals/winogrande")
    task4 = Task("gsm8k", "accuracy", "GSM8K", "base", "https://github.com/UKGovernmentBEIS/inspect_evals/tree/main/src/inspect_evals/gsm8k")
    task5 = Task("hellaswag", "accuracy", "HellaSwag", "base", "https://github.com/UKGovernmentBEIS/inspect_evals/tree/main/src/inspect_evals/hellaswag")
    task6 = Task("humaneval", "mean", "HumanEval", "base", "https://github.com/UKGovernmentBEIS/inspect_evals/tree/main/src/inspect_evals/humaneval")
    task7 = Task("ifeval", "final_acc", "IFEval", "base", "https://github.com/UKGovernmentBEIS/inspect_evals/tree/main/src/inspect_evals/ifeval")
    task8 = Task("math", "accuracy", "MATH", "base", "https://github.com/UKGovernmentBEIS/inspect_evals/tree/main/src/inspect_evals/mathematics")
    task9 = Task("mmlu", "accuracy", "MMLU", "base", "https://github.com/UKGovernmentBEIS/inspect_evals/tree/main/src/inspect_evals/mmlu")
    task10 = Task("mmlu_pro", "accuracy", "MMLU-Pro", "base", "https://github.com/UKGovernmentBEIS/inspect_evals/tree/main/src/inspect_evals/mmlu_pro")
    task11 = Task("gpqa_diamond", "accuracy", "GPQA-Diamond", "base", "https://github.com/UKGovernmentBEIS/inspect_evals/tree/main/src/inspect_evals/gpqa")
    task12 = Task("mmmu_multiple_choice", "accuracy", "MMMU-Multiple-Choice", "base", "https://github.com/UKGovernmentBEIS/inspect_evals/tree/main/src/inspect_evals/mmmu")
    task13 = Task("mmmu_open", "accuracy", "MMMU-Open-Ended", "base", "https://github.com/UKGovernmentBEIS/inspect_evals/tree/main/src/inspect_evals/mmmu")

    # agentic
    task14 = Task("gaia", "accuracy", "GAIA", "agentic", "https://github.com/UKGovernmentBEIS/inspect_evals/tree/main/src/inspect_evals/gaia")
    task15 = Task("gdm_intercode_ctf", "accuracy", "GDM-InterCode-CTF", "agentic", "https://github.com/UKGovernmentBEIS/inspect_evals/tree/main/src/inspect_evals/gdm_capabilities/intercode_ctf")
    task16 = Task("gdm_in_house_ctf", "accuracy", "GDM-In-House-CTF", "agentic", "https://github.com/UKGovernmentBEIS/inspect_evals/tree/main/src/inspect_evals/gdm_capabilities/in_house_ctf")
    task17 = Task("agentharm", "avg_score", "AgentHarm", "agentic", "https://github.com/UKGovernmentBEIS/inspect_evals/tree/main/src/inspect_evals/agentharm")
    task18 = Task("agentharm_benign", "avg_score", "AgentHarm-Benign", "agentic", "https://github.com/UKGovernmentBEIS/inspect_evals/tree/main/src/inspect_evals/agentharm")
    task19 = Task("swe_bench", "mean", "SWE-Bench", "agentic", "https://github.com/UKGovernmentBEIS/inspect_evals/tree/main/src/inspect_evals/swe_bench")


NUM_FEWSHOT = 0 # Change with your few shot
# ---------------------------------------------------



# Your leaderboard name
TITLE = """<h1 align="center" id="space-title">State of Evaluation Leaderboard</h1>"""

SINGLE_TURN_TASK_NAMES = ", ".join([f"[{task.value.col_name}]({task.value.source})" for task in Tasks if task.value.type == "base"])
AGENTIC_TASK_NAMES = ", ".join([f"[{task.value.col_name}]({task.value.source})" for task in Tasks if task.value.type == "agentic"])

# What does your leaderboard evaluate?
INTRODUCTION_TEXT = f"""
Powered by **Inspect** and **Inspect Evals**, the **Vector State of Evaluation Leaderboard** presents an objective evaluation of leading frontier models across a comprehensive suite of benchmarks. Go beyond the summary metrics: click through to interactive reporting for each model and benchmark to explore sample-level performance and detailed traces."""

# Which evaluations are you running? how can people reproduce what you have?
ABOUT_TEXT = f"""

## Vector Institute
The **Vector Institute** is dedicated to advancing the fields of artificial intelligence and machine learning through cutting-edge research, collaborative projects, and open-source contributions. Our mission is to drive excellence and innovation in AI, fostering a vibrant community of researchers, developers, and industry partners.

## 🎯 Benchmarks

This leaderboard showcases performance across a comprehensive suite of benchmarks, designed to rigorously evaluate different aspects of AI model capabilities. Let's explore the benchmarks we use:

### Inspect Evals

This leaderboard leverages [Inspect Evals](https://ukgovernmentbeis.github.io/inspect_evals/) to power evaluation. Inspect Evals is an open-source repository built upon the Inspect AI framework. Developed in collaboration between the Vector Institute, Arcadia Impact and the UK AI Safety Institute, Inspect Evals provides a comprehensive suite of high-quality benchmarks spanning diverse domains like coding, mathematics, cybersecurity, reasoning, and general knowledge.

#### Transparent and Detailed Insights

All evaluations presented on this leaderboard are run using Inspect Evals. To facilitate in-depth analysis and promote transparency, we provide [Inspect Logs](https://inspect.ai-safety-institute.org.uk/log-viewer.html) for every benchmark run. These logs offer sample and trace level reporting, allowing the community to explore the granular details of model performance. 

### ⚙️ Base Benchmarks

These benchmarks assess fundamental reasoning and knowledge capabilities of models.

<div class="benchmark-table-container">

| Benchmark           | Description                                                                      | Key Skills                                   |
|--------------------|----------------------------------------------------------------------------------|-----------------------------------------------|
| **ARC-Easy** / **ARC-Challenge** | Multiple-choice science questions measuring scientific & commonsense reasoning. | Scientific Reasoning, Commonsense Reasoning |
| **DROP**             | Reading comprehension benchmark emphasizing discrete reasoning steps.             | Reading Comprehension, Discrete Reasoning    |
| **WinoGrande**        | Commonsense reasoning challenge focused on co-reference resolution.             | Commonsense Reasoning, Co-reference Resolution|
| **GSM8K**             | Grade-school math word problems testing arithmetic & multi-step reasoning.         | Arithmetic Reasoning, Multi-step Reasoning    |
| **HellaSwag**        | Commonsense inference task centered on action completion.                        | Commonsense Inference, Action Completion       |
| **HumanEval**         | Evaluates code generation and reasoning in a programming context.                | Code Generation, Code Reasoning               |
| **IFEval**            | Specialized benchmark for incremental formal reasoning.                          | Incremental Formal Reasoning                  |
| **MATH**              | High school-level math questions requiring detailed solutions.                  | Mathematical Reasoning, Problem Solving       |
| **MMLU** / **MMLU-Pro**| Multi-subject multiple-choice tests of advanced knowledge.                     | Broad Knowledge, Multi-domain Understanding   |
| **GPQA-Diamond**      | Question-answering benchmark assessing deeper reasoning & knowledge linking.      | Deep Reasoning, Knowledge Linking             |
| **MMMU** (Multi-Choice / Open-Ended) | Multilingual & multi-domain tasks testing structured & open responses.   | Multilingual Understanding, Open-form Response |

</div>

### 🚀 Agentic Benchmarks

These benchmarks go beyond basic reasoning and evaluate more advanced, autonomous, or "agentic" capabilities of models, such as planning and interaction.

<div class="benchmark-table-container">

| Benchmark              | Description                                                                 | Key Skills                                      |
|-----------------------|-----------------------------------------------------------------------------|-------------------------------------------------|
| **GAIA**                | Evaluates autonomous reasoning, planning, problem-solving, & multi-turn interactions. | Autonomous Reasoning, Planning, Problem Solving  |
| [**InterCode-CTF**](https://ukgovernmentbeis.github.io/inspect_evals/evals/cybersecurity/in_house_ctf/)   | Capture-the-flag challenge focused on code interpretation & debugging.       | Code Interpretation, Generative Debugging       |
| **GDM-In-House-CTF**    | Capture-the-flag challenge testing web application security skills.         | Web Security Skills, Security Awareness         |
| **AgentHarm** / **AgentHarm-Benign** | Measures harmfulness of LLM agents (and benign behavior baseline).   | Harmfulness Detection, Agent Safety             |
| **SWE-Bench**           | Tests AI agent ability to solve software engineering tasks.                 | Software Engineering Tasks, Bug Fixing          |

</div>
"""

REPRODUCIBILITY_TEXT = """
## Reproduce and Extend the Leaderboard

### 1) Make sure you can load your model and tokenizer using AutoClasses:
```python
from transformers import AutoConfig, AutoModel, AutoTokenizer
config = AutoConfig.from_pretrained("your model name", revision=revision)
model = AutoModel.from_pretrained("your model name", revision=revision)
tokenizer = AutoTokenizer.from_pretrained("your model name", revision=revision)
```
If this step fails, follow the error messages to debug your model before submitting it. It's likely your model has been improperly uploaded.

Note: make sure your model is public!
Note: if your model needs `use_remote_code=True`, we do not support this option yet but we are working on adding it, stay posted!

### 2) Convert your model weights to [safetensors](https://huggingface.co/docs/safetensors/index)
It's a new format for storing weights which is safer and faster to load and use. It will also allow us to add the number of parameters of your model to the `Extended Viewer`!

### 3) Make sure your model has an open license!
This is a leaderboard for Open LLMs, and we'd love for as many people as possible to know they can use your model 🤗

### 4) Fill up your model card
When we add extra information about models to the leaderboard, it will be automatically taken from the model card

## In case of model failure
If your model is displayed in the `FAILED` category, its execution stopped.
Make sure you have followed the above steps first.
If everything is done, check you can launch the EleutherAIHarness on your model locally, using the above command without modifications (you can add `--limit` to limit the number of examples per task).
"""