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

# Geo3K

> Single-turn VLM geometry agent on the Geometry3K dataset

A vision-language agent that solves geometry problems with diagram images via the **AgentFlow protocol**. Trains on the [Geometry3K](https://huggingface.co/datasets/hiyouga/geometry3k) dataset.

## Pattern

| Aspect       | Value                                                                  |
| ------------ | ---------------------------------------------------------------------- |
| Loop shape   | Single-turn (one VLM call per task)                                    |
| Tools        | None — answer is parsed out of the response                            |
| Inputs       | Multimodal — text question + base64-encoded diagram image              |
| Termination  | Single LLM call returns; evaluator extracts `\boxed{…}`                |
| Reward shape | `1.0` if boxed answer matches ground truth (symbolic math), else `0.0` |

## Architecture

```mermaid theme={null}
flowchart TD
    A["AgentFlow.run(task, config)"] --> S["Solver"]
    S --> C["OpenAI(base_url=config.base_url).chat.completions.create(...)"]
    C --> M["messages=[system_prompt, image+question]"]
    C --> R["response"]
    R --> T["Trajectory(name="solver", steps=[Step(action=response)])"]
    T --> E["Episode(trajectories=[solver], artifacts={"answer": response})"]
    A --> E
```

The cookbook demonstrates the **multimodal content-block pattern** in an AgentFlow — the `messages` list contains a `{"type": "image_url", "image_url": {"url": f"data:image/png;base64,…"}}` block alongside the text content.

## Install

```bash theme={null}
uv pip install -e ".[tinker]"                    # rllm + tinker backend
uv pip install --no-deps -e cookbooks/geo3k      # this cookbook
rllm agent list                                  # should show "geo3k"
```

## Dataset

```bash theme={null}
rllm dataset pull geo3k
```

## Eval

```bash theme={null}
rllm eval geo3k \
    --agent geo3k \
    --evaluator geo3k \
    --model Qwen/Qwen3-VL-8B-Instruct \
    --base-url http://localhost:8000/v1 \
    --max-examples 20
```

## Training

```bash theme={null}
# Tinker (single-machine LoRA)
bash cookbooks/geo3k/train_tinker.sh

# Verl (distributed GPU)
bash cookbooks/geo3k/train_verl.sh
```

## Files

| File                                  | Description                                              |
| ------------------------------------- | -------------------------------------------------------- |
| `geo3k_flow.py`                       | Single-turn VLM AgentFlow with multimodal content blocks |
| `evaluator.py`                        | `\boxed{}` extraction + symbolic math grading            |
| `train.py` + `train_{tinker,verl}.sh` | Hydra entry points                                       |
| `pyproject.toml`                      | Plugin entry-point declarations                          |
| `test.py`                             | Unit tests                                               |

## On GitHub

<Card title="cookbooks/geo3k" icon="github" href="https://github.com/rllm-org/rllm/tree/main/cookbooks/geo3k">
  Full source, README, and runnable launch scripts
</Card>