---
viewer: false
tags:
- benchpress
- dataset-composition
---

# coding-mix

Fixed evaluation-set composition built with BenchPress. This repo stores
only the recipe (`manifest.json`) — source data is streamed at load time.

**Abilities:** algorithmic_decomposition, rule_constrained_reasoning

## Sources

| benchmark | source repo | config | split | n_samples | license | gated |
|---|---|---|---|---|---|---|
| HumanEval | [evalplus/humanevalplus](https://huggingface.co/datasets/evalplus/humanevalplus) | default | test | 100 | apache-2.0 | no |
| MBPP | [google-research-datasets/mbpp](https://huggingface.co/datasets/google-research-datasets/mbpp) | full | test | 150 | cc-by-4.0 | no |
| LiveCodeBench | [livecodebench/code_generation](https://huggingface.co/datasets/livecodebench/code_generation) | - | test | 100 | cc | no |

## Reference scores (expected on this composition)

| model | expected score |
|---|---|
| Claude-3.5-Sonnet | 0.7997 |
| Claude-Sonnet-4 | 0.8314 |
| DeepSeek-v3 | 0.604 |
| GPT-4o | 0.7206 |
| GPT-oss-20b | 0.8069 |
| Gemma-3-4B | 0.3671 |
| Llama-3.1-8B | 0.5489 |
| Phi-4-Mini | 0.5504 |
| Qwen2.5-72B | 0.784 |
| Qwen3-235B | 0.8086 |

## Usage

```python
from benchpress_hub import load_composition

ds = load_composition("seonghyeon0408/coding-mix")
```

Sources pinned to commit SHAs; loading is deterministic (fixed items, seeded order).