---
title: decision-index-leaderboard
canonical_url: "https://www.modelscope.cn/datasets/multimodalart/decision-index-leaderboard"
md_url: "https://www.modelscope.cn/datasets/multimodalart/decision-index-leaderboard.md"
repository: multimodalart/decision-index-leaderboard
last_updated: 2026-10-30
license: cc-by-4.0
storage_size: "142 KB"
downloads: 73
stars: 0
---

# decision-index-leaderboard

> decision-index-leaderboard - multimodalart 在 ModelScope 开源的数据集。The text board of the Jev Decision Index (Decision Index 0.4.1): models that read a state and a typed question and return a decision. Vision board: multimodalart/decision-index-vision-leaderboard.

multimodalart/decision-index-leaderboard 是 ModelScope 魔搭社区上的数据集，存储大小 142 KB，采用 cc-by-4.0 许可。

- **Repository**: multimodalart/decision-index-leaderboard
- **License**: cc-by-4.0
- **Storage size**: 142 KB
- **Downloads**: 73
- **Stars**: 0
- **Last updated**: 2026-10-30

Source: https://www.modelscope.cn/datasets/multimodalart/decision-index-leaderboard

---

# Decision Index

The text board of the [Jev Decision Index](https://huggingface.co/spaces/multimodalart/jev-decision-index) (Decision Index 0.4.1): models that read a state and a typed question and return a decision. Vision board: [multimodalart/decision-index-vision-leaderboard](https://huggingface.co/datasets/multimodalart/decision-index-vision-leaderboard).

Tasks (each is its own leaderboard):

- `decision_index`: the Decision Index (Full) = 0.6 × same-skills + 0.4 × new-domains, both equated on 21 stock base models. Scores within 0.9 points share a rank.
- `same_skills`: private sets that test the same skills as the public suite.
- `new_domains`: private sets from domains outside the public suite.
- `public`: the 42-benchmark public suite.

The private and new-domain sets are not released; only aggregate scores are published. The evaluation kit and submission process: [apolinario/decision-index](https://github.com/apolinario/decision-index). Methodology: https://huggingface.co/spaces/multimodalart/jev-decision-index.

This repo mirrors the board: `leaderboard` has one row per model (248 entries), `benchmarks` the per-benchmark public scores. It is regenerated whenever the board is updated.

## Top 10

| Rank | Model | Full |
|---|---|---|
| 1 | KnowLine 31B Gen1 | 71.86 |
| 2 | Fastino GLiDE 1.2 | 71.52 |
| 3 | Sieve 31B v3 | 71.10 |
| 4 | Turion Jev 31B | 70.47 |
| 5 | vLLM-SR d3-ultra (Decision 3.0, 27B) | 70.18 |
| 6 | Jeffan-31b-1710 | 69.75 |
| 7 | H2O Lightning 31B v1.1 | 68.64 |
| 8 | WeMM-JEV-27B | 68.49 |
| 9 | baselabs mammoth-1 (27B) | 68.44 |
| 10 | Ainvest-Decision-27B v1.0 | 68.23 |

Generated 2026-10-29T14:04:16Z.