---
title: Helium Benchmark Lab
emoji: 📈
colorFrom: blue
colorTo: green
sdk: gradio
sdk_version: 6.20.0
app_file: app.py
pinned: false
license: mit
---

# Helium Benchmark Lab

Try two small, inspectable evaluations:

- **Market Resolution:** sample frozen option-chain questions and compare with the 58% frontier score.
- **Cue swap:** hold the task fixed, change one political cue, and compare the responses.

Set provider keys in Space secrets. Model calls use [LiteLLM](https://docs.litellm.ai/docs/providers) strings. The demo card works without a key.

- [Market Resolution](https://huggingface.co/datasets/HeliumTrades/helium-market-resolution-benchmark)
- [Model Worldview](https://huggingface.co/datasets/HeliumTrades/helium-model-worldview-benchmark)
- [Landing](https://heliumtrades.com/benchmarks/)