---
title: AudioRubrics
emoji: 🎧
colorFrom: gray
colorTo: pink
sdk: gradio
sdk_version: 6.24.0
app_file: app.py
short_description: Audio reasoning with evolving rubric rewards
python_version: '3.12'
startup_duration_timeout: 1h
---

# AudioRubrics

Interactive demo for [Reinforcement Learning with Evolving Rubrics as Rewards for Audio Reasoning](https://huggingface.co/papers/2608.02831).

Upload an audio clip, ask a question about what you hear, and the model will reason step-by-step before answering.

**Model:** [umd-zhou-lab/AudioRubrics](https://huggingface.co/umd-zhou-lab/AudioRubrics) — a Qwen2.5-Omni-7B model fine-tuned with GRPO using self-evolving, audio-grounded rubric rewards.

## Example audio credits

The example clips in `examples/` are real recordings sourced from Wikimedia Commons (resampled to 16 kHz mono and trimmed):

- `bird_chirp.wav` — "Black-capped Chickadee 201947598" (CC0 / public domain).
- `metro_sound.wav` — "59s Metro Lisboa" by OsvaldoGago (public domain).
- `alarm_sound.wav` — "AmbulanceSound1" (CC BY-SA 4.0).