README.md

metadata

title: A2R-30B-A3B Speaker-Attributed Reasoning
emoji: 🗣️
colorFrom: purple
colorTo: green
sdk: gradio
sdk_version: 6.26.0
app_file: app.py
python_version: '3.12'
short_description: Ask who said what in a multi-speaker recording
startup_duration_timeout: 1h

A2R-30B-A3B — who said what?

Demo of PleasedPenguin/A2R-30B-A3B, the model from "HEAR Who Said What: Unlocking Speaker-Attributed Reasoning via Counterfactual Voice Grounding" (paper, code, project page).

A2R is Qwen/Qwen3-Omni-30B-A3B-Instruct post-trained with GRPO on counterfactual audio containing speaker-level hard negatives, so it grounds answers in vocal identity rather than in the transcript.

Prompt formatting, system prompt and sampling defaults (temperature=0.6, top_p=0.95) follow the authors' evaluation harness.

Example audio attribution

The bundled example clips are excerpts of the AMI Meeting Corpus, licensed CC BY 4.0 — one of the source corpora behind the HEAR benchmark. The HEAR benchmark audio itself is not redistributable and is therefore not bundled here.