---
license: apache-2.0
base_model: Qwen/Qwen3-ASR-1.7B-hf
tags:
- automatic-speech-recognition
- far-field
- noise-robust
- knowledge-distillation
language:
- en
pipeline_tag: automatic-speech-recognition
---

# tiro-qwen3-asr-1.7b-v2

Tiro is a far-field fine-tune of
[Qwen3-ASR-1.7B](https://huggingface.co/Qwen/Qwen3-ASR-1.7B-hf), with better
performance on reverberant and noisy speech, scaling as conditions get harder, while
maintaining general ASR capabilities. See benchmarks below for more details.

Currently **#1 on the [FFASR leaderboard](https://huggingface.co/spaces/treble-technologies/ffasr)**.

English only, bf16. Same architecture and parameter count as the base.

## Benchmarks

Lower is better. ✅ marks the better result.

### [FFASR Leaderboard](https://huggingface.co/spaces/treble-technologies/ffasr)

| Condition | **Base (WER)** | **Tiro (WER)** |
|---|---:|---:|
| **Average** | 13.41 | **12.01** ✅ |
| Near field speech | **3.76** ✅ | 3.80 |
| High SNR | 6.74 | **6.22** ✅ |
| Mid SNR | 13.89 | **12.00** ✅ |
| Low SNR | 29.26 | **26.02** ✅ |

### Other benchmarks

| Benchmark | Condition | **Base (WER)** | **Tiro (WER)** |
|---|---|---:|---:|
| [VOiCES](https://iqtlabs.github.io/voices/) | real rooms, replayed speech | 8.88 | **8.30** ✅ |
| [CHiME-4](https://www.chimechallenge.org/challenges/chime4/) | real noisy environments | 4.71 | **4.19** ✅ |
| CHiME-4 | simulated arm | 7.00 | **6.63** ✅ |
| [VITW-Bench](https://github.com/xzf-thu/Voices-in-the-Wild-Bench) | 8 perturbations, English | 7.09 | **6.40** ✅ |
| [NOIZEUS](https://ecs.utdallas.edu/loizou/speech/noizeus/) | additive noise, 0-15 dB | 9.23 | **7.85** ✅ |
| [Treble10-Speech](https://huggingface.co/datasets/treble-technologies/Treble10-Speech) | reverberation only | 3.29 | **3.19** ✅ |
| [AMI](https://huggingface.co/datasets/edinburghcstr/ami) | distant mic (SDM) | 19.73 | **19.33** ✅ |
| AMI | headset control (IHM) | **8.24** ✅ | 8.28 |
| [Open ASR Leaderboard](https://huggingface.co/spaces/hf-audio/open_asr_leaderboard) | mean of 8 public English splits | 4.311 | 4.310 |

<details>
<summary>Per-condition breakdown</summary>

| Benchmark | Condition | **Base (WER)** | **Tiro (WER)** |
|---|---|---:|---:|
| VOiCES | near microphone | 3.09 | **2.85** ✅ |
| VOiCES | far microphone | 14.68 | **13.76** ✅ |
| VOiCES | no distractor | 3.27 | **3.07** ✅ |
| VOiCES | music | 8.06 | **7.08** ✅ |
| VOiCES | television | 8.88 | **8.29** ✅ |
| VOiCES | babble | 15.33 | **14.77** ✅ |
| CHiME-4 real | bus | 5.98 | **5.42** ✅ |
| CHiME-4 real | cafe | 4.84 | **4.09** ✅ |
| CHiME-4 real | pedestrian | 4.18 | **3.89** ✅ |
| CHiME-4 real | street | 3.82 | **3.35** ✅ |
| NOIZEUS | 0 dB SNR | 23.24 | **20.76** ✅ |
| NOIZEUS | 5 dB SNR | 8.42 | **6.51** ✅ |
| NOIZEUS | 10 dB SNR | 3.20 | **2.74** ✅ |
| NOIZEUS | 15 dB SNR | 2.07 | **1.39** ✅ |
| VITW-Bench | real recordings | 5.30 | **4.84** ✅ |
| VITW-Bench | simulated | 7.80 | **7.03** ✅ |

</details>

<details>
<summary>Open ASR Leaderboard</summary>

Base column is the published Qwen3-ASR-1.7B result on the 8 public English
splits. The leaderboard's headline average also includes two private sets that
are not publicly runnable, so the figure shown there is higher than this one.

| Split | **Base (WER)** | **Tiro (WER)** |
|---|---:|---:|
| LibriSpeech test-clean | 1.26 | **1.25** ✅ |
| LibriSpeech test-other | **2.94** ✅ | 2.95 |
| VoxPopuli | 2.85 | **2.62** ✅ |
| AMI | **8.31** ✅ | 8.45 |
| GigaSpeech | **7.22** ✅ | 7.23 |
| Earnings22 | 5.84 | **5.70** ✅ |
| SPGISpeech | **2.57** ✅ | 2.77 |
| Monsoon | **3.50** ✅ | 3.51 |
| **mean** | 4.311 | 4.310 |

</details>

## Usage

Identical to the base model.

```python
from transformers import AutoProcessor, Qwen3ASRForConditionalGeneration

model_id = "rohansheth/tiro-qwen3-asr-1.7b-v2"
processor = AutoProcessor.from_pretrained(model_id)
model = Qwen3ASRForConditionalGeneration.from_pretrained(model_id, device_map="auto")
```

## Citations

Trained by KL distillation from the frozen base model on noise-free audio,
following:

```bibtex
@inproceedings{li2017large,
  title     = {Large-Scale Domain Adaptation via Teacher-Student Learning},
  author    = {Li, Jinyu and Seltzer, Michael L. and Wang, Xi and Zhao, Rui and Gong, Yifan},
  booktitle = {Interspeech},
  year      = {2017},
  eprint    = {1708.05466}
}

@inproceedings{mosner2019improving,
  title     = {Improving Noise Robustness of Automatic Speech Recognition via
               Parallel Data and Teacher-Student Learning},
  author    = {Mo{\v{s}}ner, Ladislav and Wu, Minhua and Raju, Anirudh and
               Parthasarathi, Sree Hari Krishnan and Kumatani, Kenichi and
               Sundaram, Shiva and Maas, Roland and Hoffmeister, Bj{\"o}rn},
  booktitle = {ICASSP},
  year      = {2019},
  eprint    = {1901.02348}
}
```

Base model:

```bibtex
@misc{qwen3asr,
  title  = {Qwen3-ASR},
  author = {Qwen Team},
  year   = {2026},
  url    = {https://huggingface.co/Qwen/Qwen3-ASR-1.7B-hf}
}
```