Update README.md

README.md

@@ -46,35 +46,33 @@ pipeline_tag: automatic-speech-recognition
 # Qwen3-ASR-0.6B — INT4 AWQ + Quantization-Aware Distillation
 
 This is a compressed and distillation-refined version of [Qwen/Qwen3-ASR-0.6B](https://huggingface.co/Qwen/Qwen3-ASR-0.6B).  
-The model applies **INT4 AWQ Activation-aware Weight Quantization** post-training quantization followed by **Quantization-Aware Distillation QAD** to recover accuracy lost during aggressive 4-bit compression — while preserving low-latency, low-memory inference.
+The model applies **INT4 AWQ (Activation-aware Weight Quantization)** post-training quantization followed by **Quantization-Aware Distillation (QAD)** to recover accuracy lost during aggressive 4-bit compression — while preserving low-latency, low-memory inference.
 
 ---
 
 ## Method Overview
 
-### Stage 1 — INT4 AWQ PTQ
-The base FP16 model is quantized to INT4 using the **AWQ awq_full** algorithm via NVIDIA ModelOpt. To preserve acoustic feature extraction and final token prediction integrity, the audio_tower and lm_head are explicitly excluded from quantization and kept in FP16. Calibration was performed dynamically across English, Chinese, and 28 other languages to ensure a balanced quantization scale mapping.
+### Stage 1 — INT4 AWQ (PTQ)
+The base FP16 model is quantized to INT4 using the **AWQ (`awq_full`)** algorithm via [NVIDIA ModelOpt](https://github.com/NVIDIA/TensorRT-Model-Optimizer). To preserve acoustic feature extraction and final token prediction integrity, the `audio_tower` and `lm_head` are explicitly excluded from quantization and kept in FP16. Calibration was performed dynamically across English, Chinese, and 28 other languages to ensure a balanced quantization scale mapping.
 
-### Stage 2 — Quantization-Aware Distillation QAD
+### Stage 2 — Quantization-Aware Distillation (QAD)
 To close the accuracy gap introduced by heavy 4-bit quantization, we apply a **knowledge distillation** fine-tuning stage where:
 
-- **Teacher**: Qwen3-ASR-1.7B FP16 model frozen
-- **Student**: the INT4-quantized 0.6B model trainable QKV/MLP quantized weights — audio encoder and LM head are frozen
+- **Teacher**: [Qwen3-ASR-1.7B](https://huggingface.co/Qwen/Qwen3-ASR-1.7B) FP16 model (frozen)
+- **Student**: the INT4-quantized 0.6B model (trainable QKV/MLP quantized weights — audio encoder and LM head are frozen)
 - **Data**: unlabeled speech data spanning **30 languages**, with pseudo-labels generated by the 1.7B teacher model
-- **Loss**: a combination of KL-divergence distillation loss alpha_kd = 0.5 and cross-entropy loss
+- **Loss**: a combination of KL-divergence distillation loss (`alpha_kd = 0.5`) and cross-entropy loss
 - **Optimizer**: AdamW with cosine decay learning rate schedule
 
-The QAD stage teaches the quantized student to match the teacher's output distribution on diverse real-world speech, without requiring any manual transcription labels.
-
 ---
 
-## Benchmark Results Trilingual Evaluation
+## Benchmark Results (Trilingual Evaluation)
 
-| Model | VIVOS Vietnamese WER ↓ | LibriSpeech English WER ↓ | Chinese Fleurs CER ↓ |
+| Model | [VIVOS](https://ailab.hcmus.edu.vn/vivos) (Viet) WER ↓ | [LibriSpeech](https://www.openslr.org/12) (Eng) WER ↓ | [Chinese Fleurs](https://huggingface.co/datasets/google/fleurs) CER ↓ |
 |---|---|---|---|
-| Teacher 1.7B FP16 base | 7.24% | 2.32% | 7.12% |
-| INT4 AWQ pre-QAD | 14.34% | 3.47% | 8.16% |
-| **INT4 + QAD this model** | **12.81%** | **3.41%** | **8.10%** |
+| Teacher 1.7B (FP16 base) | 7.24% | 2.32% | 7.12% |
+| INT4 AWQ (pre-QAD) | 14.34% | 3.47% | 8.16% |
+| **INT4 + QAD (this model)** | **12.81%** | **3.41%** | **8.10%** |
 
 > QAD successfully recovers **~1.53% absolute WER** in Vietnamese and stabilizes English/Chinese performance against aggressive 4-bit compression degradation, with no additional labeled data required.
 
@@ -122,13 +120,13 @@ print(results[0].text)
 
 | Property | Value |
 |---|---|
-| Base model | Qwen/Qwen3-ASR-0.6B |
+| Base model | [Qwen/Qwen3-ASR-0.6B](https://huggingface.co/Qwen/Qwen3-ASR-0.6B) |
 | Parameters | ~0.6B |
-| Quantization | INT4 AWQ awq_full via NVIDIA ModelOpt |
-| Audio encoder | Frozen FP16, not quantized |
-| LM head | Frozen FP16, not quantized |
+| Quantization | INT4 AWQ (`awq_full` via [NVIDIA ModelOpt](https://github.com/NVIDIA/TensorRT-Model-Optimizer)) |
+| Audio encoder | Frozen (FP16, not quantized) |
+| LM head | Frozen (FP16, not quantized) |
 | Quantized scope | Transformer decoder layers |
-| Distillation data | Multilingual unlabeled speech 30 languages |
+| Distillation data | Multilingual unlabeled speech (30 languages) |
 | License | Apache 2.0 |
 
 ---
@@ -150,5 +148,5 @@ If you use this model, please also cite the original Qwen3-ASR work:
 
 ## Acknowledgements
 
-- Qwen Team for the base ASR model
-- NVIDIA ModelOpt for quantization tooling
+- [Qwen Team](https://huggingface.co/Qwen) for the base ASR model
+- [NVIDIA ModelOpt](https://github.com/NVIDIA/TensorRT-Model-Optimizer) for quantization tooling