Update v2: README.md

README.md

@@ -24,7 +24,7 @@ model-index:
     metrics:
     - name: Test WER
       type: wer
-      value: 9.47
+      value: 7.94
   - task:
       name: Automatic Speech Recognition
       type: automatic-speech-recognition
@@ -37,7 +37,7 @@ model-index:
     metrics:
     - name: Test WER
       type: wer
-      value: 10.61
+      value: 8.86
   - task:
       name: Automatic Speech Recognition
       type: automatic-speech-recognition
@@ -50,7 +50,7 @@ model-index:
     metrics:
     - name: Test WER
       type: wer
-      value: 8.02
+      value: 8.16
   - task:
       name: Automatic Speech Recognition
       type: automatic-speech-recognition
@@ -64,7 +64,7 @@ model-index:
     metrics:
     - name: Test WER
       type: wer
-      value: 1.43
+      value: 1.13
   - task:
       name: Automatic Speech Recognition
       type: automatic-speech-recognition
@@ -78,7 +78,7 @@ model-index:
     metrics:
     - name: Test WER
       type: wer
-      value: 2.82
+      value: 2.66
   - task:
       name: Automatic Speech Recognition
       type: automatic-speech-recognition
@@ -92,7 +92,7 @@ model-index:
     metrics:
     - name: Test WER
       type: wer
-      value: 2.23
+      value: 2.15
   - task:
       name: Automatic Speech Recognition
       type: automatic-speech-recognition
@@ -106,7 +106,7 @@ model-index:
     metrics:
     - name: Test WER
       type: wer
-      value: 2.32
+      value: 2.95
   - task:
       name: Automatic Speech Recognition
       type: automatic-speech-recognition
@@ -120,13 +120,32 @@ model-index:
     metrics:
     - name: Test WER
       type: wer
-      value: 5.99
+      value: 6.07
 ---
 
 # Higgs Audio v3 STT
 
 A speech-to-text model combining a Whisper-Large-v3 encoder with a Qwen3 decoder (2.68B total parameters).
 
+## Update (June 2026)
+
+This repository now hosts an updated checkpoint (v2 weights). Changes:
+
+- Fine-tuning data refreshed: public train splits of AMI (IHM), VoxPopuli (en),
+  SPGISpeech, LibriSpeech, TED-LIUM, GigaSpeech, plus the public Earnings22
+  train split (`sanchit-gandhi/earnings22_split`) with all rows from source
+  recordings that appear in the ESB/Open-ASR test sets excluded.
+- `transcribe.py` adds a phrase-level repetition-loop collapse
+  (`ngram_loop_fix.py`) alongside the existing word-repetition cap; both are
+  deterministic and applied uniformly to every dataset.
+- The reported WERs below were measured on the complete ESB test suite
+  (82,185 utterances) with the Open-ASR-Leaderboard scoring pipeline
+  (bundled normalizer + compound-pair normalization), batched inference
+  (batch 32, VoxPopuli 16). Earlier reported figures came from a previous
+  checkpoint and evaluation setup and are superseded.
+
+The previous weights remain available via the git revision history.
+
 ## Usage
 
 **Important:** This model uses a custom architecture. You must pass `trust_remote_code=True` when loading.