Upload README.md with huggingface_hub

README.md

@@ -58,6 +58,8 @@ Total: **60.25 minutes**.
 - **Phoneme coverage**: English 39 (100%), Telugu 44 (88%).
 - **Perceptual quality** (DNSMOS OVRL, published set): EN 3.09 (58% pass>3.0), TE 3.16 (81% pass>3.0). Filter `dnsmos_pass=True` for a stricter subset.
 - **Transcript–audio alignment** (MMS forced-align): median confidence EN 0.954, TE 0.937.
+- **Emotion-label agreement** (Krippendorff alpha): 0.4442 between the two LLM raters (0.4+ is the field norm). A 3-rater panel adding SER models drops near zero, since off-the-shelf SER clusters toward neutral and does not transfer to Telugu. Per-clip VAD (valence, arousal, dominance) is included.
+- **LLM-as-judge cross-check** (independent model, 499 clips): 75% of transcripts judged clean and 81% suitable to train on. Each clip also has a topic; the set is mostly storytelling (mythology, folk tales, audiobook fiction).
 
 See the project report (GitHub repo) for full methodology and figures.