Argon Model Identifies a Russian Accent From a 3-Minute Clip With Phoneme-Level Analysis
soumitrashukla9 · x · 2026-11-10
Ben Golub tested the Argon model with a 3-minute audio clip (4,620 tokens), asking it to identify the speaker's non-English first language by ear alone.
The model judged the speaker's English as extremely fluent General American, consistent with an early bilingual who moved to the US in childhood — but pinned the accent as Eastern European/Slavic, most likely Russian, citing:
- Velarized "dark l" in onset positions ("legible," "line," "location," "lies")
- Retracted /ɪ/ shifting toward the Slavic [ɨ], as in "within," "picture," "typical"
- Tense, monophthongal back vowels with far-back rounded /u:/ in "groups" and "two"
- Unreduced vowels in unstressed syllables
The analysis cited exact timestamps and phoneme-level detail, showcasing rare fine-grained phonetic listening ability.
More from Models
- Claude Fable 5.2 video demo wows on pacing, but AI-sounding music persists — jiayuan_jy · 2026-11-10
- Gemini Notebooks keeps pulling context from old conversations, user reports — AnonymousDork929 · 2026-11-10
- User complains model takes four minutes to compare two 16-token completions — cephaloform · 2026-11-10
- DavidAU releases Qwen3.8-40B-Ultimate-Heretic-Uncensored-Turbo-Brilliance-50x-AGI-NEO-GGUF merge — mp3m4k3r · 2026-11-10
- Uncensored Gemma 4 31B 'heretic' build hits Hugging Face, 11.9GB quantized — alexcovo_eth · 2026-11-10
- Same demo, $1.4 on DeepSeek vs $40 on Fable-5.2: is it worth it? — karminski3 · 2026-11-10