Atlas

Benchmarks

← All benchmarks

TED-LIUM 3 (ESB test)

Multimodal · 2022-10-24

The TED-LIUM release 3 test split as packaged by the ESB benchmark and used by the Hugging Face Open ASR Leaderboard: 1,155 utterances of English TED talk speech. Scores are corpus word error rate in percent after Whisper English normalization; lower is better.

Top models (lower is better)

ModelScore
LFM2.5-Audio-1.5B3.5
LFM2-Audio-1.5B3.6
Loading Atlas data…