George Saon

The architecture, training methodology and inference speedups of Granite 5.0 Turbo CTC result in a model that is on the speed-accuracy Pareto frontier of the Open ASR leaderboard for English short-form ASR while being twice as fast as the fastest competitor.

Forget phoneme sequences and G2P systems: this work shows you can boost ASR accuracy for rare words by cleverly leveraging acoustic cues from common words with similar sounds.

Speaker-attributed ASR gets a serious boost from jointly training speaker cluster tags within a speech-aware LLM, outperforming traditional pipelines.

Ditch slow, sequential decoding: NLE achieves 27x speedup over autoregressive ASR by using a non-autoregressive, LLM-based transcript editing approach.