Loading
Life sciences · Preprint
arXiv · September 9, 2026
Early or partial results. Treat as a signal, not a conclusion.
Orukeet is a modified automatic speech recognition model that replaces half of Parakeet's temporal filters with frozen Gabor kernels and retrains on multilingual data. On standard benchmarks (FLEURS, LibriSpeech), it reports consistent reductions in word error rate across multiple languages and test conditions, but the work is unreviewed and lacks statistical significance testing or held-out test validation.
Comparative technical evaluation on standard benchmarks. Multilingual speech audio from FLEURS dataset (25 languages) and LibriSpeech English test sets (test-clean and test-other).. Intervention: Orukeet: Parakeet encoder with half of temporal filters replaced by 12,288 frozen Gabor kernels; remaining parameters trained on multilingual and multi-accent data.. Compared with: Parakeet: original baseline model.. n = 20,146.
Pooled WER across 20,146 FLEURS recordings in 25 languages: 9.85% (Orukeet) vs. 11.01% (Parakeet), a 10.6% relative reduction LibriSpeech test-clean: 1.46% WER (Orukeet) vs. 1.53% (Parakeet) LibriSpeech test-other: 2.86% WER (Orukeet) vs. 3.14% (Parakeet)
Safety was not reported in the material analysed. Check the source before drawing any conclusion about harm.
The source did not state who this applies to in practice.
This is an unreviewed preprint describing a technical modification to an ASR model with reported improvements on multilingual benchmarks, but lacking statistical testing, confidence intervals, or peer review.
As stated by the source record.
Quoted from the source exactly as published.
Graded across the dimensions that decide whether you should act, each from what the source actually supports. There is no single score, and where a dimension was not assessed it says so.
Orukeet replaces half of an adapted Parakeet encoder's temporal filters with 12,288 fitted Gabor kernels, freezes these replacements, and trains the remaining parameters on multilingual and multi-accent data. Final adaptation and checkpoint selection use LibriSpeech test-other. Across 20,146 FLEURS recordings in 25 languages, pooled word error rate (WER) falls from Parakeet's 11.01% to Orukeet's 9.85%, a 10.6% relative reduction. Orukeet has lower WER on 23 of the 25 languages. Orukeet outperforms Parakeet on 61 out of 74 tested splits, including LibriSpeech test-clean (1.46% vs. 1.53% WER), test-other (2.86% vs. 3.14%), and FLEURS English (3.82% vs. 4.28%). All comparisons decode the same audio with matched NeMo settings. The fitted kernels are stored as ordinary convolution weights, retaining Parakeet's architecture and inference operators.
Taken from the source record, never inferred. Follow any of these and new work involving them reaches your briefing.