# Hypothesis-mode word alignment on Buckeye 5h (test) Source: [Hypothesis-mode word alignment on Buckeye 5h (test)]() Collection index: [llms.txt]() Data updated: 2026-07-09T06:35:08.371Z Evaluation of Hypothesis-mode Word Alignment performance on the Buckeye 5h test set, measured using metrics such as Reference Match percentage, Word Error Rate (WER), and Boundary Error metrics. A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage. | Scope | Recorded value | | --- | --- | | Task | Hypothesis-mode word alignment | | Dataset (including any stated split) | Buckeye 5h (test) | | Metric | Reference Match Rate | | Unit | PERCENT | | Direction | Higher is better | | Results loaded | 29 of 29 | | Coverage | All results currently returned for this benchmark | Evaluation settings may differ between results. The original papers and recorded configurations are needed to establish comparability. A shared metric name alone does not establish an identical protocol. ## Metrics - [Reference Match Rate](): Higher is better; unit: PERCENT. - [Word Error Rate (WER)](): Lower is better; unit: PERCENT. - [Matched Word Boundary Error (ms)](): Lower is better; unit: TIME. - [F1 Score (\<= 50ms)](): Higher is better; unit: PERCENT. ## Recorded results | Method | Reference Match Rate | Result date | Report paper | | --- | --- | --- | --- | | Canary-Qwen | 93.1 | Not recorded | 6a4f2aec32fb937314875ad4 | | Canary-Qwen | 93.1 | Not recorded | 6a4f2aec32fb937314875ad4 | | Phi-4-MM | 92.8 | Not recorded | 6a4f2aec32fb937314875ad4 | | Phi-4-MM | 92.8 | Not recorded | 6a4f2aec32fb937314875ad4 | | Crisper-Whisper | 92.6 | Not recorded | 6a4f2aec32fb937314875ad4 | | Parakeet CTC | 92.3 | Not recorded | 6a4f2aec32fb937314875ad4 | | Parakeet CTC | 92.3 | Not recorded | 6a4f2aec32fb937314875ad4 | | Crisper-Whisper | 92 | Not recorded | 6a4f2aec32fb937314875ad4 | | Crisper-Whisper | 92 | Not recorded | 6a4f2aec32fb937314875ad4 | | Parakeet TDT | 91.8 | Not recorded | 6a4f2aec32fb937314875ad4 | | Parakeet TDT | 91.8 | Not recorded | 6a4f2aec32fb937314875ad4 | | Parakeet RNN-T | 89.8 | Not recorded | 6a4f2aec32fb937314875ad4 | | Parakeet RNN-T | 89.8 | Not recorded | 6a4f2aec32fb937314875ad4 | | OWSM-CTC | 89.2 | Not recorded | 6a4f2aec32fb937314875ad4 | | OWSM-CTC | 89.2 | Not recorded | 6a4f2aec32fb937314875ad4 | | Voxtral | 88.3 | Not recorded | 6a4f2aec32fb937314875ad4 | | Voxtral | 88.3 | Not recorded | 6a4f2aec32fb937314875ad4 | | Whisper-large-v3 | 88.1 | Not recorded | 6a4f2aec32fb937314875ad4 | | Whisper-large-v3 | 88.1 | Not recorded | 6a4f2aec32fb937314875ad4 | | FastConformer (streaming) | 87.1 | Not recorded | 6a4f2aec32fb937314875ad4 | | FastConformer (streaming) | 87.1 | Not recorded | 6a4f2aec32fb937314875ad4 | | FastConformer (streaming) | 86.5 | Not recorded | 6a4f2aec32fb937314875ad4 | | FastConformer (streaming) | 86.5 | Not recorded | 6a4f2aec32fb937314875ad4 | | OWLS-1B | 85.7 | Not recorded | 6a4f2aec32fb937314875ad4 | | OWLS-1B | 85.7 | Not recorded | 6a4f2aec32fb937314875ad4 | | Whisper-base | 85.5 | Not recorded | 6a4f2aec32fb937314875ad4 | | Whisper-base | 85.5 | Not recorded | 6a4f2aec32fb937314875ad4 | | Emformer (streaming) | 72.1 | Not recorded | 6a4f2aec32fb937314875ad4 | | Emformer (streaming) | 72.1 | Not recorded | 6a4f2aec32fb937314875ad4 | ### Canary-Qwen Reference Match Rate: 93.1. Result ID: 6a4f410f60b88bb339e40e7c. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model Type":"Speech LLM","Align method":"Gradients"}. This record does not assert independent reproduction. ### Canary-Qwen Reference Match Rate: 93.1. Result ID: 6a4f410f60b88bb339e40e7d. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model Type":"Speech LLM","Align method":"Self-att."}. This record does not assert independent reproduction. ### Phi-4-MM Reference Match Rate: 92.8. Result ID: 6a4f410f60b88bb339e40e7a. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model Type":"Speech LLM","Align method":"Gradients"}. This record does not assert independent reproduction. ### Phi-4-MM Reference Match Rate: 92.8. Result ID: 6a4f410f60b88bb339e40e7b. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model Type":"Speech LLM","Align method":"Self-att."}. This record does not assert independent reproduction. ### Crisper-Whisper Reference Match Rate: 92.6. Result ID: 6a4f410f60b88bb339e40e75. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model Type":"AED","Align method":"Official"}. This record does not assert independent reproduction. ### Parakeet CTC Reference Match Rate: 92.3. Result ID: 6a4f410f60b88bb339e40e61. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model Type":"CTC","Align method":"Gradients"}. This record does not assert independent reproduction. ### Parakeet CTC Reference Match Rate: 92.3. Result ID: 6a4f410f60b88bb339e40e62. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model Type":"CTC","Align method":"Posteriors"}. This record does not assert independent reproduction. ### Crisper-Whisper Reference Match Rate: 92. Result ID: 6a4f410f60b88bb339e40e73. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model Type":"AED","Align method":"Gradients"}. This record does not assert independent reproduction. ### Crisper-Whisper Reference Match Rate: 92. Result ID: 6a4f410f60b88bb339e40e74. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model Type":"AED","Align method":"Cross-att."}. This record does not assert independent reproduction. ### Parakeet TDT Reference Match Rate: 91.8. Result ID: 6a4f410f60b88bb339e40e69. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model Type":"Transd.","Align method":"Gradients"}. This record does not assert independent reproduction. ### Parakeet TDT Reference Match Rate: 91.8. Result ID: 6a4f410f60b88bb339e40e6a. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model Type":"Transd.","Align method":"Posteriors"}. This record does not assert independent reproduction. ### Parakeet RNN-T Reference Match Rate: 89.8. Result ID: 6a4f410f60b88bb339e40e67. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model Type":"Transd.","Align method":"Gradients"}. This record does not assert independent reproduction. ### Parakeet RNN-T Reference Match Rate: 89.8. Result ID: 6a4f410f60b88bb339e40e68. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model Type":"Transd.","Align method":"Posteriors"}. This record does not assert independent reproduction. ### OWSM-CTC Reference Match Rate: 89.2. Result ID: 6a4f410f60b88bb339e40e63. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model Type":"CTC","Align method":"Gradients"}. This record does not assert independent reproduction. ### OWSM-CTC Reference Match Rate: 89.2. Result ID: 6a4f410f60b88bb339e40e64. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model Type":"CTC","Align method":"Posteriors"}. This record does not assert independent reproduction. ### Voxtral Reference Match Rate: 88.3. Result ID: 6a4f410f60b88bb339e40e78. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model Type":"Speech LLM","Align method":"Gradients"}. This record does not assert independent reproduction. ### Voxtral Reference Match Rate: 88.3. Result ID: 6a4f410f60b88bb339e40e79. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model Type":"Speech LLM","Align method":"Self-att."}. This record does not assert independent reproduction. ### Whisper-large-v3 Reference Match Rate: 88.1. Result ID: 6a4f410f60b88bb339e40e71. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model Type":"AED","Align method":"Gradients"}. This record does not assert independent reproduction. ### Whisper-large-v3 Reference Match Rate: 88.1. Result ID: 6a4f410f60b88bb339e40e72. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model Type":"AED","Align method":"Cross-att."}. This record does not assert independent reproduction. ### FastConformer (streaming) Reference Match Rate: 87.1. Result ID: 6a4f410f60b88bb339e40e6d. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model Type":"Transd.","Align method":"Gradients"}. This record does not assert independent reproduction. ### FastConformer (streaming) Reference Match Rate: 87.1. Result ID: 6a4f410f60b88bb339e40e6e. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model Type":"Transd.","Align method":"Posteriors"}. This record does not assert independent reproduction. ### FastConformer (streaming) Reference Match Rate: 86.5. Result ID: 6a4f410f60b88bb339e40e65. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model Type":"CTC","Align method":"Gradients"}. This record does not assert independent reproduction. ### FastConformer (streaming) Reference Match Rate: 86.5. Result ID: 6a4f410f60b88bb339e40e66. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model Type":"CTC","Align method":"Posteriors"}. This record does not assert independent reproduction. ### OWLS-1B Reference Match Rate: 85.7. Result ID: 6a4f410f60b88bb339e40e76. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model Type":"AED","Align method":"Gradients"}. This record does not assert independent reproduction. ### OWLS-1B Reference Match Rate: 85.7. Result ID: 6a4f410f60b88bb339e40e77. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model Type":"AED","Align method":"Cross-att."}. This record does not assert independent reproduction. ### Whisper-base Reference Match Rate: 85.5. Result ID: 6a4f410f60b88bb339e40e6f. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model Type":"AED","Align method":"Gradients"}. This record does not assert independent reproduction. ### Whisper-base Reference Match Rate: 85.5. Result ID: 6a4f410f60b88bb339e40e70. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model Type":"AED","Align method":"Cross-att."}. This record does not assert independent reproduction. ### Emformer (streaming) Reference Match Rate: 72.1. Result ID: 6a4f410f60b88bb339e40e6b. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model Type":"Transd.","Align method":"Gradients"}. This record does not assert independent reproduction. ### Emformer (streaming) Reference Match Rate: 72.1. Result ID: 6a4f410f60b88bb339e40e6c. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model Type":"Transd.","Align method":"Posteriors"}. This record does not assert independent reproduction. Page 1 of 1. This is an index page, not the full collection.