# Speech Recognition and Diarization on AMI IHM

Source: [Speech Recognition and Diarization on AMI IHM](<https://www.sota2.com/research/sota/speech-recognition-and-diarization-on-ami-ihm>)

Collection index: [llms.txt](<https://www.sota2.com/research/llms.txt>)

Data updated: 2026-06-02T11:06:58.320Z

Evaluation combining Speech Recognition (WER) and Speaker Diarization (DER, cpWER, tcpWER) performance on the AMI IHM dataset.

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 | Speech Recognition and Diarization |
| Dataset (including any stated split) | AMI IHM |
| Metric | WER |
| Unit | SCALAR |
| Direction | Lower is better |
| Results loaded | 6 of 6 |
| 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

- [WER](<https://www.sota2.com/research/sota/speech-recognition-and-diarization-on-ami-ihm.md?metric=WER>): Lower is better; unit: SCALAR.

- [DER](<https://www.sota2.com/research/sota/speech-recognition-and-diarization-on-ami-ihm.md?metric=DER>): Lower is better; unit: SCALAR.

- [cpWER](<https://www.sota2.com/research/sota/speech-recognition-and-diarization-on-ami-ihm.md?metric=cpWER>): Lower is better; unit: SCALAR.

- [tcpWER](<https://www.sota2.com/research/sota/speech-recognition-and-diarization-on-ami-ihm.md?metric=tcpWER>): Lower is better; unit: SCALAR.

## Recorded results

| Method | WER | Result date | Report paper |
| --- | --- | --- | --- |
| VibeVoice Single-pass baseline | 19.21 | Not recorded | 6a1e63d3dbd7150506f85445 |
| MURMUR: Chunked, no eviction | 19.8 | Not recorded | 6a1e63d3dbd7150506f85445 |
| MURMUR: Output token eviction | 19.89 | Not recorded | 6a1e63d3dbd7150506f85445 |
| MURMUR: Both eviction | 20.23 | Not recorded | 6a1e63d3dbd7150506f85445 |
| MURMUR: Speech token eviction | 20.41 | Not recorded | 6a1e63d3dbd7150506f85445 |
| WhisperX | 22.51 | Not recorded | 6a1e63d3dbd7150506f85445 |

### VibeVoice Single-pass baseline

WER: 19.21. Result ID: 6a1eb8fefbc373592c19cda5.

[Reporting paper](<https://www.sota2.com/research/paper/6a1e63d3dbd7150506f85445.md>)

[Method origin paper](<https://www.sota2.com/research/paper/697846c6a2b28e761d87c6d8.md>)

Recorded evaluation configuration: {"Chunk size":"300s","Batch size":"16","Total (s)":"370.73","Speed-up":"baseline"}. This record does not assert independent reproduction.

### MURMUR: Chunked, no eviction

WER: 19.8. Result ID: 6a1eb8fefbc373592c19cda7.

[Reporting paper](<https://www.sota2.com/research/paper/6a1e63d3dbd7150506f85445.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6a1e63d3dbd7150506f85445.md>)

Recorded evaluation configuration: {"Chunk size":"300s","Batch size":"16","KV cache eviction strategy":"None","Total (s)":"100.8","Speed-up":"3.68×"}. This record does not assert independent reproduction.

### MURMUR: Output token eviction

WER: 19.89. Result ID: 6a1eb8fefbc373592c19cda8.

[Reporting paper](<https://www.sota2.com/research/paper/6a1e63d3dbd7150506f85445.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6a1e63d3dbd7150506f85445.md>)

Recorded evaluation configuration: {"Chunk size":"300s","Batch size":"16","KV cache eviction strategy":"s4 w1024","Total (s)":"97.6","Speed-up":"3.79×"}. This record does not assert independent reproduction.

### MURMUR: Both eviction

WER: 20.23. Result ID: 6a1eb8fefbc373592c19cdaa.

[Reporting paper](<https://www.sota2.com/research/paper/6a1e63d3dbd7150506f85445.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6a1e63d3dbd7150506f85445.md>)

Recorded evaluation configuration: {"Chunk size":"300s","Batch size":"16","KV cache eviction strategy":"Output and Speech eviction","Total (s)":"96.1","Speed-up":"3.85×"}. This record does not assert independent reproduction.

### MURMUR: Speech token eviction

WER: 20.41. Result ID: 6a1eb8fefbc373592c19cda9.

[Reporting paper](<https://www.sota2.com/research/paper/6a1e63d3dbd7150506f85445.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6a1e63d3dbd7150506f85445.md>)

Recorded evaluation configuration: {"Chunk size":"300s","Batch size":"16","KV cache eviction strategy":"d512 w1024","Total (s)":"88.43","Speed-up":"4.20×"}. This record does not assert independent reproduction.

### WhisperX

WER: 22.51. Result ID: 6a1eb8fefbc373592c19cda6.

[Reporting paper](<https://www.sota2.com/research/paper/6a1e63d3dbd7150506f85445.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6ffe1e209131cea48624.md>)

Recorded evaluation configuration: {"Chunk size":"300s","Batch size":"16","Total (s)":"38.81","Speed-up":"9.55×"}. This record does not assert independent reproduction.

Page 1 of 1. This is an index page, not the full collection.
