hzhongresearch/auditoryhum_supplementary

AuditoryHuM supplementary data

AuditoryHuM: Auditory Scene Label Generation and Clustering using Human-MLLM Collaboration. This is the supplementary material used to generate the results in the paper. The Keras models require the presence of https://www.kaggle.com/api/v1/models/google/yamnet/tensorFlow2/yamnet/1/download Download yamnet-tensorflow2-yamnet-v1.tar.gz and extract this model to a directory named yamnet_model. mkdir yamnet_model && tar -xvzf yamnet-tensorflow2-yamnet-v1.tar.gz --directory yamnet_model

• Website
• Paper
• Code
• Supplementary Data
• Demo Website With Alignment Samples

Description of data

Files	Description
advance_file_list.csv	List of files and the order with which they are processed.
ahead-ds_file_list.csv
tau2019_file_list.csv
advance_gemma3n_e2b_labels.csv	Labels generated by MLLMs.
advance_qwen2a_labels.csv
advance_qwen2_5o_labels.csv
ahead-ds_gemma3n_e2b_labels.csv
ahead-ds_qwen2a_labels.csv
ahead-ds_qwen2_5o_labels.csv
tau2019_gemma3n_e2b_labels.csv
tau2019_qwen2a_labels.csv
tau2019_qwen2_5o_labels.csv
advance_gemma3n_e2b_labels_annotations.csv	Human provided labels.
advance_qwen2a_labels_annotations.csv
advance_qwen2_5o_labels_annotations.csv
ahead-ds_gemma3n_e2b_labels_annotations.csv
ahead-ds_qwen2a_labels_annotations.csv
ahead-ds_qwen2_5o_labels_annotations.csv
tau2019_gemma3n_e2b_labels_annotations.csv
tau2019_qwen2a_labels_annotations.csv
tau2019_qwen2_5o_labels_annotations.csv
advance_captions.csv	Captions generated by MLLM.
ahead-ds_captions.csv
tau2019_captions.csv
advance_qwen2_5o_human_strat.csv	More labels for testing human labelling strategy.
advance_provided_labels.csv	Dataset provided labels.
ahead-ds_provided_labels.csv
tau2019_provided_labels.csv
advance_provided_train.csv	Dataset provided labels for downstream training.
advance_provided_validation.csv
advance_provided_test.csv
ahead-ds_provided_train.csv
ahead-ds_provided_validation.csv
ahead-ds_provided_test.csv
tau2019_provided_train.csv
tau2019_provided_validation.csv
tau2019_provided_test.csv
advance_auditoryhum_train.csv	AuditoryHuM labels for downstream training.
advance_auditoryhum_validation.csv
advance_auditoryhum_test.csv
ahead-ds_auditoryhum_train.csv
ahead-ds_auditoryhum_validation.csv
ahead-ds_auditoryhum_test.csv
tau2019_auditoryhum_train.csv
tau2019_auditoryhum_validation.csv
tau2019_auditoryhum_test.csv
yamnetp_advance_provided.keras	OpenYAMNet models trained on dataset provided labels.
yamnetp_ahead-ds_provided.keras
yamnetp_tau2019_provided.keras
yamnetp_advance_auditoryhum.keras	OpenYAMNet models trained on AuditoryHuM labels.
yamnetp_ahead-ds_auditoryhum.keras
yamnetp_tau2019_auditoryhum.keras

Licence

Licenced under CC-BY-4.0. See LICENCE.txt.

Attribution.

@misc{zhong2026auditoryhumauditoryscenelabel,
      title={AuditoryHuM: Auditory Scene Label Generation and Clustering using Human-MLLM Collaboration}, 
      author={Henry Zhong and Jörg M. Buchholz and Julian Maclaren and Simon Carlile and Richard F. Lyon},
      year={2026},
      eprint={2602.19409},
      archivePrefix={arXiv},
      primaryClass={cs.SD},
      url={https://arxiv.org/abs/2602.19409}, 
}