BidirLM: From Text to Omnimodal Bidirectional Encoders by Adapting and Composing Causal LLMs
Paper • 2604.02045 • Published • 38
positive dict | __key__ stringlengths 8 11 | __url__ stringclasses 1
value | anchor stringlengths 0 538 |
|---|---|---|---|
{"array":[0.11658812314271927,0.13368800282478333,0.1213090792298317,0.1218159943819046,0.1008025407(...TRUNCATED) | sample_1 | hf://datasets/laion/LAION-Audio-300M@main/flash_15_2_random_snippets_190.tar | "The audio consists primarily of a low, resonant hum, similar to the previous file but slightly loud(...TRUNCATED) |
{"array":[-0.3148557245731354,-0.3485443890094757,-0.31956517696380615,-0.34786057472229004,-0.32258(...TRUNCATED) | sample_3 | hf://datasets/laion/LAION-Audio-300M@main/flash_15_2_random_snippets_190.tar | "This clip is dominated by a persistent, low-pitched hum. The sound is fairly consistent in volume a(...TRUNCATED) |
{"array":[0.13354352116584778,0.1366734802722931,0.11123676598072052,0.11053533852100372,0.087821707(...TRUNCATED) | sample_4 | hf://datasets/laion/LAION-Audio-300M@main/flash_15_2_random_snippets_190.tar | "A low-level, continuous hum is the main auditory feature of this file. It's less prominent than in(...TRUNCATED) |
{"array":[0.011666100472211838,0.03146066889166832,0.039140257984399796,0.03170940279960632,0.023465(...TRUNCATED) | sample_9 | hf://datasets/laion/LAION-Audio-300M@main/flash_15_2_random_snippets_190.tar | "This short audio clip features a low, consistent hum. The sound is quiet and lacks distinct charac(...TRUNCATED) |
{"array":[0.11658812314271927,0.13368800282478333,0.1213090792298317,0.1218159943819046,0.1008025407(...TRUNCATED) | sample_11 | hf://datasets/laion/LAION-Audio-300M@main/flash_15_2_random_snippets_190.tar | "This file features a low, resonant hum that is slightly more pronounced than in some previous files(...TRUNCATED) |
{"array":[-0.3148557245731354,-0.3485443890094757,-0.31956517696380615,-0.34786057472229004,-0.32258(...TRUNCATED) | sample_13 | hf://datasets/laion/LAION-Audio-300M@main/flash_15_2_random_snippets_190.tar | "A low-frequency hum is the primary auditory element in this file. The sound is relatively stable a(...TRUNCATED) |
{"array":[-0.0005944287404417992,0.0034230872988700867,0.006422434933483601,0.0033243969082832336,-0(...TRUNCATED) | sample_18 | hf://datasets/laion/LAION-Audio-300M@main/flash_15_2_random_snippets_190.tar | "The file is dominated by a low, sustained hum. It's quiet and lacks strong character. The recordin(...TRUNCATED) |
{"array":[0.031681496649980545,0.016386352479457855,-0.06533752381801605,-0.09438370168209076,-0.093(...TRUNCATED) | sample_20 | hf://datasets/laion/LAION-Audio-300M@main/flash_15_2_random_snippets_190.tar | "Upbeat, playful music with a clear, bright sound. The genre seems to be a type of pop or upbeat ch(...TRUNCATED) |
{"array":[0.09359826147556305,0.12612353265285492,0.12834006547927856,0.16644428670406342,0.09237072(...TRUNCATED) | sample_21 | hf://datasets/laion/LAION-Audio-300M@main/flash_15_2_random_snippets_190.tar | "This audio clip features a similar musical style to the previous one: a bright, cheerful, potential(...TRUNCATED) |
{"array":[0.05242771655321121,0.05457395315170288,-0.005757701583206654,-0.02586378902196884,-0.0408(...TRUNCATED) | sample_22 | hf://datasets/laion/LAION-Audio-300M@main/flash_15_2_random_snippets_190.tar | "This piece of music maintains the same upbeat and cheerful style. A prominent keyboard is still pr(...TRUNCATED) |
End of preview. Expand in Data Studio
⚠️ Part of the BidirLM-Omni Collection > This dataset is a specific modality sub-sample of the corpus used to train the BidirLM-Omni models.
Looking for the full training mixture? > If you want to access the complete, balanced 1.8M sample omnimodal dataset (integrating text, image, audio), please visit the global integration hub here:
👉 BidirLM/BidirLM-Omni-Contrastive
📜 Citation
If you use this processed dataset or the broader BidirLM mixture in your research, please cite our work:
@misc{boizard2026bidirlmtextomnimodalbidirectional,
title={BidirLM: From Text to Omnimodal Bidirectional Encoders by Adapting and Composing Causal LLMs},
author={Nicolas Boizard and Théo Deschamps-Berger and Hippolyte Gisserot-Boukhlef and Céline Hudelot and Pierre Colombo},
year={2026},
eprint={2604.02045},
archivePrefix={arXiv},
primaryClass={cs.CL},
url={https://arxiv.org/abs/2604.02045},
}
- Downloads last month
- 1,375