aiqbench analysis
This research investigates the semantic token codebook of Moshi, demonstrating how its 2048 tokens map to phonetic units like phones and triphones when realigned with standard TIMIT transcriptions. The study reveals that initial ABX experiments failed to capture these mappings accurately but offers a corrected interpretation for understanding neural codec representations in language models.
Using the Mimi codec for metalinguistic representations
arXiv:2608.15799v1 Announce Type: new Abstract: In this paper, we focus on the dictionary of 2048 tokens used in Mimi semantic token codebook, the neural codec of the Moshi language model. We show that the ABX experiment carried out with Mimi fails to capture the mapping of the semantic tokens to phone realisations. By realigning Mimi representations to the TIMIT corpus transcriptions, we show that the 2048 tokens IDs of the semantic codebook map to quadphone, triphone, biphone, phone and subphone realisations.
Abstract:In this paper, we focus on the dictionary of 2048 tokens used in Mimi semantic token codebook, the neural codec of the Moshi language model. We show that the ABX experiment carried out with Mimi fails to capture the mapping of the semantic tokens to phone realisations. By realigning Mimi representations to the TIMIT corpus transcriptions, we show that the 2048 tokens IDs of the semantic codebook map to quadphone, triphone, biphone, phone and subphone realisations.
Computation and Language (cs.CL)
arXiv:2608.15799 [cs.CL]
(or arXiv:2608.15799v1 [cs.CL] for this version)
https://doi.org/10.48550/arXiv.2608.15799 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Full text under the publisher's open licence · 171 words
Related coverage
Masked Autoregressive Speech Enhancement with Continuous Neural Audio Codec Representations
arXiv · Sep 5, 2026
Data-driven Video Codec with Implicit Neural Representations
arXiv · Jul 20, 2026
What Survives the Codec Shift: Pooled No-Vocals Residuals for Speech Deepfake Detection
arXiv · Sep 30, 2026
CoVisco: Codec-Native Vision Encoder with Native Token Compression for Unified Image-Video Understanding
arXiv · Oct 1, 2026