Line of inquiry
Inquiring lines›What determines when language mode…›What internal mechanisms drive lan…›this line of inquiry
Does learning articulatory physics explain speech communication superiority?
A broader line of inquiry — a family of 52 specific questions the research asks around this. Follow one into its inquiring-line page, or move sideways to a related line below.
Questions in this line of inquiry 52
Specific inquiring lines the field asks around this — ordered from the most general framing down to the most specific angle.
- Do discrete tokenized modalities preserve information better than continuous embeddings?
- Can speech embeddings carry articulatory structure that text cannot?
- How does the articulatory substrate explain direct speech-to-speech superiority over transcription pipelines?
- Do speech encoders actually learn the physics of how vocal tracts produce sound?
- Can we balance interpretability with the efficiency gains of compressed inter-model communication?
- Can speech encoders learn vocal tract physics without phonetic labels?
- Can feature disentanglement in gesture synthesis generalize to completely unseen voice distributions?
- How do speech encoders learn articulatory physics without phonetic labels?
- Do speech models learn the articulatory processes that produce acoustic signals?
- Do self-supervised speech models learn dialect-agnostic representations across languages?
- Can spectral eigenvector ordering serve as a model-agnostic interpretability probe?
- How does causal multimodal modeling differ from encoder-decoder architectures?
- Do high-entropy RLVR tokens correspond to MI-peak tokens during inference?
- Can next-token prediction train models to optimize for communication efficiency?
- Why is latent-level prediction more sample-efficient than token-level prediction?
- How do hidden embeddings preserve more information than discrete tokens?
- Why does articulatory probing predict SSL model performance better than phonetic probing?
- Why do image captions create different friction than pure video data?
- What makes internal embeddings useful as multimodal input for language model training?
- How do different speech encoder layers capture different types of gesture information?
- What compression explains why syntax fits in low-dimensional subspaces?
- What scaling exponent would audio or other modalities require in a truly multimodal system?
- How should speech recognition systems be audited for demographic bias?
- Can decoder-only models become effective text encoders with training?
- Does the same spectral signature appear across different embedding models?
- Why does emotion-guided diffusion outperform discrete emotion category selection for gesture?
- Why does forcing single labels on emotions destroy information similar to language?
- What information does transcription destroy that direct speech-to-speech models preserve?
- Why does transcription destroy prosodic information in speech processing?
- Can generative reconstruction preserve latent manifold structure better than geometric compression?
- Why does statistical compression destroy literary connotation and meaning?
- Can articulatory inversion serve as a window into what speech models have learned?
- Why do handcrafted acoustic features outperform neural speaker embeddings for personality?
- How does entropy-based patching compare to fixed token vocabularies in practice?
- How do encode-decode contractive biases create stable attractors in latent space?
- What moves become possible when you represent ASR as a noisy observation model?
- Can transcription quality metrics adequately measure fairness across demographic groups?
- What physical structure does a Gaussian-regularized latent space actually encode?
- What spectral signatures distinguish hierarchy-driven geometry from corpus-driven geometry?
- What information does transcription destroy that direct speech pathways preserve?
- How do commercial ASR systems produce disparate error rates by speaker race?
- How does iconicity detection work within static embeddings before any attention?
- What training data do modern commercial ASR systems actually contain?
- Why does text encoding create different subspaces across domains?
- Do generic kernel-decay assumptions alone explain coarse-to-fine spectral ordering?
- Why does masking the penultimate token outperform random token masking?
- How does data entropy inflate compression estimates in prequential coding?
- What paired speech data is needed to train end-to-end models?
- Why do cascade pipelines fail to capture global motion structure?
- What temporal and spatial constraints does Space-Time U-Net solve?
- Why does entropy-based frame sampling work better than uniform stride selection?
- How does epiplexity measure extractable value differently from compression codelength?