AI Paper Atlas

Research map / Image generation / Autoregressive and latent generation

Continuous Audio Language Models

Simon Rouard, Manu Orsini, Axel Roebel, Neil Zeghidour, Alexandre Défossez

ICLR 2026 · Image generation › Autoregressive and latent generation · approach: Diffusion audio synthesis · 11 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
Replace 8-RVQ RQ-Transformer head with 1-step consistency headSpeech continuation LM · Speech continuationAcoustic quality MOS (temp 0.8): 2.75 -> 3.45Improved
Lower Gaussian sampling temperature from 1.0 to 0.8 (ablation)CALM speech continuation · Speech continuationPPX: 42.9 -> 23.8Improved
Add noise augmentation to the backbone context (ablation)CALM music, consistency 4 steps · Music continuation ablationFAD: 1.63 -> 0.93Improved
Add short-context transformer (ablation)CALM music, consistency 4 steps · Music continuation ablationFAD: 4.03 -> 0.93Improved
Add head batch multiplier (ablation)CALM music, consistency 4 steps · Music continuation ablationFAD: 1.32 -> 0.93Improved
Use consistency head 1 step instead of RQ-Transformer 32 RVQMusic continuation LM · Music continuationFAD: 1.06 -> 0.83Improved
Use 4 consistency steps instead of 1 (ablation)CALM music · Music continuationFAD: 0.83 -> 0.71Improved
Use TrigFlow 100-step head instead of consistency (ablation)CALM music · Music continuationFAD / speed: FAD 0.64 vs 0.71; overall speedup x0.3 vs x1.9Mixed
Add short-context transformer and noise injection for speech (ablation)CALM speech continuation · Speech continuationqualityNo clear effect
Reduce VAE to 32 dims for music (ablation)Music VAE · Music compressionVisQOL: 4.01 (128 dims) -> 2.23 (32 dims)Hurt
Use continuous VAE (32 dims) instead of 8-RVQ Mimi codec for speechSpeech codec · Speech compressionPESQ: 2.13 -> 2.42Improved
CALM TTS vs DSM on speaker similarityCALM TTS with LSD · LibriSpeech test-cleanWavLM speaker similarity: 0.67 -> 0.52Hurt

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy