Searcharxiv⌕ Search

arXiv subjects

Luca Cerovaz

Publications and source records attributed to Luca Cerovaz.

5 recordsLinked to original sources

SAGE: Semantic Audio Generative Encoder

Audio autoencoders compress waveforms into compact latent representations that serve as the interface between raw audio and downstream models. Current systems navigate a three-way trade-off between reconstruction quality, semantic structure of the latent space, and inference speed, typically favoring one or two of these at the expense of the others. This paper introduces SAGE, Semantic Audio Generative Encoder: a compact variational autoencoder, trained solely on publicly available music, that shapes its latent by distilling embeddings from a pretrained audio-text model. This 105M-parameter model runs about 3x faster than Stable Audio Open and reaches the listening-test quality of SAME-L, an autoencoder 8x larger and over 2x slower, while surpassing both on objective perceptual and distributional metrics of reconstruction. Furthermore, it sets the state of the art on all nineteen probing tasks of latent semantics, in domain and out of domain. These results establish SAGE as a lightweight audio autoencoder that strikes the best balance of the three-way trade-off among those we evaluate, combining high reconstruction fidelity, state-of-the-art semantic structure, and fast inference.

cs.SD↗

Limite 1B - Violetto: Technical Report

We present Limite 1B - Violetto, a one-billion-parameter model designed for high-throughput mathematical reasoning. Limite is trained from scratch on 237 billion tokens from a carefully curated, mathematics-centric corpus and supports context lengths of up to 131k tokens. Its architecture and training recipe were developed together to maximize reasoning capability while keeping training and inference costs low. Post-training combines knowledge distillation, targeted supervised fine-tuning, and RLVR on a curated curriculum. Limite achieves 94.01% on AIME 2026 and 74.25% on BeyondAIME, placing it among models many times its size on our evaluation suite. The model is deliberately only lightly instruction-tuned: it is intended primarily as a single-turn mathematical solver and as a fast, parallel component of larger systems, rather than as a general-purpose assistant. We release the model, its base and value checkpoints, and an optimized serving implementation under the Apache 2.0 license.

cs.CL↗

How Neural Losses Shape VAE Latents

Modern VAEs are rarely trained with the pointwise likelihood implied by the standard $β$-VAE objective. In practice, pointwise reconstruction is often combined with perceptual and adversarial losses, despite a lack of understanding of how this changes the latent dynamics of the model. We show that the choice of reconstruction loss reshapes the rate-distortion problem itself, altering both the information content and the geometry of the learned latent space in ways that may be invisible from reconstructions alone. First, we prove and verify empirically that augmenting pointwise reconstruction with neural terms, such as perceptual and adversarial objectives, reduces the amount of information stored in the latent representations. Second, we show that neural reconstruction losses systematically change the geometry of the latent space: they make representations more isotropic and distribute uncertainty more evenly across latent dimensions, producing different posterior variance profiles. These findings highlight how the rate-distortion tradeoff is not a comprehensive lens to understand the behavior of VAEs, and we propose a more mechanistic approach to investigate how the choice of a distortion metric reshapes the optimization problem.

cs.LG↗

PHALAR: Phasors for Learned Musical Audio Representations

Stem retrieval, the task of matching missing stems to a given audio submix, is a key challenge currently limited by models that discard temporal information. We introduce PHALAR, a contrastive framework achieving a relative accuracy increase of up to $\approx 70\%$ over the state-of-the-art while requiring $<50\%$ of the parameters and a 7$\times$ training speedup. By utilizing a Learned Spectral Pooling layer and a complex-valued head, PHALAR enforces pitch-equivariant and phase-equivariant biases. PHALAR establishes new retrieval state-of-the-art across MoisesDB, Slakh, and ChocoChorales, correlating significantly higher with human coherence judgment than semantic baselines. Finally, zero-shot beat tracking and linear chord probing confirm that PHALAR captures robust musical structures beyond the retrieval task.

cs.SD↗

EuleroDec: A Complex-Valued RVQ-VAE for Efficient and Robust Audio Coding

Audio codecs power discrete music generative modelling, music streaming and immersive media by shrinking PCM audio to bandwidth-friendly bit-rates. Recent works have gravitated towards processing in the spectral domain; however, spectrogram-domains typically struggle with phase modeling which is naturally complex-valued. Most frequency-domain neural codecs either disregard phase information or encode it as two separate real-valued channels, limiting spatial fidelity. This entails the need to introduce adversarial discriminators at the expense of convergence speed and training stability to compensate for the inadequate representation power of the audio signal. In this work we introduce an end-to-end complex-valued RVQ-VAE audio codec that preserves magnitude-phase coupling across the entire analysis-quantization-synthesis pipeline and removes adversarial discriminators and diffusion post-filters. Without GANs or diffusion we match or surpass much longer-trained baselines in-domain and reach SOTA out-of-domain performance. Compared to standard baselines that train for hundreds of thousands of steps, our model reducing training budget by an order of magnitude is markedly more compute-efficient while preserving high perceptual quality.

cs.SD↗