mlsanityData to 14 Oct 2026

I'm Sorry for Your Loss: Spectrally-Based Audio Distances Are Bad at Pitch

arXiv 2012.04572 ↗First posted 8 Dec 2020cs.SD cs.AI cs.LGcited by 17

Not on the map. Why?

Abstract (arXiv)

Growing research demonstrates that synthetic failure modes imply poor generalization. We compare commonly used audio-to-audio losses on a synthetic benchmark, measuring the pitch distance between two stationary sinusoids. The results are surprising: many have poor sense of pitch direction. These shortcomings are exposed using simple rank assumptions. Our task is trivial for humans but difficult for these audio distances, suggesting significant progress can be made in self-supervised audio learning by improving current losses.

Key facts. Cited by 17 papers in the map (data through 14 Oct 2026). First posted on arXiv 8 Dec 2020.

Builds on

No reference list yet.

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2021: 2; 2022: 3; 2023: 4; 2024: 3; 2025: 2; 2026: 3.

17 papers.

Sort:

By area:

  1. 2022DDX7: Differentiable FM Synthesis of Musical Instrument Soundscited by 13
  2. 2021One Billion Audio Sounds from GPU-enabled Modular Synthesiscited by 11
  3. 2023Unsupervised Harmonic Parameter Estimation Using Differentiable DSP and Spectral Optimal Transportcited by 8
  4. 2024DiffMoog: a Differentiable Modular Synthesizer for Sound Matchingcited by 7
  5. 2025Audio synthesizer inversion in symmetric parameter spaces with approximately equivariant flow matchingcited by 6
  6. 2021Towards Lightweight Controllable Audio Synthesis with Conditional Implicit Neural Representationscited by 5
  7. 2022Improving Speech Enhancement through Fine-Grained Speech Characteristicscited by 5
  8. 2024Differentiable Modal Synthesis for Physical Modeling of Planar String Sound and Motion Simulationcited by 4
  9. 2024Real-time Timbre Remapping with Differentiable DSPcited by 3
  10. 2025Fast Differentiable Modal Simulation of Non-linear Strings, Membranes, and Platescited by 3
  11. 2023Mesostructures: Beyond Spectrogram Loss in Differentiable Time-Frequency Analysiscited by 2
  12. 2023TAPLoss: A Temporal Acoustic Parameter Loss for Speech Enhancementcited by 2
5 more
  1. 2023PAAPLoss: A Phonetic-Aligned Acoustic Parameter Loss for Speech Enhancementcited by 2
  2. 2022R-MelNet: Reduced Mel-Spectral Modeling for Neural TTScited by 0
  3. 2026Architecture and Affordances of PLAUD: Performative Latents and Unsupervised DDSPcited by 0
  4. 2026Evaluating Loss Functions in Differentiable Out-of-Domain Sound-Matching with Partial Parameter Distancecited by 0
  5. 2026Off-manifold robustness in synthesizer inversion with joint distribution flow matchingcited by 0