ImpactMat
ImpactMat: Continuous Material Estimation for Inverse Impact Sound Rendering
A synthetic dataset and benchmark of single and blended material impact sounds paired with ground-truth continuous physical parameters for inverse sound rendering evaluation.
How to run
$ npm i -g lettokenburn-cli$ ltb bench info impactmat
External harness. No Harbor adapter has been generated for this benchmark yet. `ltb bench info` prints the known links.
Runs happen on your machine with your own subscription; only the score summary is pushed. CLI docs
Scores
2 sourced entries| Harness | Source | Confidence | Link | |||
|---|---|---|---|---|---|---|
| Qwen2.5-Omni-7B | 1.3 | icl | Independent | medium | Oct 5, 2026 | arxiv.org |
| Audio Flamingo 3 | 12.4 | icl | Independent | medium | Oct 5, 2026 | arxiv.org |
Every number links to where it was published. Low-confidence rows are muted and need per-cell verification.
Reproduced by the community
All runsAbout
ImpactMat contains synthetic impact sounds generated from 1,100 objects across eight base materials and their blends. Each sample is paired with a five-dimensional vector of continuous physical parameters (Young’s modulus, density, Poisson’s ratio, damping coefficients).
The benchmark evaluates feed-forward models on normalized Mean Absolute Error (NMAE) of parameter estimation. The proposed method outperforms baselines like DiffSound and audio-language models in recovering continuous material properties from audio.