Music Lenses: ISMIR paper uses sparse autoencoders for steerable concept-level music retrieval
umpedronosapato · x · 2026-10-27
An ISMIR 2026 paper introduces Music Lenses, combining sparse inversion and sparse autoencoders to enable concept-conditional similarity over a dense music embedding space. Instead of merely finding similar tracks, users steer individual concepts via text prompts—more piano, more intimate, less gritty—to modify a query, e.g. matching vocal character rather than whole songs. An interactive demo lets users pan/zoom a song map and click waveforms to listen; built-in lenses retain 80% of member variance.
More from Research
- Math advisory group AGMAI urges OpenAI to immediately release Yang–Mills AI outputs — LucaAmb · 2026-10-27
- Ex-Vercel engineer Shu Ding: linear approximation gives free layout animations — shuding · 2026-10-27
- Treat designs as points in N-d space: geodesic transitions where no card jumps rows — shuding · 2026-10-27
- Interval arithmetic can prove a UI never fails across infinitely many props — shuding · 2026-10-27
- Vercel Labs compiles React UI into a pure math function, unlocking calculus for interfaces — shuding · 2026-10-27
- AuthorizationBench: LLM agents overstep user authorization in 19.2% of runs — rohanpaul_ai · 2026-10-27