<strong>Accepted to ECCV 2026 (Poster)</strong></p>\n<p>We study sinusoidal recurrence as an iterative mechanism for harmonic spectral enrichment in implicit neural representations (INRs). Our analysis reveals that sinusoidal activations induce a harmonic line spectrum, providing a spectral account of how recurrent unrolling enriches the effective spectral support. We realize this principle with a shared sinusoidal block that iteratively refines the latent representation. We empirically validate the resulting spectral behavior against feed-forward INRs, non-sinusoidal recurrent variants, and equilibrium-style sinusoidal models. Complementing this analysis, we evaluate the proposed architecture across image and 3D representation tasks. On RGB image benchmarks, our method achieves higher fidelity than feed-forward baselines with fewer parameters and fewer optimization steps, and it further transfers favorably to super-resolution, NeRF, and SDF tasks.</p>\n","updatedAt":"2026-07-24T04:33:18.317Z","author":{"_id":"680ed019a9918bb1cbc66248","avatarUrl":"/avatars/904d243f2fad99341f11795e93788993.svg","fullname":"Hyunmin Cho","name":"hyeoncho01","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"isUserFollowing":false}},"numEdits":0,"identifiedLanguage":{"language":"en","probability":0.8397355675697327},"editors":["hyeoncho01"],"editorAvatarUrls":["/avatars/904d243f2fad99341f11795e93788993.svg"],"reactions":[],"isReport":false}}],"primaryEmailConfirmed":false,"paper":{"id":"2607.21485","authors":[{"_id":"6a62eabd2ee212ed0e2a1544","user":{"_id":"680ed019a9918bb1cbc66248","avatarUrl":"/avatars/904d243f2fad99341f11795e93788993.svg","isPro":false,"fullname":"Hyunmin Cho","user":"hyeoncho01","type":"user","name":"hyeoncho01"},"name":"Hyunmin Cho","status":"claimed_verified","statusLastChangedAt":"2026-07-24T08:45:04.316Z","hidden":false},{"_id":"6a62eabd2ee212ed0e2a1545","name":"Jaejun Yoo","hidden":false},{"_id":"6a62eabd2ee212ed0e2a1546","name":"Kyong Hwan Jin","hidden":false}],"publishedAt":"2026-07-23T00:00:00.000Z","submittedOnDailyAt":"2026-07-24T00:00:00.000Z","title":"Recurrent Sinusoidal INRs for Efficient High-Fidelity Representation","submittedOnDailyBy":{"_id":"680ed019a9918bb1cbc66248","avatarUrl":"/avatars/904d243f2fad99341f11795e93788993.svg","isPro":false,"fullname":"Hyunmin Cho","user":"hyeoncho01","type":"user","name":"hyeoncho01"},"summary":"We study sinusoidal recurrence as an iterative mechanism for harmonic spectral enrichment in implicit neural representations (INRs). Our analysis reveals that sinusoidal activations induce a harmonic line spectrum, providing a spectral account of how recurrent unrolling enriches the effective spectral support. We realize this principle with a shared sinusoidal block that iteratively refines the latent representation. We empirically validate the resulting spectral behavior against feed-forward INRs, non-sinusoidal recurrent variants, and equilibrium-style sinusoidal models. Complementing this analysis, we evaluate the proposed architecture across image and 3D representation tasks. On RGB image benchmarks, our method achieves higher fidelity than feed-forward baselines with fewer parameters and fewer optimization steps, and it further transfers favorably to super-resolution, NeRF, and SDF tasks.","upvotes":6,"discussionId":"6a62eabd2ee212ed0e2a1547","projectPage":"https://hyeon-cho.github.io/Harmonic-line-Spectrum/","githubRepo":"https://github.com/hyeon-cho/Harmonic-Siren","githubRepoAddedBy":"user","githubStars":0},"canReadDatabase":false,"canManagePapers":false,"canSubmit":false,"hasHfLevelAccess":false,"upvoted":false,"upvoters":[{"_id":"680ed019a9918bb1cbc66248","avatarUrl":"/avatars/904d243f2fad99341f11795e93788993.svg","isPro":false,"fullname":"Hyunmin Cho","user":"hyeoncho01","type":"user"},{"_id":"69ccfc64eb9cdf88f2a5f88c","avatarUrl":"/avatars/b9b9c773c730a66398a176220e42c05e.svg","isPro":false,"fullname":"Donghyeon Choi","user":"mydongh","type":"user"},{"_id":"691a8bc64679e966a59ec135","avatarUrl":"/avatars/3ed0e4f30790c01b305d4838d7b15d42.svg","isPro":false,"fullname":"suno","user":"suniverse77","type":"user"},{"_id":"66207af41c64c883eb865634","avatarUrl":"/avatars/167a6e19602877fa4d12e2e8070a0b1b.svg","isPro":false,"fullname":"grennkim","user":"greenx9","type":"user"},{"_id":"68ec7bb51dfe9ec7bd329d15","avatarUrl":"/avatars/552be99565053fb0a1b0d0fc895f5395.svg","isPro":false,"fullname":"Byeongho Moon","user":"bhomoon","type":"user"},{"_id":"6a158e5d0c9550b6e1325cec","avatarUrl":"/avatars/0eaf86a1a48e63e2c6a05e004fbd7371.svg","isPro":false,"fullname":"朱 雨桐","user":"aria-martinez43","type":"user"}],"acceptLanguages":["en"],"dailyPaperRank":0,"markdownContentUrl":"https://huggingface.co/buckets/huggingchat/papers-content/resolve/2607/2607.21485.md","query":{}}">

Get this paper in your agent:

Don't have the latest CLI?

Cite arxiv.org/abs/2607.21485 in a model README.md to link it from this page.

Cite arxiv.org/abs/2607.21485 in a dataset README.md to link it from this page.

Cite arxiv.org/abs/2607.21485 in a Space README.md to link it from this page.

Add this paper to a collection to link it from this page.