{"id":"arxiv:2511.08544","title":"LeJEPA: Provable and Scalable Self-Supervised Learning Without the Heuristics","date":"2025-11-11","section":"Self-supervised learning and vision","quality":0.332039,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/e2c5835a26dabc321e6364380a37c47c10425ee830ec522564b0604733025a94/","source_url":"https://arxiv.org/abs/2511.08544","text_available":true,"authors":["Randall Balestriero","Yann LeCun"],"outgoing":[{"source":"arxiv:2511.08544","target":"arxiv:2104.14294","evidence":"title","id":"arxiv:2104.14294","title":"Emerging Properties in Self-Supervised Vision Transformers","url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/4372186d141ca514b08e93bb3b1e26fdbcba9ce62187fb772c14bf223160c06d/"},{"source":"arxiv:2511.08544","target":"arxiv:2105.04906","evidence":"id","id":"arxiv:2105.04906","title":"VICReg: Variance-Invariance-Covariance Regularization for Self-Supervised Learning","url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/822b6414fb02bd6c746e0d559fd3dcc60766476142ebae7795b02bc6c45223ce/"},{"source":"arxiv:2511.08544","target":"arxiv:2301.08243","evidence":"title","id":"arxiv:2301.08243","title":"Self-Supervised Learning from Images with a Joint-Embedding Predictive Architecture","url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/475bac0a3b29edd42642bbe10d1a41bfb18a3022ac95b9b1a3db9e5d5d0b95fe/"},{"source":"arxiv:2511.08544","target":"arxiv:2304.07193","evidence":"id","id":"arxiv:2304.07193","title":"DINOv2: Learning Robust Visual Features without Supervision","url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/b15933596df35497024556ae88bfa696f3b0e8372b678194f04f1bfb9f3be553/"},{"source":"arxiv:2511.08544","target":"arxiv:2304.09355","evidence":"title","id":"arxiv:2304.09355","title":"To Compress or Not to Compress- Self-Supervised Learning and Information Theory: A Review","url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/6de01284dffdeadbd92020fc614ff702f27bc9c49215a1658ea3ca8a4a0c36ce/"},{"source":"arxiv:2511.08544","target":"arxiv:2304.12210","evidence":"id","id":"arxiv:2304.12210","title":"A Cookbook of Self-Supervised Learning","url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/518a026819227af3e1838911bfadf26c1515786a3af305de5c2e77efb052ecfd/"}],"incoming":[{"source":"arxiv:2605.26379","target":"arxiv:2511.08544","evidence":"id","id":"arxiv:2605.26379","title":"When Does LeJEPA Learn a World Model?","url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/f140b8fcc13ea085ce6ad170e090380c8d373416f791960778c4b216f6e5c8f6/"},{"source":"arxiv:2606.02572","target":"arxiv:2511.08544","evidence":"id","id":"arxiv:2606.02572","title":"VISReg: Variance-Invariance-Sketching Regularization for JEPA training","url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/262c1938984761f99c55d9ae1237e686b2d6ab94b69d7181d11dceab2a7210fb/"}],"text_parts":[{"start":0,"end":15344,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/e2c5835a26dabc321e6364380a37c47c10425ee830ec522564b0604733025a94/part-1.html"},{"start":15344,"end":29910,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/e2c5835a26dabc321e6364380a37c47c10425ee830ec522564b0604733025a94/part-2.html"},{"start":29910,"end":45369,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/e2c5835a26dabc321e6364380a37c47c10425ee830ec522564b0604733025a94/part-3.html"},{"start":45369,"end":60506,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/e2c5835a26dabc321e6364380a37c47c10425ee830ec522564b0604733025a94/part-4.html"},{"start":60506,"end":74849,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/e2c5835a26dabc321e6364380a37c47c10425ee830ec522564b0604733025a94/part-5.html"},{"start":74849,"end":90244,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/e2c5835a26dabc321e6364380a37c47c10425ee830ec522564b0604733025a94/part-6.html"},{"start":90244,"end":103101,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/e2c5835a26dabc321e6364380a37c47c10425ee830ec522564b0604733025a94/part-7.html"},{"start":103101,"end":117672,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/e2c5835a26dabc321e6364380a37c47c10425ee830ec522564b0604733025a94/part-8.html"},{"start":117672,"end":132698,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/e2c5835a26dabc321e6364380a37c47c10425ee830ec522564b0604733025a94/part-9.html"},{"start":132698,"end":139605,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/e2c5835a26dabc321e6364380a37c47c10425ee830ec522564b0604733025a94/part-10.html"}],"contents":[{"title":"1 Introduction","offset":2937,"level":2,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/e2c5835a26dabc321e6364380a37c47c10425ee830ec522564b0604733025a94/part-1.html#text"},{"title":"2 Background and Notations","offset":9690,"level":2,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/e2c5835a26dabc321e6364380a37c47c10425ee830ec522564b0604733025a94/part-1.html#text"},{"title":"2.1 Notations and Definitions","offset":9938,"level":3,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/e2c5835a26dabc321e6364380a37c47c10425ee830ec522564b0604733025a94/part-1.html#text"},{"title":"2.2 The Need for Reliable Pretraining","offset":14049,"level":3,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/e2c5835a26dabc321e6364380a37c47c10425ee830ec522564b0604733025a94/part-1.html#text"},{"title":"2.3 The Need for Actionable Theory","offset":15304,"level":3,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/e2c5835a26dabc321e6364380a37c47c10425ee830ec522564b0604733025a94/part-1.html#text"},{"title":"3 Latent Euclidean: Embeddings Should be Isotropic Gaussian","offset":17969,"level":2,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/e2c5835a26dabc321e6364380a37c47c10425ee830ec522564b0604733025a94/part-2.html#text"},{"title":"3.1 Linear Probing","offset":18528,"level":3,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/e2c5835a26dabc321e6364380a37c47c10425ee830ec522564b0604733025a94/part-2.html#text"},{"title":"3.2 Nonlinear Probing","offset":20692,"level":3,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/e2c5835a26dabc321e6364380a37c47c10425ee830ec522564b0604733025a94/part-2.html#text"},{"title":"3.3 Geometric and Practical Insights","offset":23161,"level":3,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/e2c5835a26dabc321e6364380a37c47c10425ee830ec522564b0604733025a94/part-2.html#text"},{"title":"4 SIGReg: Reliable Isotropic Gaussian Regularization in High-Dimension","offset":24724,"level":2,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/e2c5835a26dabc321e6364380a37c47c10425ee830ec522564b0604733025a94/part-2.html#text"},{"title":"4.1 Hypothesis Testing as a Judge","offset":25554,"level":3,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/e2c5835a26dabc321e6364380a37c47c10425ee830ec522564b0604733025a94/part-2.html#text"},{"title":"4.2 SIGReg: Sketching the Epps-Pulley Test is Stable and Scalable","offset":29065,"level":3,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/e2c5835a26dabc321e6364380a37c47c10425ee830ec522564b0604733025a94/part-2.html#text"},{"title":"4.2.1 Moments are Unstable and Insufficient","offset":29860,"level":4,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/e2c5835a26dabc321e6364380a37c47c10425ee830ec522564b0604733025a94/part-2.html#text"},{"title":"4.2.2 Cumulative Density Functions are Impractical","offset":33355,"level":4,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/e2c5835a26dabc321e6364380a37c47c10425ee830ec522564b0604733025a94/part-3.html#text"},{"title":"4.2.3 Characteristic Functions are Stable, Scalable and Identifiable","offset":35624,"level":4,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/e2c5835a26dabc321e6364380a37c47c10425ee830ec522564b0604733025a94/part-3.html#text"},{"title":"4.3 How SIGReg Beats the Curse of Dimensionality","offset":39025,"level":3,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/e2c5835a26dabc321e6364380a37c47c10425ee830ec522564b0604733025a94/part-3.html#text"},{"title":"Smoothness Beats the Curse of Dimensionality","offset":39326,"level":4,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/e2c5835a26dabc321e6364380a37c47c10425ee830ec522564b0604733025a94/part-3.html#text"},{"title":"SGD Beats the Curse of Dimensionality","offset":41294,"level":4,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/e2c5835a26dabc321e6364380a37c47c10425ee830ec522564b0604733025a94/part-3.html#text"},{"title":"Empirical Validation on Synthetic Data","offset":42025,"level":4,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/e2c5835a26dabc321e6364380a37c47c10425ee830ec522564b0604733025a94/part-3.html#text"},{"title":"5 LeJEPA: Stable and Scalable Implementation","offset":43388,"level":2,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/e2c5835a26dabc321e6364380a37c47c10425ee830ec522564b0604733025a94/part-3.html#text"},{"title":"5.1 LeJEPA: SIGReg + Prediction Loss","offset":43909,"level":3,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/e2c5835a26dabc321e6364380a37c47c10425ee830ec522564b0604733025a94/part-3.html#text"},{"title":"5.2 Relation to Prior Work","offset":47684,"level":3,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/e2c5835a26dabc321e6364380a37c47c10425ee830ec522564b0604733025a94/part-4.html#text"},{"title":"6 LeJEPA: Empirical Validation","offset":49782,"level":2,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/e2c5835a26dabc321e6364380a37c47c10425ee830ec522564b0604733025a94/part-4.html#text"},{"title":"6.1 LeJEPA’s Stability Across Hyper-Parameters and Architectures","offset":50392,"level":3,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/e2c5835a26dabc321e6364380a37c47c10425ee830ec522564b0604733025a94/part-4.html#text"},{"title":"6.2 LeJEPA’s Training Loss is Informative of Downstream Performance","offset":55109,"level":3,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/e2c5835a26dabc321e6364380a37c47c10425ee830ec522564b0604733025a94/part-4.html#text"},{"title":"6.3 In-Domain LeJEPA Outperforms Frontier Model Transfer Learning","offset":57313,"level":3,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/e2c5835a26dabc321e6364380a37c47c10425ee830ec522564b0604733025a94/part-4.html#text"},{"title":"6.4 LeJEPA Scales Across Data and Models","offset":59187,"level":3,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/e2c5835a26dabc321e6364380a37c47c10425ee830ec522564b0604733025a94/part-4.html#text"},{"title":"6.5 Emergent Semantic Structure in LeJEPA Representations","offset":60443,"level":3,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/e2c5835a26dabc321e6364380a37c47c10425ee830ec522564b0604733025a94/part-4.html#text"},{"title":"7 Conclusion","offset":61709,"level":2,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/e2c5835a26dabc321e6364380a37c47c10425ee830ec522564b0604733025a94/part-5.html#text"},{"title":"Acknowledgments","offset":62685,"level":2,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/e2c5835a26dabc321e6364380a37c47c10425ee830ec522564b0604733025a94/part-5.html#text"},{"title":"Appendix A Additional Details on Nonlinear Probing","offset":62808,"level":2,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/e2c5835a26dabc321e6364380a37c47c10425ee830ec522564b0604733025a94/part-5.html#text"},{"title":"A.1 kNN Probing","offset":62863,"level":3,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/e2c5835a26dabc321e6364380a37c47c10425ee830ec522564b0604733025a94/part-5.html#text"},{"title":"A.2 Kernel Probing","offset":65560,"level":3,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/e2c5835a26dabc321e6364380a37c47c10425ee830ec522564b0604733025a94/part-5.html#text"},{"title":"Appendix B Proofs","offset":68371,"level":2,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/e2c5835a26dabc321e6364380a37c47c10425ee830ec522564b0604733025a94/part-5.html#text"},{"title":"B.1 Proof of Section 3.1","offset":68393,"level":3,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/e2c5835a26dabc321e6364380a37c47c10425ee830ec522564b0604733025a94/part-5.html#text"},{"title":"Proof.","offset":68423,"level":6,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/e2c5835a26dabc321e6364380a37c47c10425ee830ec522564b0604733025a94/part-5.html#text"},{"title":"B.2 Proof of Section 3.1","offset":71474,"level":3,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/e2c5835a26dabc321e6364380a37c47c10425ee830ec522564b0604733025a94/part-5.html#text"},{"title":"Proof.","offset":71504,"level":6,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/e2c5835a26dabc321e6364380a37c47c10425ee830ec522564b0604733025a94/part-5.html#text"},{"title":"B.3 Proof of Section A.1","offset":73940,"level":3,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/e2c5835a26dabc321e6364380a37c47c10425ee830ec522564b0604733025a94/part-5.html#text"},{"title":"Proof.","offset":73970,"level":6,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/e2c5835a26dabc321e6364380a37c47c10425ee830ec522564b0604733025a94/part-5.html#text"},{"title":"B.4 Proof of Section A.1","offset":79709,"level":3,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/e2c5835a26dabc321e6364380a37c47c10425ee830ec522564b0604733025a94/part-6.html#text"},{"title":"Proof.","offset":79739,"level":6,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/e2c5835a26dabc321e6364380a37c47c10425ee830ec522564b0604733025a94/part-6.html#text"},{"title":"1) The score-gradient term 𝔼 ⁡ [ A ​ ( X ) 2 ] \\mathbb{E}[A(X)^{2}] .","offset":81661,"level":5,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/e2c5835a26dabc321e6364380a37c47c10425ee830ec522564b0604733025a94/part-6.html#text"},{"title":"2) The cross term 2 ​ 𝔼 ​ [ A ⁡ ( X ) ​ C ​ ( X ) ] 2\\mathbb{E}[A(X)C(X)] .","offset":83197,"level":5,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/e2c5835a26dabc321e6364380a37c47c10425ee830ec522564b0604733025a94/part-6.html#text"},{"title":"3) The curvature term 𝔼 ⁡ [ C ​ ( X ) 2 ] \\mathbb{E}[C(X)^{2}] .","offset":84283,"level":5,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/e2c5835a26dabc321e6364380a37c47c10425ee830ec522564b0604733025a94/part-6.html#text"},{"title":"Putting it together.","offset":84757,"level":5,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/e2c5835a26dabc321e6364380a37c47c10425ee830ec522564b0604733025a94/part-6.html#text"},{"title":"Proof.","offset":86018,"level":6,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/e2c5835a26dabc321e6364380a37c47c10425ee830ec522564b0604733025a94/part-6.html#text"},{"title":"Step 2: Optimizing over covariance shapes under scalar constraints","offset":87302,"level":3,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/e2c5835a26dabc321e6364380a37c47c10425ee830ec522564b0604733025a94/part-6.html#text"},{"title":"(a) Trace constraint.","offset":87760,"level":5,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/e2c5835a26dabc321e6364380a37c47c10425ee830ec522564b0604733025a94/part-6.html#text"},{"title":"(b) Determinant constraint.","offset":88394,"level":5,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/e2c5835a26dabc321e6364380a37c47c10425ee830ec522564b0604733025a94/part-6.html#text"},{"title":"(c) Frobenius-norm constraint.","offset":89128,"level":5,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/e2c5835a26dabc321e6364380a37c47c10425ee830ec522564b0604733025a94/part-6.html#text"},{"title":"(d) Spectral-radius constraint.","offset":90205,"level":5,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/e2c5835a26dabc321e6364380a37c47c10425ee830ec522564b0604733025a94/part-6.html#text"},{"title":"Conclusion: Isotropic Gaussian is optimal","offset":90982,"level":3,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/e2c5835a26dabc321e6364380a37c47c10425ee830ec522564b0604733025a94/part-7.html#text"},{"title":"Proof.","offset":91089,"level":6,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/e2c5835a26dabc321e6364380a37c47c10425ee830ec522564b0604733025a94/part-7.html#text"},{"title":"B.5 Proof of Section A.2","offset":91494,"level":3,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/e2c5835a26dabc321e6364380a37c47c10425ee830ec522564b0604733025a94/part-7.html#text"},{"title":"Proof.","offset":91524,"level":6,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/e2c5835a26dabc321e6364380a37c47c10425ee830ec522564b0604733025a94/part-7.html#text"},{"title":"B.6 Proof of Equation 6 to Equation 7","offset":97250,"level":3,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/e2c5835a26dabc321e6364380a37c47c10425ee830ec522564b0604733025a94/part-7.html#text"},{"title":"Proof.","offset":97293,"level":6,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/e2c5835a26dabc321e6364380a37c47c10425ee830ec522564b0604733025a94/part-7.html#text"},{"title":"B.7 Proof of Section A.2","offset":100452,"level":3,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/e2c5835a26dabc321e6364380a37c47c10425ee830ec522564b0604733025a94/part-7.html#text"},{"title":"Proof.","offset":100482,"level":6,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/e2c5835a26dabc321e6364380a37c47c10425ee830ec522564b0604733025a94/part-7.html#text"},{"title":"B.8 Proof of Section 4.1","offset":102909,"level":3,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/e2c5835a26dabc321e6364380a37c47c10425ee830ec522564b0604733025a94/part-7.html#text"},{"title":"Proof.","offset":102939,"level":6,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/e2c5835a26dabc321e6364380a37c47c10425ee830ec522564b0604733025a94/part-7.html#text"},{"title":"B.9 Proof of Section 4.1","offset":106422,"level":3,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/e2c5835a26dabc321e6364380a37c47c10425ee830ec522564b0604733025a94/part-8.html#text"},{"title":"Proof.","offset":106452,"level":6,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/e2c5835a26dabc321e6364380a37c47c10425ee830ec522564b0604733025a94/part-8.html#text"},{"title":"B.10 Proof of Section 4.3","offset":108503,"level":3,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/e2c5835a26dabc321e6364380a37c47c10425ee830ec522564b0604733025a94/part-8.html#text"},{"title":"Proof.","offset":108534,"level":6,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/e2c5835a26dabc321e6364380a37c47c10425ee830ec522564b0604733025a94/part-8.html#text"},{"title":"B.11 Proof of Section 4.2.1","offset":111695,"level":3,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/e2c5835a26dabc321e6364380a37c47c10425ee830ec522564b0604733025a94/part-8.html#text"},{"title":"B.12 Proof of Section 4.2.3","offset":112611,"level":3,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/e2c5835a26dabc321e6364380a37c47c10425ee830ec522564b0604733025a94/part-8.html#text"},{"title":"Proof.","offset":112644,"level":6,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/e2c5835a26dabc321e6364380a37c47c10425ee830ec522564b0604733025a94/part-8.html#text"},{"title":"B.13 Proof of Section 5.1","offset":117626,"level":3,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/e2c5835a26dabc321e6364380a37c47c10425ee830ec522564b0604733025a94/part-8.html#text"},{"title":"Proof.","offset":117657,"level":6,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/e2c5835a26dabc321e6364380a37c47c10425ee830ec522564b0604733025a94/part-8.html#text"},{"title":"B.14 Proof of VICReg’s Recovery","offset":122922,"level":3,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/e2c5835a26dabc321e6364380a37c47c10425ee830ec522564b0604733025a94/part-9.html#text"},{"title":"Proof.","offset":122959,"level":6,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/e2c5835a26dabc321e6364380a37c47c10425ee830ec522564b0604733025a94/part-9.html#text"},{"title":"Part I: 𝔼 ⁡ [ 𝐗 ] = 𝟎 \\mathbb{E}[\\mathbf{X}]=\\mathbf{0}","offset":123010,"level":5,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/e2c5835a26dabc321e6364380a37c47c10425ee830ec522564b0604733025a94/part-9.html#text"},{"title":"Part II: Cov ⁡ ( 𝐗 ) = 𝐈 d \\mathrm{Cov}(\\mathbf{X})=\\mathbf{I}_{d}","offset":124480,"level":5,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/e2c5835a26dabc321e6364380a37c47c10425ee830ec522564b0604733025a94/part-9.html#text"},{"title":"Appendix C Background","offset":127235,"level":2,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/e2c5835a26dabc321e6364380a37c47c10425ee830ec522564b0604733025a94/part-9.html#text"},{"title":"Appendix D Details on Low-Discrepancy Sequences","offset":128396,"level":2,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/e2c5835a26dabc321e6364380a37c47c10425ee830ec522564b0604733025a94/part-9.html#text"},{"title":"Appendix E Shapiro-Wilk Test","offset":129369,"level":2,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/e2c5835a26dabc321e6364380a37c47c10425ee830ec522564b0604733025a94/part-9.html#text"},{"title":"Appendix F Multivariate Statistics","offset":132659,"level":2,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/e2c5835a26dabc321e6364380a37c47c10425ee830ec522564b0604733025a94/part-9.html#text"},{"title":"Instructions for reporting errors","offset":138641,"level":2,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/e2c5835a26dabc321e6364380a37c47c10425ee830ec522564b0604733025a94/part-10.html#text"}],"text_code_points":139605,"source_markdown":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/e2c5835a26dabc321e6364380a37c47c10425ee830ec522564b0604733025a94/source.md"}