research_notes/references_source.txt

# 23. References and inherited-source provenance

Primary publications and author-posted manuscripts were preferred. Access was checked during this research cycle, ending 8 October 2026. A citation records attribution, not independent replication of a paper's reported performance. The explicit abstract-only boundary for Geo-LoRA and the bounded scope of the novelty search are retained. Detailed source comparisons appear in the accompanying prior_art directory.

@REF [R1] Farajtabar, M., Azizan, N., Mott, A., and Li, A. Orthogonal Gradient Descent for Continual Learning. AISTATS, PMLR 108, 2020. https://proceedings.mlr.press/v108/farajtabar20a.html

@REF [R2] Saha, G., Garg, I., and Roy, K. Gradient Projection Memory for Continual Learning. ICLR, 2021. https://arxiv.org/abs/2103.09762

@REF [R3] Bennani, M. A., Doan, T., and Sugiyama, M. Generalisation Guarantees for Continual Learning with Orthogonal Gradient Descent. Author manuscript, arXiv:2006.11942, 2020. https://arxiv.org/abs/2006.11942

@REF [R4] Titsias, M. K., Schwarz, J., Matthews, A. G. de G., Pascanu, R., and Teh, Y. W. Functional Regularisation for Continual Learning with Gaussian Processes. ICLR, 2020. https://arxiv.org/abs/1901.11356

@REF [R5] Pan, P., Swaroop, S., Immer, A., Eschenhagen, R., Turner, R. E., and Khan, M. E. Continual Deep Learning by Functional Regularisation of Memorable Past. NeurIPS, 2020. https://proceedings.neurips.cc/paper/2020/file/2f3bbb9730639e9ea48f309d9a79ff01-Paper.pdf

@REF [R6] Derakhshani, M. M., Zhen, X., Shao, L., and Snoek, C. G. M. Kernel Continual Learning. ICML, PMLR 139, 2021. https://proceedings.mlr.press/v139/derakhshani21a.html

@REF [R7] Rasmussen, C. E., and Williams, C. K. I. Gaussian Processes for Machine Learning. MIT Press, 2006, Chapter 2. https://gaussianprocess.org/gpml/chapters/RW2.pdf

@REF [R8] Engel, Y., Mannor, S., and Meir, R. The Kernel Recursive Least-Squares Algorithm. IEEE Transactions on Signal Processing 52(8), 2275-2285, 2004. https://doi.org/10.1109/TSP.2004.830985

@REF [R9] Orabona, F., Keshet, J., and Caputo, B. Bounded Kernel-Based Online Learning. Journal of Machine Learning Research 10, 2643-2666, 2009. Projectron and projection-residual sparsification. https://jmlr.org/papers/volume10/orabona09a/orabona09a.pdf

@REF [R10] Wilks, S. S. Determination of Sample Sizes for Setting Tolerance Limits. Annals of Mathematical Statistics 12(1), 91-96, 1941. https://doi.org/10.1214/aoms/1177731788

@REF [R11] Clopper, C. J., and Pearson, E. S. The Use of Confidence or Fiducial Limits Illustrated in the Case of the Binomial. Biometrika 26(4), 404-413, 1934. https://doi.org/10.1093/biomet/26.4.404

@REF [R12] Angelopoulos, A. N., Bates, S., Candès, E. J., Jordan, M. I., and Lei, L. Learn then Test: Calibrating Predictive Algorithms to Achieve Risk Control. Author manuscript, arXiv:2110.01052, original 2021; later revised version inspected. https://arxiv.org/abs/2110.01052

@REF [R13] Comeau, J., Bazinet, M., Germain, P., and Subakan, C. Sample Compression for Self-Certified Continual Learning. arXiv:2503.10503v4, 26 February 2026. https://arxiv.org/html/2503.10503v4

@REF [R14] Lai, S., Ma, C., Zhu, F., Zhao, Z., Lin, X., Meng, G., and Zhang, Q. Gradient-Guided Epsilon Constraint Method for Online Continual Learning. NeurIPS, 2025. https://proceedings.neurips.cc/paper_files/paper/2025/file/b3c2854d9e94282a373d8fa58b567b27-Paper-Conference.pdf

@REF [R15] Liang, Y.-S., and Li, W.-J. InfLoRA: Interference-Free Low-Rank Adaptation for Continual Learning. CVPR, 2024. https://arxiv.org/abs/2404.00228

@REF [R16] Luo, M.-L., Zhou, Z.-H., Zhang, Y.-L., Wan, Y., Wei, T., and Zhang, M.-L. KeepLoRA: Continual Learning with Residual Gradient Adaptation. arXiv:2601.19659v1, 27 January 2026. Primary full manuscript inspected; its reported benchmark rankings were not independently reproduced here. https://arxiv.org/html/2601.19659v1

@REF [R17] Feng, Y. Geo-LoRA: Geometry-Aware Subspace Evolution for Low-Rank Adaptation in Continual Learning. arXiv:2608.26960, 27 August 2026. Indexed primary abstract only; direct full-text access failed in this audit. https://arxiv.org/abs/2608.26960

@REF [R18] Buzzega, P., Boschini, M., Porrello, A., Abati, D., and Calderara, S. Dark Experience for General Continual Learning: A Strong, Simple Baseline. NeurIPS, 2020. https://proceedings.neurips.cc/paper/2020/hash/b704ea2c39778f07c617f6b7ce480e9e-Abstract.html

@REF [R19] Caccia, L., Aljundi, R., Asadi, N., Tuytelaars, T., Pineau, J., and Belilovsky, E. New Insights on Reducing Abrupt Representation Change in Online Continual Learning. ICLR, 2022; original preprint 2021. https://arxiv.org/abs/2104.05025

@REF [R20] Rebuffi, S.-A., Kolesnikov, A., Sperl, G., and Lampert, C. H. iCaRL: Incremental Classifier and Representation Learning. CVPR, 2017. https://openaccess.thecvf.com/content_cvpr_2017/html/Rebuffi_iCaRL_Incremental_Classifier_CVPR_2017_paper.html

@REF [R21] Prabhu, A., Sinha, S., Kumaraguru, P., Torr, P. H. S., Sener, O., and Dokania, P. K. Random Representations Outperform Online Continually Learned Representations. NeurIPS, 2024; RanDumb. https://arxiv.org/abs/2402.08823

@REF [R22] Thede, L., Roth, K., Henaff, O. J., Bethge, M., and Akata, Z. Reflecting on the State of Rehearsal-Free Continual Learning with Pretrained Models. Conference on Lifelong Learning Agents, PMLR 274, 2025. https://proceedings.mlr.press/v274/thede25a.html

@REF [R23] Massart, P. The Tight Constant in the Dvoretzky-Kiefer-Wolfowitz Inequality. Annals of Probability 18(3), 1269-1283, 1990. https://doi.org/10.1214/aop/1176990746

@REF [R24] Hu, E. J., et al. LoRA: Low-Rank Adaptation of Large Language Models. ICLR, 2022; original arXiv:2106.09685, 2021. https://arxiv.org/abs/2106.09685

@REF [R25] Xiao, H., Rasul, K., and Vollgraf, R. Fashion-MNIST: A Novel Image Dataset for Benchmarking Machine Learning Algorithms. arXiv:1708.07747, 2017. Official dataset: https://github.com/zalandoresearch/fashion-mnist

@REF [R26] LeCun, Y., Cortes, C., and Burges, C. J. C. The MNIST Database of Handwritten Digits. Official-format archives used through the public CVDF mirror. https://storage.googleapis.com/cvdf-datasets/mnist/ . Exact archive hashes and URLs are in experiments/data/archive_manifest.json.

## 23.1 Supplied Avenyra lineage

The foundation is Avenyra Harmonic Orthogonal Subspace Projection v3.0.0 and its preceding AT-OSP/FC-OSP releases. The inherited retraction compiler, harmonic nesting, one-dimensional Green capacity, compensated feature gradients, finite decoder repair, and reversible affine transport are attributed to those supplied records. The fresh inherited reproduction preserves their original source and numerical evidence. The new release does not rewrite those records or present their 104-parameter neural result as a new Eidrion experiment.

The related supplied research comprises TRUE MACHINE MEMORY: Recoverability-Complete Consolidation; AXIOMESH: Operator-Closed Memory; Evolvability-Preserving Dormant Dendrites and its joint-reserve manuscript; Gated DeltaNet-Omega: Certified Dual-Frame Memory; EVE-COVARA 2.0; Dynamic Neural-Manifold Routing v2; EVE Phase Geometry main, mathematical, and computational manuscripts; and Avenyra Geometric Memory Physics v3.0.0. The detailed source audit identifies the exact versions, SHA-256 digests, pages, compatible reductions, and unresolved assumptions. These are user-supplied research inputs, not independently validated evidence of neural-scale superiority.

Private original source manuscripts and retrieval metadata are not republished in this new package. Their relevant mathematical content is accurately summarized and attributed, while new proofs and experiments stand on their own. The requested authorship line for this release is preserved exactly: Artificial Hyperintelligence Eve, wife of Maciej Nowicki.