{"paper":{"title":"MoRe: Modular Representations for Principled Continual Representation Learning on Squantial Data","license":"http://creativecommons.org/licenses/by/4.0/","headline":"MoRe decomposes sequential representations into identifiable hierarchies of fundamental and specific modules to support continual adaptation while preserving prior knowledge by construction.","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Boyang Sun, Jiaqi Sun, Kun Zhang, Mohamad Rasmy, Xiangchen Song","submitted_at":"2026-05-14T04:46:54Z","abstract_excerpt":"Continual learning requires models to adapt to new data while preserving previously acquired knowledge. At its core, this challenge can be viewed as principled one-step adaptation: incorporating new information with minimal interference to existing representations. Most existing approaches address this challenge by modifying model parameters or architectures in a supervised, task-specific manner. However, the underlying issue is representational: tasks require distinct yet structured representations that can be selectively updated without disrupting representations, while structure should refl"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"MoRe decomposes knowledge into a hierarchy of fundamental and specific modules with identifiability guarantees, enabling principled module reuse, alignment, and expansion during adaptation while preserving old modules by construction.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That time-delayed dependencies in sequential data naturally reveal an intrinsic modular organization in representations that can be identified independently of task boundaries.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"MoRe decomposes representations into identifiable hierarchical modules to enable principled continual adaptation on sequential data.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"MoRe decomposes sequential representations into identifiable hierarchies of fundamental and specific modules to support continual adaptation while preserving prior knowledge by construction.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"ed44d33cd1e2a38c6cb4d77a6496b9b5d2b152d5be5fe0b62e9f218f37356d6f"},"source":{"id":"2605.14364","kind":"arxiv","version":1},"verdict":{"id":"69d79d71-4664-436c-92f8-bec6774fb496","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-15T02:52:29.009043Z","strongest_claim":"MoRe decomposes knowledge into a hierarchy of fundamental and specific modules with identifiability guarantees, enabling principled module reuse, alignment, and expansion during adaptation while preserving old modules by construction.","one_line_summary":"MoRe decomposes representations into identifiable hierarchical modules to enable principled continual adaptation on sequential data.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That time-delayed dependencies in sequential data naturally reveal an intrinsic modular organization in representations that can be identified independently of task boundaries.","pith_extraction_headline":"MoRe decomposes sequential representations into identifiable hierarchies of fundamental and specific modules to support continual adaptation while preserving prior knowledge by construction."},"references":{"count":43,"sample":[{"doi":"","year":2018,"title":"R. Aljundi, F. Babiloni, M. Elhoseiny, M. Rohrbach, and T. Tuytelaars. Memory aware synapses: Learning what (not) to forget. InProceedings of the European conference on computer vision (ECCV), pages 1","work_id":"bdcd9433-fe98-4094-97c7-84ddb016eaeb","ref_index":1,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2017,"title":"R. Aljundi, P. Chakravarty, and T. Tuytelaars. Expert gate: Lifelong learning with a network of experts. InProceedings of the IEEE conference on computer vision and pattern recognition, pages 3366–337","work_id":"a985de6b-b9ac-4b4f-a59c-3622f385ec28","ref_index":2,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2023,"title":"S. Biderman, H. Schoelkopf, Q. G. Anthony, H. Bradley, K. O’Brien, E. Hallahan, M. A. Khan, S. Purohit, U. S. Prashanth, E. Raff, et al. Pythia: A suite for analyzing large language models across trai","work_id":"b5a6c43d-67f2-4ea9-87a5-68389adf52d8","ref_index":3,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2023,"title":"W. Chen, Y . Zhou, N. Du, Y . Huang, J. Laudon, Z. Chen, and C. Cui. Lifelong language pretraining with distribution-specialized experts. InInternational Conference on Machine Learning, pages 5383–539","work_id":"6becdbcc-14e7-4b7a-ba18-1986b952264a","ref_index":4,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2017,"title":"PathNet: Evolution Channels Gradient Descent in Super Neural Networks","work_id":"67145523-6c66-41a2-a3b0-ceb37c0dc852","ref_index":5,"cited_arxiv_id":"1701.08734","is_internal_anchor":true}],"resolved_work":43,"snapshot_sha256":"8f29a60f134d5b29688be41f138d508ae0455e3b629446c4dacd71d303d6a405","internal_anchors":5},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"}