{"paper":{"title":"A Unified Perspective for Learning Graph Representations Across Multi-Level Abstractions","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"A unified contrastive framework learns graph representations by linearly combining node, proximity, cluster, and graph level signals with a parameter-free self-weighting mechanism.","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Abdoulaye Banir\\'e Diallo, Mohamed Bouguessa, Mohamed Mahmoud Amar, Nairouz Mrabah","submitted_at":"2026-05-12T19:33:39Z","abstract_excerpt":"Graph Self-Supervised Learning (GSSL) has emerged as a powerful paradigm for generating high-quality representations for graph-structured data. While multi-scale graph contrastive learning has received increasing attention, many existing methods still predominantly focus on a single graph abstraction level. To address this limitation, we propose a unified contrastive framework that can target node-level, proximity-level, cluster-level, and graph-level information and integrate them through a linear combination of similarity scores on positive pairs and dissimilarity scores (i.e., similarity sc"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"Our approach not only enhances optimization flexibility but also eliminates the computational overhead of hyperparameter tuning in conventional multi-task GSSL methods. Comprehensive experiments on real-world datasets show that our methods consistently outperform state-of-the-art approaches across downstream tasks, including classification, clustering, and link prediction, in both single-level and multi-level scenarios.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That a linear combination of per-level similarity and dissimilarity scores, modulated by the proposed self-weighting, captures complementary multi-level information without destructive interference or the need for level-specific tuning.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"A multi-level graph contrastive framework with adaptive self-weighting outperforms prior single-level and multi-task GSSL methods on classification, clustering, and link prediction.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"A unified contrastive framework learns graph representations by linearly combining node, proximity, cluster, and graph level signals with a parameter-free self-weighting mechanism.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"4165d5c4a16d4ee6835e5764ee633c407ca175c63de606d7d8ef7c38558d9851"},"source":{"id":"2605.12685","kind":"arxiv","version":1},"verdict":{"id":"0607749a-e518-4620-a903-8109b68471b4","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-14T21:05:19.597474Z","strongest_claim":"Our approach not only enhances optimization flexibility but also eliminates the computational overhead of hyperparameter tuning in conventional multi-task GSSL methods. Comprehensive experiments on real-world datasets show that our methods consistently outperform state-of-the-art approaches across downstream tasks, including classification, clustering, and link prediction, in both single-level and multi-level scenarios.","one_line_summary":"A multi-level graph contrastive framework with adaptive self-weighting outperforms prior single-level and multi-task GSSL methods on classification, clustering, and link prediction.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That a linear combination of per-level similarity and dissimilarity scores, modulated by the proposed self-weighting, captures complementary multi-level information without destructive interference or the need for level-specific tuning.","pith_extraction_headline":"A unified contrastive framework learns graph representations by linearly combining node, proximity, cluster, and graph level signals with a parameter-free self-weighting mechanism."},"references":{"count":63,"sample":[{"doi":"","year":2025,"title":"Re- thinking deep clustering paradigms: Self-supervision is all you need,","work_id":"ccfd593a-e0d1-4ad3-81c2-68eadb62ea0d","ref_index":1,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2020,"title":"Big self-supervised models are strong semi-supervised learners,","work_id":"513d3240-0fb8-405d-8153-170946d79204","ref_index":2,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2023,"title":"Toward convex manifolds: A geometric perspective for deep graph clus- tering of single-cell rna-seq data","work_id":"1a78b8db-4cf3-49c6-8d71-ba8a43ef62e2","ref_index":3,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2023,"title":"Exploring the interaction between local and global latent configurations for clustering single-cell rna-seq: a unified per- spective,","work_id":"f7f39add-2782-41a9-b4e6-5575c165a68c","ref_index":4,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2022,"title":"Graph self-supervised learning: A survey,","work_id":"9ae3777f-5e13-46dd-a82b-33f6487f25c1","ref_index":5,"cited_arxiv_id":"","is_internal_anchor":false}],"resolved_work":63,"snapshot_sha256":"9e193197d17575f88b1bd1bfa636ad0653033eb3cbcfb3b4dd028ff3f72a473f","internal_anchors":1},"formal_canon":{"evidence_count":2,"snapshot_sha256":"62748dd041f8adec6ffbb31f47a6e727c5a0f5c1a86bd5cfeb126bce6da81280"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"}