{"paper":{"title":"QLAM: A Quantum Long-Attention Memory Approach to Long-Sequence Token Modeling","license":"http://creativecommons.org/licenses/by/4.0/","headline":"QLAM represents sequence memory as a quantum superposition state evolved by input-conditioned circuits to capture global dependencies in linear time.","cross_cats":["cs.CV"],"primary_cat":"cs.LG","authors_text":"Hoang-Quan Nguyen, Khoa Luu, Sankalp Pandey","submitted_at":"2026-05-13T17:56:20Z","abstract_excerpt":"Modeling long-range dependencies in sequential data remains a central challenge in machine learning. Transformers address this challenge through attention mechanisms, but their quadratic complexity with respect to sequence length limits scalability to long contexts. State-space models (SSMs) provide an efficient alternative with linear-time computation by evolving a latent state through recurrent updates, but their memory is typically formed via additive or linear transitions, which can limit their ability to capture complex global interactions across tokens. In this work, we introduce one of "},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"Across all tasks, QLAM consistently improves over recurrent baselines and transformer-based models.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That the parameterized quantum circuits can evolve the superposition state to capture complex global token interactions more effectively than classical additive or linear transitions, and that this advantage can be realized at practical simulation cost.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"QLAM extends state-space models with quantum superposition in the hidden state for linear-time long-sequence modeling and reports consistent gains over RNN and transformer baselines on sequential image tasks.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"QLAM represents sequence memory as a quantum superposition state evolved by input-conditioned circuits to capture global dependencies in linear time.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"fcb742a48f7a1d455ee76b1595877d7a793e9f71043db27a8addd2251f600eb1"},"source":{"id":"2605.13833","kind":"arxiv","version":1},"verdict":{"id":"9cc8b76f-d4ee-4b43-bccd-5f22a76edd03","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-14T19:07:13.496227Z","strongest_claim":"Across all tasks, QLAM consistently improves over recurrent baselines and transformer-based models.","one_line_summary":"QLAM extends state-space models with quantum superposition in the hidden state for linear-time long-sequence modeling and reports consistent gains over RNN and transformer baselines on sequential image tasks.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That the parameterized quantum circuits can evolve the superposition state to capture complex global token interactions more effectively than classical additive or linear transitions, and that this advantage can be realized at practical simulation cost.","pith_extraction_headline":"QLAM represents sequence memory as a quantum superposition state evolved by input-conditioned circuits to capture global dependencies in linear time."},"references":{"count":62,"sample":[{"doi":"","year":1997,"title":"Long short-term memory","work_id":"907ceddc-90b3-4a1b-949d-86b07efe30ca","ref_index":1,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":1994,"title":"Learning long-term dependencies with gradient descent is difficult,","work_id":"5ffe9241-a117-4d44-b88a-d63159d1f954","ref_index":2,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2017,"title":"Attention is all you need","work_id":"4f585692-8ef8-4f20-8f75-5e0e32647d76","ref_index":3,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":1904,"title":"Generating Long Sequences with Sparse Transformers","work_id":"c5b81688-45ee-4a9a-b095-e6290f45cb6c","ref_index":4,"cited_arxiv_id":"1904.10509","is_internal_anchor":true},{"doi":"","year":2020,"title":"Transformers are rnns: Fast autoregressive transformers with linear attention,","work_id":"009df75c-9fa0-41b2-afb9-de7d506cc950","ref_index":5,"cited_arxiv_id":"","is_internal_anchor":false}],"resolved_work":62,"snapshot_sha256":"453523713579c668c6b544a705c385b72a5c0fe3a9bb44652744bb2c82fbc5e9","internal_anchors":11},"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"}