{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:6ZO2YXP5DGVOU44MRJOXTZGTUP","short_pith_number":"pith:6ZO2YXP5","schema_version":"1.0","canonical_sha256":"f65dac5dfd19aaea738c8a5d79e4d3a3d9df55bc002a88ad7aba4b9ed41ba963","source":{"kind":"arxiv","id":"2410.17897","version":5},"attestation_state":"computed","paper":{"title":"Value Residual Learning","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Fares Obeid, Tianyi Wu, Zhanchao Zhou, Zhenzhong Lan, Zhiyun Jiang","submitted_at":"2024-10-23T14:15:07Z","abstract_excerpt":"While Transformer models have achieved remarkable success in various domains, the effectiveness of information propagation through deep networks remains a critical challenge. Standard hidden state residuals often fail to adequately preserve initial token-level information in deeper layers. This paper introduces ResFormer, a novel architecture that enhances information flow by incorporating value residual connections in addition to hidden state residuals. And a variant is SVFormer, where all layers share the first layer's value embedding. Comprehensive empirical evidence demonstrates ResFormer "},"verification_status":{"content_addressed":true,"pith_receipt":true,"author_attested":false,"weak_author_claims":0,"strong_author_claims":0,"externally_anchored":false,"storage_verified":false,"citation_signatures":0,"replication_records":0,"graph_snapshot":true,"references_resolved":false,"formal_links_present":false},"canonical_record":{"source":{"id":"2410.17897","kind":"arxiv","version":5},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2024-10-23T14:15:07Z","cross_cats_sorted":[],"title_canon_sha256":"6211ab6a7bb6c66468d678b4d39a5406a70f9a39ad141f9e0fa057fcc99aeeb4","abstract_canon_sha256":"b9c31e1c182967ba9abed8bb820cbc750658ed4ab9dd0292124069d49048239a"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:17:36.662610Z","signature_b64":"L9E9lKdC0VBFmD7yTYfUBWMJUGklsQPaVcFzaAZsDKfrOEICae5ol4dl8qPpbe1Om87XLD48KsbzwjqSj/L+DQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"f65dac5dfd19aaea738c8a5d79e4d3a3d9df55bc002a88ad7aba4b9ed41ba963","last_reissued_at":"2026-07-05T11:17:36.662097Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:17:36.662097Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Value Residual Learning","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Fares Obeid, Tianyi Wu, Zhanchao Zhou, Zhenzhong Lan, Zhiyun Jiang","submitted_at":"2024-10-23T14:15:07Z","abstract_excerpt":"While Transformer models have achieved remarkable success in various domains, the effectiveness of information propagation through deep networks remains a critical challenge. Standard hidden state residuals often fail to adequately preserve initial token-level information in deeper layers. This paper introduces ResFormer, a novel architecture that enhances information flow by incorporating value residual connections in addition to hidden state residuals. And a variant is SVFormer, where all layers share the first layer's value embedding. Comprehensive empirical evidence demonstrates ResFormer "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2410.17897","kind":"arxiv","version":5},"verdict":{"id":null,"model_set":{},"created_at":null,"strongest_claim":"","one_line_summary":"","pipeline_version":null,"weakest_assumption":"","pith_extraction_headline":""},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2410.17897/integrity.json","findings":[],"available":true,"detectors_run":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938"},"references":{"count":0,"sample":[],"resolved_work":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","internal_anchors":0},"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"},"aliases":[{"alias_kind":"arxiv","alias_value":"2410.17897","created_at":"2026-07-05T11:17:36.662166+00:00"},{"alias_kind":"arxiv_version","alias_value":"2410.17897v5","created_at":"2026-07-05T11:17:36.662166+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2410.17897","created_at":"2026-07-05T11:17:36.662166+00:00"},{"alias_kind":"pith_short_12","alias_value":"6ZO2YXP5DGVO","created_at":"2026-07-05T11:17:36.662166+00:00"},{"alias_kind":"pith_short_16","alias_value":"6ZO2YXP5DGVOU44M","created_at":"2026-07-05T11:17:36.662166+00:00"},{"alias_kind":"pith_short_8","alias_value":"6ZO2YXP5","created_at":"2026-07-05T11:17:36.662166+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":7,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.25198","citing_title":"Heuresis: Search Strategies for Autonomous AI Research Agents Across Quality, Diversity and Novelty","ref_index":73,"is_internal_anchor":false},{"citing_arxiv_id":"2606.25198","citing_title":"Heuresis: Search Strategies for Autonomous AI Research Agents Across Quality, Diversity and Novelty","ref_index":74,"is_internal_anchor":false},{"citing_arxiv_id":"2606.02780","citing_title":"Do Value Vectors in Deep Layers Need Context from the Residual Stream?","ref_index":111,"is_internal_anchor":false},{"citing_arxiv_id":"2606.07604","citing_title":"Contribution Weights: A Geometrical Analysis of Self-Attention Transformers","ref_index":46,"is_internal_anchor":false},{"citing_arxiv_id":"2605.20798","citing_title":"Most Transformer Modifications Still Do Not Transfer at 1-3B: A 2020-2026 Update to Narang et al. (2021) with Downstream Evaluation and a Noise Floor","ref_index":37,"is_internal_anchor":false},{"citing_arxiv_id":"2509.18993","citing_title":"CR-Net: Scaling Parameter-Efficient Training with Cross-Layer Low-Rank Structure","ref_index":65,"is_internal_anchor":false},{"citing_arxiv_id":"2604.21724","citing_title":"Beyond N-gram: Data-Aware X-GRAM Extraction for Efficient Embedding Parameter Scaling","ref_index":17,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/6ZO2YXP5DGVOU44MRJOXTZGTUP","json":"https://pith.science/pith/6ZO2YXP5DGVOU44MRJOXTZGTUP.json","graph_json":"https://pith.science/api/pith-number/6ZO2YXP5DGVOU44MRJOXTZGTUP/graph.json","events_json":"https://pith.science/api/pith-number/6ZO2YXP5DGVOU44MRJOXTZGTUP/events.json","paper":"https://pith.science/paper/6ZO2YXP5"},"agent_actions":{"view_html":"https://pith.science/pith/6ZO2YXP5DGVOU44MRJOXTZGTUP","download_json":"https://pith.science/pith/6ZO2YXP5DGVOU44MRJOXTZGTUP.json","view_paper":"https://pith.science/paper/6ZO2YXP5","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2410.17897&json=true","fetch_graph":"https://pith.science/api/pith-number/6ZO2YXP5DGVOU44MRJOXTZGTUP/graph.json","fetch_events":"https://pith.science/api/pith-number/6ZO2YXP5DGVOU44MRJOXTZGTUP/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/6ZO2YXP5DGVOU44MRJOXTZGTUP/action/timestamp_anchor","attest_storage":"https://pith.science/pith/6ZO2YXP5DGVOU44MRJOXTZGTUP/action/storage_attestation","attest_author":"https://pith.science/pith/6ZO2YXP5DGVOU44MRJOXTZGTUP/action/author_attestation","sign_citation":"https://pith.science/pith/6ZO2YXP5DGVOU44MRJOXTZGTUP/action/citation_signature","submit_replication":"https://pith.science/pith/6ZO2YXP5DGVOU44MRJOXTZGTUP/action/replication_record"}},"created_at":"2026-07-05T11:17:36.662166+00:00","updated_at":"2026-07-05T11:17:36.662166+00:00"}