{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:AAJENRCJURKDRQA4327TSY3R4S","short_pith_number":"pith:AAJENRCJ","schema_version":"1.0","canonical_sha256":"001246c449a45438c01cdebf396371e497cf6b7e867d19415225e6df41060353","source":{"kind":"arxiv","id":"2405.20985","version":1},"attestation_state":"computed","paper":{"title":"DeCo: Decoupling Token Compression from Semantic Abstraction in Multimodal Large Language Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Lean Wang, Lei Li, Linli Yao, Lu Hou, Shuhuai Ren, Xu Sun, Yuanxin Liu","submitted_at":"2024-05-31T16:31:38Z","abstract_excerpt":"The visual projector, which bridges the vision and language modalities and facilitates cross-modal alignment, serves as a crucial component in MLLMs. However, measuring the effectiveness of projectors in vision-language alignment remains under-explored, which currently can only be inferred from the performance of MLLMs on downstream tasks. Motivated by the problem, this study examines the projector module by interpreting the vision-language semantic flow within MLLMs. Specifically, we trace back the semantic relevance flow from generated language tokens to raw visual encoder patches and the in"},"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":"2405.20985","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CV","submitted_at":"2024-05-31T16:31:38Z","cross_cats_sorted":[],"title_canon_sha256":"ed8951b2078458a1437c9da9dad6916dda54cdbff8f30c19ca651add61630e7e","abstract_canon_sha256":"e94e2b543bd76d4f79a25154d25cbce411cf0f65e083605e2202e9bd85c139f4"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T08:25:44.400742Z","signature_b64":"BXT9WCQ1K9SFHNbMtaC6fixVDyke0EHEkqbeZVnecqTghz3+4XQ1LxE5TvpTpIWHjssTIGEPSFhHK1Rz8IBKAA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"001246c449a45438c01cdebf396371e497cf6b7e867d19415225e6df41060353","last_reissued_at":"2026-07-05T08:25:44.400397Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T08:25:44.400397Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"DeCo: Decoupling Token Compression from Semantic Abstraction in Multimodal Large Language Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Lean Wang, Lei Li, Linli Yao, Lu Hou, Shuhuai Ren, Xu Sun, Yuanxin Liu","submitted_at":"2024-05-31T16:31:38Z","abstract_excerpt":"The visual projector, which bridges the vision and language modalities and facilitates cross-modal alignment, serves as a crucial component in MLLMs. However, measuring the effectiveness of projectors in vision-language alignment remains under-explored, which currently can only be inferred from the performance of MLLMs on downstream tasks. Motivated by the problem, this study examines the projector module by interpreting the vision-language semantic flow within MLLMs. Specifically, we trace back the semantic relevance flow from generated language tokens to raw visual encoder patches and the in"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2405.20985","kind":"arxiv","version":1},"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/2405.20985/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":"2405.20985","created_at":"2026-07-05T08:25:44.400455+00:00"},{"alias_kind":"arxiv_version","alias_value":"2405.20985v1","created_at":"2026-07-05T08:25:44.400455+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2405.20985","created_at":"2026-07-05T08:25:44.400455+00:00"},{"alias_kind":"pith_short_12","alias_value":"AAJENRCJURKD","created_at":"2026-07-05T08:25:44.400455+00:00"},{"alias_kind":"pith_short_16","alias_value":"AAJENRCJURKDRQA4","created_at":"2026-07-05T08:25:44.400455+00:00"},{"alias_kind":"pith_short_8","alias_value":"AAJENRCJ","created_at":"2026-07-05T08:25:44.400455+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":10,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.24422","citing_title":"EgoSAT: A Comprehensive Benchmark of Egocentric Streaming Interaction Understanding","ref_index":46,"is_internal_anchor":false},{"citing_arxiv_id":"2606.02300","citing_title":"Beyond Isolated Behaviors: Hierarchical User Modeling for LLM Personalization","ref_index":75,"is_internal_anchor":false},{"citing_arxiv_id":"2412.18158","citing_title":"Semantics Disentanglement and Composition for Universal Image Coding with Efficiently LLM Reasoning and Generative Diffusion","ref_index":50,"is_internal_anchor":false},{"citing_arxiv_id":"2410.04417","citing_title":"SparseVLM: Visual Token Sparsification for Efficient Vision-Language Model Inference","ref_index":42,"is_internal_anchor":false},{"citing_arxiv_id":"2410.17247","citing_title":"PyramidDrop: Accelerating Your Large Vision-Language Models via Pyramid Visual Redundancy Reduction","ref_index":50,"is_internal_anchor":false},{"citing_arxiv_id":"2603.27383","citing_title":"Decompose, Mix, Adapt: A Unified Framework for Parameter-Efficient Neural Network Recombination and Compression","ref_index":87,"is_internal_anchor":false},{"citing_arxiv_id":"2604.02689","citing_title":"Efficient3D: A Unified Framework for Adaptive and Debiased Token Reduction in 3D MLLMs","ref_index":71,"is_internal_anchor":false},{"citing_arxiv_id":"2604.22281","citing_title":"DocPrune:Efficient Document Question Answering via Background, Question, and Comprehension-aware Token Pruning","ref_index":35,"is_internal_anchor":false},{"citing_arxiv_id":"2604.11240","citing_title":"Decoupled Similarity for Task-Aware Token Pruning in Large Vision-Language Models","ref_index":46,"is_internal_anchor":false},{"citing_arxiv_id":"2605.07568","citing_title":"Tracing the Arrow of Time: Diagnosing Temporal Information Flow in Video-LLMs","ref_index":49,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/AAJENRCJURKDRQA4327TSY3R4S","json":"https://pith.science/pith/AAJENRCJURKDRQA4327TSY3R4S.json","graph_json":"https://pith.science/api/pith-number/AAJENRCJURKDRQA4327TSY3R4S/graph.json","events_json":"https://pith.science/api/pith-number/AAJENRCJURKDRQA4327TSY3R4S/events.json","paper":"https://pith.science/paper/AAJENRCJ"},"agent_actions":{"view_html":"https://pith.science/pith/AAJENRCJURKDRQA4327TSY3R4S","download_json":"https://pith.science/pith/AAJENRCJURKDRQA4327TSY3R4S.json","view_paper":"https://pith.science/paper/AAJENRCJ","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2405.20985&json=true","fetch_graph":"https://pith.science/api/pith-number/AAJENRCJURKDRQA4327TSY3R4S/graph.json","fetch_events":"https://pith.science/api/pith-number/AAJENRCJURKDRQA4327TSY3R4S/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/AAJENRCJURKDRQA4327TSY3R4S/action/timestamp_anchor","attest_storage":"https://pith.science/pith/AAJENRCJURKDRQA4327TSY3R4S/action/storage_attestation","attest_author":"https://pith.science/pith/AAJENRCJURKDRQA4327TSY3R4S/action/author_attestation","sign_citation":"https://pith.science/pith/AAJENRCJURKDRQA4327TSY3R4S/action/citation_signature","submit_replication":"https://pith.science/pith/AAJENRCJURKDRQA4327TSY3R4S/action/replication_record"}},"created_at":"2026-07-05T08:25:44.400455+00:00","updated_at":"2026-07-05T08:25:44.400455+00:00"}