{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:BLXO7MROVZTVC4BCB7RFC2MKEB","short_pith_number":"pith:BLXO7MRO","schema_version":"1.0","canonical_sha256":"0aeeefb22eae675170220fe251698a205e07df74f6a1ec6d0a7cc079a140ac5b","source":{"kind":"arxiv","id":"2407.02477","version":1},"attestation_state":"computed","paper":{"title":"Understanding Alignment in Multimodal LLMs: A Comprehensive Study","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.CV","authors_text":"Afshin Dehghan, Christian Kerl, Christoph Roesmann, Elmira Amirloo, Jean-Philippe Fauconnier, Peter Grasch, Rinu Boney, Yinfei Yang, Yusu Qian, Zhe Gan, Zirui Wang","submitted_at":"2024-07-02T17:55:03Z","abstract_excerpt":"Preference alignment has become a crucial component in enhancing the performance of Large Language Models (LLMs), yet its impact in Multimodal Large Language Models (MLLMs) remains comparatively underexplored. Similar to language models, MLLMs for image understanding tasks encounter challenges like hallucination. In MLLMs, hallucination can occur not only by stating incorrect facts but also by producing responses that are inconsistent with the image content. A primary objective of alignment for MLLMs is to encourage these models to align responses more closely with image information. Recently,"},"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":"2407.02477","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CV","submitted_at":"2024-07-02T17:55:03Z","cross_cats_sorted":["cs.CL"],"title_canon_sha256":"8357eacd0d8e1fd478cecfc420b12b55fabbe85ec781fb79b2bad5b41a42429b","abstract_canon_sha256":"17a59ebc9b4f157f9279ecc07ca07be76b6d5fdce78fee4c218bfd4c07446df5"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T08:39:24.487066Z","signature_b64":"2O427AE0ZefyU1V+vBmhIMHjdKcY6FpT/vp62eiBmZfzCqtr62wuw4H4iBq+BF7RIUekKa19t70EsC8lIt2nAQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"0aeeefb22eae675170220fe251698a205e07df74f6a1ec6d0a7cc079a140ac5b","last_reissued_at":"2026-07-05T08:39:24.486663Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T08:39:24.486663Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Understanding Alignment in Multimodal LLMs: A Comprehensive Study","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.CV","authors_text":"Afshin Dehghan, Christian Kerl, Christoph Roesmann, Elmira Amirloo, Jean-Philippe Fauconnier, Peter Grasch, Rinu Boney, Yinfei Yang, Yusu Qian, Zhe Gan, Zirui Wang","submitted_at":"2024-07-02T17:55:03Z","abstract_excerpt":"Preference alignment has become a crucial component in enhancing the performance of Large Language Models (LLMs), yet its impact in Multimodal Large Language Models (MLLMs) remains comparatively underexplored. Similar to language models, MLLMs for image understanding tasks encounter challenges like hallucination. In MLLMs, hallucination can occur not only by stating incorrect facts but also by producing responses that are inconsistent with the image content. A primary objective of alignment for MLLMs is to encourage these models to align responses more closely with image information. Recently,"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2407.02477","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/2407.02477/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":"2407.02477","created_at":"2026-07-05T08:39:24.486719+00:00"},{"alias_kind":"arxiv_version","alias_value":"2407.02477v1","created_at":"2026-07-05T08:39:24.486719+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2407.02477","created_at":"2026-07-05T08:39:24.486719+00:00"},{"alias_kind":"pith_short_12","alias_value":"BLXO7MROVZTV","created_at":"2026-07-05T08:39:24.486719+00:00"},{"alias_kind":"pith_short_16","alias_value":"BLXO7MROVZTVC4BC","created_at":"2026-07-05T08:39:24.486719+00:00"},{"alias_kind":"pith_short_8","alias_value":"BLXO7MRO","created_at":"2026-07-05T08:39:24.486719+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":4,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.09038","citing_title":"Personalization Meets Safety:Mechanisms,Risks,and Mitigations in Personalized LLMs","ref_index":259,"is_internal_anchor":false},{"citing_arxiv_id":"2605.10622","citing_title":"Vocabulary Hijacking in LVLMs: Unveiling Critical Attention Heads by Excluding Inert Tokens to Mitigate Hallucination","ref_index":10,"is_internal_anchor":false},{"citing_arxiv_id":"2604.10527","citing_title":"STORM: End-to-End Referring Multi-Object Tracking in Videos","ref_index":1,"is_internal_anchor":false},{"citing_arxiv_id":"2505.07062","citing_title":"Seed1.5-VL Technical Report","ref_index":4,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/BLXO7MROVZTVC4BCB7RFC2MKEB","json":"https://pith.science/pith/BLXO7MROVZTVC4BCB7RFC2MKEB.json","graph_json":"https://pith.science/api/pith-number/BLXO7MROVZTVC4BCB7RFC2MKEB/graph.json","events_json":"https://pith.science/api/pith-number/BLXO7MROVZTVC4BCB7RFC2MKEB/events.json","paper":"https://pith.science/paper/BLXO7MRO"},"agent_actions":{"view_html":"https://pith.science/pith/BLXO7MROVZTVC4BCB7RFC2MKEB","download_json":"https://pith.science/pith/BLXO7MROVZTVC4BCB7RFC2MKEB.json","view_paper":"https://pith.science/paper/BLXO7MRO","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2407.02477&json=true","fetch_graph":"https://pith.science/api/pith-number/BLXO7MROVZTVC4BCB7RFC2MKEB/graph.json","fetch_events":"https://pith.science/api/pith-number/BLXO7MROVZTVC4BCB7RFC2MKEB/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/BLXO7MROVZTVC4BCB7RFC2MKEB/action/timestamp_anchor","attest_storage":"https://pith.science/pith/BLXO7MROVZTVC4BCB7RFC2MKEB/action/storage_attestation","attest_author":"https://pith.science/pith/BLXO7MROVZTVC4BCB7RFC2MKEB/action/author_attestation","sign_citation":"https://pith.science/pith/BLXO7MROVZTVC4BCB7RFC2MKEB/action/citation_signature","submit_replication":"https://pith.science/pith/BLXO7MROVZTVC4BCB7RFC2MKEB/action/replication_record"}},"created_at":"2026-07-05T08:39:24.486719+00:00","updated_at":"2026-07-05T08:39:24.486719+00:00"}