{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2023:KQA7DYW6Q33DSO7XNLAI3WK5WW","short_pith_number":"pith:KQA7DYW6","schema_version":"1.0","canonical_sha256":"5401f1e2de86f6393bf76ac08dd95db5951b4e73f2e5971006e2967837513244","source":{"kind":"arxiv","id":"2312.10665","version":1},"attestation_state":"computed","paper":{"title":"Silkie: Preference Distillation for Large Visual Language Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.CV","authors_text":"Benyou Wang, Lei Li, Liang Chen, Lingpeng Kong, Mukai Li, Peiyi Wang, Shunian Chen, Yazheng Yang, Zhihui Xie","submitted_at":"2023-12-17T09:44:27Z","abstract_excerpt":"This paper explores preference distillation for large vision language models (LVLMs), improving their ability to generate helpful and faithful responses anchoring the visual context. We first build a vision-language feedback (VLFeedback) dataset utilizing AI annotation. Specifically, responses are generated by models sampled from 12 LVLMs, conditioned on multi-modal instructions sourced from various datasets. We adopt GPT-4V to assess the generated outputs regarding helpfulness, visual faithfulness, and ethical considerations. Furthermore, the preference supervision is distilled into Qwen-VL-C"},"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":"2312.10665","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CV","submitted_at":"2023-12-17T09:44:27Z","cross_cats_sorted":["cs.CL"],"title_canon_sha256":"12534fd8a702fd445eb56a183266aa9a2c828ea84200314b9564df4e3f059c09","abstract_canon_sha256":"114ec248dddbf3686820f868b3fa8dd86ac396b4be9bdda6bfe523bdbae3397c"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T07:25:10.406966Z","signature_b64":"WrfkF97xA6EAl/9GKhXTbSpVWXddcvKJ0zGGJ+RdNpDljTYSr8ycr8sUliQ2D0qY2ObKrGvD8V0mhisuBJL9BA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"5401f1e2de86f6393bf76ac08dd95db5951b4e73f2e5971006e2967837513244","last_reissued_at":"2026-07-05T07:25:10.406496Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T07:25:10.406496Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Silkie: Preference Distillation for Large Visual Language Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.CV","authors_text":"Benyou Wang, Lei Li, Liang Chen, Lingpeng Kong, Mukai Li, Peiyi Wang, Shunian Chen, Yazheng Yang, Zhihui Xie","submitted_at":"2023-12-17T09:44:27Z","abstract_excerpt":"This paper explores preference distillation for large vision language models (LVLMs), improving their ability to generate helpful and faithful responses anchoring the visual context. We first build a vision-language feedback (VLFeedback) dataset utilizing AI annotation. Specifically, responses are generated by models sampled from 12 LVLMs, conditioned on multi-modal instructions sourced from various datasets. We adopt GPT-4V to assess the generated outputs regarding helpfulness, visual faithfulness, and ethical considerations. Furthermore, the preference supervision is distilled into Qwen-VL-C"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2312.10665","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/2312.10665/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":"2312.10665","created_at":"2026-07-05T07:25:10.406552+00:00"},{"alias_kind":"arxiv_version","alias_value":"2312.10665v1","created_at":"2026-07-05T07:25:10.406552+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2312.10665","created_at":"2026-07-05T07:25:10.406552+00:00"},{"alias_kind":"pith_short_12","alias_value":"KQA7DYW6Q33D","created_at":"2026-07-05T07:25:10.406552+00:00"},{"alias_kind":"pith_short_16","alias_value":"KQA7DYW6Q33DSO7X","created_at":"2026-07-05T07:25:10.406552+00:00"},{"alias_kind":"pith_short_8","alias_value":"KQA7DYW6","created_at":"2026-07-05T07:25:10.406552+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":18,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.25398","citing_title":"MAPL: Multi-Objective Preference Learning for Robot Locomotion","ref_index":20,"is_internal_anchor":false},{"citing_arxiv_id":"2605.25343","citing_title":"Toward Native Multimodal Modeling: A Roadmap","ref_index":171,"is_internal_anchor":false},{"citing_arxiv_id":"2507.12455","citing_title":"Mitigating Object Hallucinations via Sentence-Level Early Intervention","ref_index":28,"is_internal_anchor":false},{"citing_arxiv_id":"2402.11684","citing_title":"ALLaVA: Harnessing GPT4V-Synthesized Data for Lite Vision-Language Models","ref_index":106,"is_internal_anchor":false},{"citing_arxiv_id":"2605.15300","citing_title":"Deep Pre-Alignment for VLMs","ref_index":74,"is_internal_anchor":false},{"citing_arxiv_id":"2402.13116","citing_title":"A Survey on Knowledge Distillation of Large Language Models","ref_index":34,"is_internal_anchor":false},{"citing_arxiv_id":"2402.11411","citing_title":"Aligning Modalities in Vision Large Language Models via Preference Fine-tuning","ref_index":161,"is_internal_anchor":false},{"citing_arxiv_id":"2407.03320","citing_title":"InternLM-XComposer-2.5: A Versatile Large Vision Language Model Supporting Long-Contextual Input and Output","ref_index":73,"is_internal_anchor":false},{"citing_arxiv_id":"2411.10442","citing_title":"Enhancing the Reasoning Ability of Multimodal Large Language Models via Mixed Preference Optimization","ref_index":48,"is_internal_anchor":false},{"citing_arxiv_id":"2306.13549","citing_title":"A Survey on Multimodal Large Language Models","ref_index":117,"is_internal_anchor":false},{"citing_arxiv_id":"2305.10415","citing_title":"PMC-VQA: Visual Instruction Tuning for Medical Visual Question Answering","ref_index":31,"is_internal_anchor":false},{"citing_arxiv_id":"2603.13054","citing_title":"Topo-R1: Detecting Topological Anomalies via Vision-Language Models","ref_index":41,"is_internal_anchor":false},{"citing_arxiv_id":"2404.07972","citing_title":"OSWorld: Benchmarking Multimodal Agents for Open-Ended Tasks in Real Computer Environments","ref_index":26,"is_internal_anchor":false},{"citing_arxiv_id":"2605.00323","citing_title":"Online Self-Calibration Against Hallucination in Vision-Language Models","ref_index":15,"is_internal_anchor":false},{"citing_arxiv_id":"2404.18930","citing_title":"Hallucination of Multimodal Large Language Models: A Survey","ref_index":105,"is_internal_anchor":false},{"citing_arxiv_id":"2604.18034","citing_title":"SignDPO: Multi-level Direct Preference Optimisation for Skeleton-based Gloss-free Sign Language Translation","ref_index":17,"is_internal_anchor":false},{"citing_arxiv_id":"2604.13029","citing_title":"Visual Preference Optimization with Rubric Rewards","ref_index":26,"is_internal_anchor":false},{"citing_arxiv_id":"2604.10966","citing_title":"You Only Judge Once: Multi-response Reward Modeling in a Single Forward Pass","ref_index":3,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/KQA7DYW6Q33DSO7XNLAI3WK5WW","json":"https://pith.science/pith/KQA7DYW6Q33DSO7XNLAI3WK5WW.json","graph_json":"https://pith.science/api/pith-number/KQA7DYW6Q33DSO7XNLAI3WK5WW/graph.json","events_json":"https://pith.science/api/pith-number/KQA7DYW6Q33DSO7XNLAI3WK5WW/events.json","paper":"https://pith.science/paper/KQA7DYW6"},"agent_actions":{"view_html":"https://pith.science/pith/KQA7DYW6Q33DSO7XNLAI3WK5WW","download_json":"https://pith.science/pith/KQA7DYW6Q33DSO7XNLAI3WK5WW.json","view_paper":"https://pith.science/paper/KQA7DYW6","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2312.10665&json=true","fetch_graph":"https://pith.science/api/pith-number/KQA7DYW6Q33DSO7XNLAI3WK5WW/graph.json","fetch_events":"https://pith.science/api/pith-number/KQA7DYW6Q33DSO7XNLAI3WK5WW/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/KQA7DYW6Q33DSO7XNLAI3WK5WW/action/timestamp_anchor","attest_storage":"https://pith.science/pith/KQA7DYW6Q33DSO7XNLAI3WK5WW/action/storage_attestation","attest_author":"https://pith.science/pith/KQA7DYW6Q33DSO7XNLAI3WK5WW/action/author_attestation","sign_citation":"https://pith.science/pith/KQA7DYW6Q33DSO7XNLAI3WK5WW/action/citation_signature","submit_replication":"https://pith.science/pith/KQA7DYW6Q33DSO7XNLAI3WK5WW/action/replication_record"}},"created_at":"2026-07-05T07:25:10.406552+00:00","updated_at":"2026-07-05T07:25:10.406552+00:00"}