{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2022:5ZWFXW5RGGA7WHGTPPC5KGJMIM","short_pith_number":"pith:5ZWFXW5R","schema_version":"1.0","canonical_sha256":"ee6c5bdbb13181fb1cd37bc5d5192c433624cc3987703324a096607fcd187bef","source":{"kind":"arxiv","id":"2212.09849","version":6},"attestation_state":"computed","paper":{"title":"Dataless Knowledge Fusion by Merging Weights of Language Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CL","authors_text":"Daniel Preotiuc-Pietro, Pengxiang Cheng, Xiang Ren, Xisen Jin","submitted_at":"2022-12-19T20:46:43Z","abstract_excerpt":"Fine-tuning pre-trained language models has become the prevalent paradigm for building downstream NLP models. Oftentimes fine-tuned models are readily available but their training data is not, due to data privacy or intellectual property concerns. This creates a barrier to fusing knowledge across individual models to yield a better single model. In this paper, we study the problem of merging individual models built on different training data sets to obtain a single model that performs well both across all data set domains and can generalize on out-of-domain data. We propose a dataless knowledg"},"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":"2212.09849","kind":"arxiv","version":6},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2022-12-19T20:46:43Z","cross_cats_sorted":["cs.LG"],"title_canon_sha256":"724852c2ab7b1e64ceddd8aa840fd497ca46c29d1795b0e839cae7d739c1f8cc","abstract_canon_sha256":"83f9f90b031f259d0e84c54dab0f6240544dec47aabf938d8d582e7352218445"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:06:59.465184Z","signature_b64":"HnShSx+7NDH5r7gCrGmMkHXKjb9NBHpG81nySXHBjuG30FMIgWAQf2U0fTu0qQFDWztyV2QcN3juIW5QXfx6Dw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"ee6c5bdbb13181fb1cd37bc5d5192c433624cc3987703324a096607fcd187bef","last_reissued_at":"2026-07-05T11:06:59.464729Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:06:59.464729Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Dataless Knowledge Fusion by Merging Weights of Language Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CL","authors_text":"Daniel Preotiuc-Pietro, Pengxiang Cheng, Xiang Ren, Xisen Jin","submitted_at":"2022-12-19T20:46:43Z","abstract_excerpt":"Fine-tuning pre-trained language models has become the prevalent paradigm for building downstream NLP models. Oftentimes fine-tuned models are readily available but their training data is not, due to data privacy or intellectual property concerns. This creates a barrier to fusing knowledge across individual models to yield a better single model. In this paper, we study the problem of merging individual models built on different training data sets to obtain a single model that performs well both across all data set domains and can generalize on out-of-domain data. We propose a dataless knowledg"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2212.09849","kind":"arxiv","version":6},"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/2212.09849/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":"2212.09849","created_at":"2026-07-05T11:06:59.464793+00:00"},{"alias_kind":"arxiv_version","alias_value":"2212.09849v6","created_at":"2026-07-05T11:06:59.464793+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2212.09849","created_at":"2026-07-05T11:06:59.464793+00:00"},{"alias_kind":"pith_short_12","alias_value":"5ZWFXW5RGGA7","created_at":"2026-07-05T11:06:59.464793+00:00"},{"alias_kind":"pith_short_16","alias_value":"5ZWFXW5RGGA7WHGT","created_at":"2026-07-05T11:06:59.464793+00:00"},{"alias_kind":"pith_short_8","alias_value":"5ZWFXW5R","created_at":"2026-07-05T11:06:59.464793+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":14,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.19549","citing_title":"Predicting Mergeability of Parameter-Efficient Fine-Tuning Updates","ref_index":39,"is_internal_anchor":false},{"citing_arxiv_id":"2607.01689","citing_title":"Model Merging as Probabilistic Inference in Fine-Tuning Parameter Space","ref_index":51,"is_internal_anchor":false},{"citing_arxiv_id":"2606.03391","citing_title":"When Model Merging Breaks Routing: Training-Free Calibration for MoE","ref_index":6,"is_internal_anchor":false},{"citing_arxiv_id":"2606.02606","citing_title":"ReLoRA: Knowledge-Reusing Adaptation for Fast Rollout of Evolving LLM Services","ref_index":42,"is_internal_anchor":false},{"citing_arxiv_id":"2605.18904","citing_title":"Dynamic Model Merging Made Slim","ref_index":43,"is_internal_anchor":false},{"citing_arxiv_id":"2509.24244","citing_title":"Model Merging Scaling Laws in Large Language Models","ref_index":11,"is_internal_anchor":false},{"citing_arxiv_id":"2603.02945","citing_title":"ACE-Merging: Data-Free Model Merging with Adaptive Covariance Estimation","ref_index":7,"is_internal_anchor":false},{"citing_arxiv_id":"2605.14350","citing_title":"Distributionally Robust Multi-Task Reinforcement Learning via Adaptive Task Sampling","ref_index":236,"is_internal_anchor":false},{"citing_arxiv_id":"2604.02633","citing_title":"Analytic Drift Resister for Non-Exemplar Continual Graph Learning","ref_index":34,"is_internal_anchor":false},{"citing_arxiv_id":"2604.27155","citing_title":"Generalizing the Geometry of Model Merging Through Frechet Averages","ref_index":4,"is_internal_anchor":false},{"citing_arxiv_id":"2604.27155","citing_title":"Generalizing the Geometry of Model Merging Through Frechet Averages","ref_index":4,"is_internal_anchor":false},{"citing_arxiv_id":"2604.23121","citing_title":"Breaking Lock-In: Preserving Steerability under Low-Data VLA Post-Training","ref_index":21,"is_internal_anchor":false},{"citing_arxiv_id":"2604.20985","citing_title":"Differentially Private Model Merging","ref_index":8,"is_internal_anchor":false},{"citing_arxiv_id":"2604.14016","citing_title":"MAny: Merge Anything for Multimodal Continual Instruction Tuning","ref_index":5,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/5ZWFXW5RGGA7WHGTPPC5KGJMIM","json":"https://pith.science/pith/5ZWFXW5RGGA7WHGTPPC5KGJMIM.json","graph_json":"https://pith.science/api/pith-number/5ZWFXW5RGGA7WHGTPPC5KGJMIM/graph.json","events_json":"https://pith.science/api/pith-number/5ZWFXW5RGGA7WHGTPPC5KGJMIM/events.json","paper":"https://pith.science/paper/5ZWFXW5R"},"agent_actions":{"view_html":"https://pith.science/pith/5ZWFXW5RGGA7WHGTPPC5KGJMIM","download_json":"https://pith.science/pith/5ZWFXW5RGGA7WHGTPPC5KGJMIM.json","view_paper":"https://pith.science/paper/5ZWFXW5R","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2212.09849&json=true","fetch_graph":"https://pith.science/api/pith-number/5ZWFXW5RGGA7WHGTPPC5KGJMIM/graph.json","fetch_events":"https://pith.science/api/pith-number/5ZWFXW5RGGA7WHGTPPC5KGJMIM/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/5ZWFXW5RGGA7WHGTPPC5KGJMIM/action/timestamp_anchor","attest_storage":"https://pith.science/pith/5ZWFXW5RGGA7WHGTPPC5KGJMIM/action/storage_attestation","attest_author":"https://pith.science/pith/5ZWFXW5RGGA7WHGTPPC5KGJMIM/action/author_attestation","sign_citation":"https://pith.science/pith/5ZWFXW5RGGA7WHGTPPC5KGJMIM/action/citation_signature","submit_replication":"https://pith.science/pith/5ZWFXW5RGGA7WHGTPPC5KGJMIM/action/replication_record"}},"created_at":"2026-07-05T11:06:59.464793+00:00","updated_at":"2026-07-05T11:06:59.464793+00:00"}