{"paper":{"title":"SynCABEL: Synthetic Contextualized Augmentation for Biomedical Entity Linking","license":"http://creativecommons.org/licenses/by-sa/4.0/","headline":"SynCABEL generates synthetic training examples with large language models to overcome data scarcity in biomedical entity linking and reaches new state-of-the-art results on three multilingual benchmarks with up to 60 percent less human-anno","cross_cats":["cs.AI","cs.IR","cs.LG"],"primary_cat":"cs.CL","authors_text":"Adam Remaki, Christel G\\'erardin, Eul\\`alia Farr\\'e-Maduell, Martin Krallinger, Xavier Tannier","submitted_at":"2026-01-27T14:47:17Z","abstract_excerpt":"We present SynCABEL (Synthetic Contextualized Augmentation for Biomedical Entity Linking), a framework that addresses a central bottleneck in supervised biomedical entity linking (BEL): the scarcity of expert-annotated training data. SynCABEL leverages large language models to generate context-rich synthetic training examples for all candidate concepts in a target knowledge base, providing broad supervision without manual annotation. We demonstrate that SynCABEL, when combined with decoder-only models and guided inference, establishes new state-of-the-art results across three widely used multi"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"SynCABEL, when combined with decoder-only models and guided inference, establishes new state-of-the-art results across three widely used multilingual benchmarks: MedMentions for English, QUAERO for French, and SPACCC for Spanish.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"The assumption that LLM-generated synthetic examples are sufficiently representative of real biomedical text distributions and do not introduce systematic biases or hallucinations that would degrade downstream linking performance.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"SynCABEL generates LLM-based synthetic data for all candidate concepts in biomedical entity linking, reaching new SOTA results on MedMentions, QUAERO, and SPACCC with up to 60% less human-annotated data.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"SynCABEL generates synthetic training examples with large language models to overcome data scarcity in biomedical entity linking and reaches new state-of-the-art results on three multilingual benchmarks with up to 60 percent less human-anno","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"9bf3a9d6244abd4915dcd9ed74c436faae953d4f4e26ba3c7f58cbd88a2571ae"},"source":{"id":"2601.19667","kind":"arxiv","version":2},"verdict":{"id":"76bc4c1f-786b-4a46-86e5-a03019a397a1","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-16T10:56:45.841996Z","strongest_claim":"SynCABEL, when combined with decoder-only models and guided inference, establishes new state-of-the-art results across three widely used multilingual benchmarks: MedMentions for English, QUAERO for French, and SPACCC for Spanish.","one_line_summary":"SynCABEL generates LLM-based synthetic data for all candidate concepts in biomedical entity linking, reaching new SOTA results on MedMentions, QUAERO, and SPACCC with up to 60% less human-annotated data.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"The assumption that LLM-generated synthetic examples are sufficiently representative of real biomedical text distributions and do not introduce systematic biases or hallucinations that would degrade downstream linking performance.","pith_extraction_headline":"SynCABEL generates synthetic training examples with large language models to overcome data scarcity in biomedical entity linking and reaches new state-of-the-art results on three multilingual benchmarks with up to 60 percent less human-anno"},"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"}