{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:VHKEUZ3IXPWC47FBKG3Q5PAIQV","short_pith_number":"pith:VHKEUZ3I","schema_version":"1.0","canonical_sha256":"a9d44a6768bbec2e7ca151b70ebc08857a337829f40148218e7a0a63763d2e35","source":{"kind":"arxiv","id":"2604.11225","version":1},"attestation_state":"computed","paper":{"title":"Sign Language Recognition in the Age of LLMs","license":"http://creativecommons.org/licenses/by/4.0/","headline":"Open-source VLMs lag far behind supervised classifiers in zero-shot sign language recognition but capture partial visual-semantic alignment.","cross_cats":["cs.CL"],"primary_cat":"cs.CV","authors_text":"Ivan Gruber, Jakub Honzik, Marek Hruz, Tomas Zelezny, Vaclav Javorek","submitted_at":"2026-04-13T09:26:16Z","abstract_excerpt":"Recent Vision Language Models (VLMs) have demonstrated strong performance across a wide range of multimodal reasoning tasks. This raises the question of whether such general-purpose models can also address specialized visual recognition problems such as isolated sign language recognition (ISLR) without task-specific training. In this work, we investigate the capability of modern VLMs to perform ISLR in a zero-shot setting. We evaluate several open-source and proprietary VLMs on the WLASL300 benchmark. Our experiments show that, under prompt-only zero-shot inference, current open-source VLMs re"},"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":true,"formal_links_present":false},"canonical_record":{"source":{"id":"2604.11225","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CV","submitted_at":"2026-04-13T09:26:16Z","cross_cats_sorted":["cs.CL"],"title_canon_sha256":"5e0b1669158c91ef7efc2a3a42ac10d4b48af8a06f3d4bfab51d06a35b80f7ee","abstract_canon_sha256":"44ced6aa60d597cb50792d3a19e7842afcfdf86471a96419bf6386deb79fbfd0"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-26T01:15:51.644096Z","signature_b64":"OYp8mqCZYG8hwhqDzVwYQRIWtdMMl3HQq0cs2yJAKe64Y/CpbQDLIaFgq3j1Vnks4F5aAK1vnCkfwXqWaF53Cg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"a9d44a6768bbec2e7ca151b70ebc08857a337829f40148218e7a0a63763d2e35","last_reissued_at":"2026-06-26T01:15:51.643721Z","signature_status":"signed_v1","first_computed_at":"2026-06-26T01:15:51.643721Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Sign Language Recognition in the Age of LLMs","license":"http://creativecommons.org/licenses/by/4.0/","headline":"Open-source VLMs lag far behind supervised classifiers in zero-shot sign language recognition but capture partial visual-semantic alignment.","cross_cats":["cs.CL"],"primary_cat":"cs.CV","authors_text":"Ivan Gruber, Jakub Honzik, Marek Hruz, Tomas Zelezny, Vaclav Javorek","submitted_at":"2026-04-13T09:26:16Z","abstract_excerpt":"Recent Vision Language Models (VLMs) have demonstrated strong performance across a wide range of multimodal reasoning tasks. This raises the question of whether such general-purpose models can also address specialized visual recognition problems such as isolated sign language recognition (ISLR) without task-specific training. In this work, we investigate the capability of modern VLMs to perform ISLR in a zero-shot setting. We evaluate several open-source and proprietary VLMs on the WLASL300 benchmark. Our experiments show that, under prompt-only zero-shot inference, current open-source VLMs re"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"under prompt-only zero-shot inference, current open-source VLMs remain far behind classic supervised ISLR classifiers by a wide margin. However, follow-up experiments reveal that these models capture partial visual-semantic alignment between signs and text descriptions. Larger proprietary models achieve substantially higher accuracy.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That the chosen prompts, WLASL300 benchmark splits, and evaluation protocol provide an unbiased test of zero-shot capability without hidden advantages from prompt engineering or dataset characteristics.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"Zero-shot VLM evaluation on WLASL300 reveals open-source models lag far behind supervised ISLR baselines, but proprietary models improve with scale and exhibit some visual-semantic alignment.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"Open-source VLMs lag far behind supervised classifiers in zero-shot sign language recognition but capture partial visual-semantic alignment.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"47c486ded64478a4b522fc93a6e852140e5bba4ec15f99554bca667f72c0cfc1"},"source":{"id":"2604.11225","kind":"arxiv","version":1},"verdict":{"id":"46a620be-b5fe-4623-91aa-1020545e9560","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-10T15:31:21.413668Z","strongest_claim":"under prompt-only zero-shot inference, current open-source VLMs remain far behind classic supervised ISLR classifiers by a wide margin. However, follow-up experiments reveal that these models capture partial visual-semantic alignment between signs and text descriptions. Larger proprietary models achieve substantially higher accuracy.","one_line_summary":"Zero-shot VLM evaluation on WLASL300 reveals open-source models lag far behind supervised ISLR baselines, but proprietary models improve with scale and exhibit some visual-semantic alignment.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That the chosen prompts, WLASL300 benchmark splits, and evaluation protocol provide an unbiased test of zero-shot capability without hidden advantages from prompt engineering or dataset characteristics.","pith_extraction_headline":"Open-source VLMs lag far behind supervised classifiers in zero-shot sign language recognition but capture partial visual-semantic alignment."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2604.11225/integrity.json","findings":[],"available":true,"detectors_run":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938"},"references":{"count":44,"sample":[{"doi":"","year":2022,"title":"Flamingo: a visual language model for few- shot learning","work_id":"8077b03e-3b58-4984-9d17-7025a3593594","ref_index":1,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":null,"title":"Pyav: Pythonic bindings for ffm- peg.https://github.com/PyAV-Org/PyAV","work_id":"88b9e280-46e2-47f8-a71f-ffe11a16c2fe","ref_index":2,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":1901,"title":"Language models are few-shot learn- ers.Advances in neural information processing sys- tems, 33:1877–1901","work_id":"c4b7fbe8-38ff-4f9c-921c-223ff3a20fda","ref_index":3,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2015,"title":"Pillow (pil fork) documentation","work_id":"9253d432-575e-4e41-9d6f-fb7b47e99862","ref_index":4,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2025,"title":"Emerging Properties in Unified Multimodal Pretraining","work_id":"e0cfd82c-f5d4-44fd-b531-ec73ab0a805b","ref_index":5,"cited_arxiv_id":"2505.14683","is_internal_anchor":true}],"resolved_work":44,"snapshot_sha256":"ed46b9d1b80be4198b042f54434acf306f0e168166e1dffe575823d5a284f375","internal_anchors":2},"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":"2604.11225","created_at":"2026-06-26T01:15:51.643776+00:00"},{"alias_kind":"arxiv_version","alias_value":"2604.11225v1","created_at":"2026-06-26T01:15:51.643776+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2604.11225","created_at":"2026-06-26T01:15:51.643776+00:00"},{"alias_kind":"pith_short_12","alias_value":"VHKEUZ3IXPWC","created_at":"2026-06-26T01:15:51.643776+00:00"},{"alias_kind":"pith_short_16","alias_value":"VHKEUZ3IXPWC47FB","created_at":"2026-06-26T01:15:51.643776+00:00"},{"alias_kind":"pith_short_8","alias_value":"VHKEUZ3I","created_at":"2026-06-26T01:15:51.643776+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":0,"internal_anchor_count":0,"sample":[]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/VHKEUZ3IXPWC47FBKG3Q5PAIQV","json":"https://pith.science/pith/VHKEUZ3IXPWC47FBKG3Q5PAIQV.json","graph_json":"https://pith.science/api/pith-number/VHKEUZ3IXPWC47FBKG3Q5PAIQV/graph.json","events_json":"https://pith.science/api/pith-number/VHKEUZ3IXPWC47FBKG3Q5PAIQV/events.json","paper":"https://pith.science/paper/VHKEUZ3I"},"agent_actions":{"view_html":"https://pith.science/pith/VHKEUZ3IXPWC47FBKG3Q5PAIQV","download_json":"https://pith.science/pith/VHKEUZ3IXPWC47FBKG3Q5PAIQV.json","view_paper":"https://pith.science/paper/VHKEUZ3I","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2604.11225&json=true","fetch_graph":"https://pith.science/api/pith-number/VHKEUZ3IXPWC47FBKG3Q5PAIQV/graph.json","fetch_events":"https://pith.science/api/pith-number/VHKEUZ3IXPWC47FBKG3Q5PAIQV/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/VHKEUZ3IXPWC47FBKG3Q5PAIQV/action/timestamp_anchor","attest_storage":"https://pith.science/pith/VHKEUZ3IXPWC47FBKG3Q5PAIQV/action/storage_attestation","attest_author":"https://pith.science/pith/VHKEUZ3IXPWC47FBKG3Q5PAIQV/action/author_attestation","sign_citation":"https://pith.science/pith/VHKEUZ3IXPWC47FBKG3Q5PAIQV/action/citation_signature","submit_replication":"https://pith.science/pith/VHKEUZ3IXPWC47FBKG3Q5PAIQV/action/replication_record"}},"created_at":"2026-06-26T01:15:51.643776+00:00","updated_at":"2026-06-26T01:15:51.643776+00:00"}