{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:HIOEVHICW4D3AJO7WCLPMVFBOD","short_pith_number":"pith:HIOEVHIC","schema_version":"1.0","canonical_sha256":"3a1c4a9d02b707b025dfb096f654a170f0e78f3bcc102fa70c829e0ff46fc295","source":{"kind":"arxiv","id":"2605.16991","version":1},"attestation_state":"computed","paper":{"title":"Response-free item difficulty modelling for multiple-choice items with fine-tuned transformers: Component-wise representation and multi-task learning","license":"http://creativecommons.org/licenses/by/4.0/","headline":"Fine-tuned transformers predict multiple-choice item difficulty directly from wording, with multi-task learning aiding small-sample cases.","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Jan Net\\'ik, Patr\\'icia Martinkov\\'a","submitted_at":"2026-05-16T13:22:57Z","abstract_excerpt":"Response-free item difficulty modelling promises to reduce reliance on response-based calibration but is intrinsically difficult on reading-comprehension multiple-choice items, where difficulty depends on inferential demands across wording components. Whereas most existing approaches extract item-text features and pass them to a separate statistical or machine-learning model, we fine-tune transformer encoders end-to-end on the item wording, eliminating the manual feature engineering and preprocessing that discards information. Moreover, two extensions to this joint-encoding approach are propos"},"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":true},"canonical_record":{"source":{"id":"2605.16991","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2026-05-16T13:22:57Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"12f0ea05281bedd77611a52f9e40ab488bf728daeddb807c78d858a49e8a6325","abstract_canon_sha256":"d10c676ce8323e7d68784b3a0ea0a7f1fffaabc94a48e2258174993e0971f891"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-20T00:03:34.869620Z","signature_b64":"PZoFPSOoEzvprp1E51EsVzn+/C6DhlXIAx/2GWUg3SCy+gs1ycQMOx+d3dMKhGixf7pBw6zx/IDT93Z6leGQBA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"3a1c4a9d02b707b025dfb096f654a170f0e78f3bcc102fa70c829e0ff46fc295","last_reissued_at":"2026-05-20T00:03:34.868769Z","signature_status":"signed_v1","first_computed_at":"2026-05-20T00:03:34.868769Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Response-free item difficulty modelling for multiple-choice items with fine-tuned transformers: Component-wise representation and multi-task learning","license":"http://creativecommons.org/licenses/by/4.0/","headline":"Fine-tuned transformers predict multiple-choice item difficulty directly from wording, with multi-task learning aiding small-sample cases.","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Jan Net\\'ik, Patr\\'icia Martinkov\\'a","submitted_at":"2026-05-16T13:22:57Z","abstract_excerpt":"Response-free item difficulty modelling promises to reduce reliance on response-based calibration but is intrinsically difficult on reading-comprehension multiple-choice items, where difficulty depends on inferential demands across wording components. Whereas most existing approaches extract item-text features and pass them to a separate statistical or machine-learning model, we fine-tune transformer encoders end-to-end on the item wording, eliminating the manual feature engineering and preprocessing that discards information. Moreover, two extensions to this joint-encoding approach are propos"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"the multi-task variant delivers significant paired improvements in the smallest-sample regime. Transformer fine-tuning, especially if regularised by a suitable auxiliary task, recovers a substantial share of the wording-derivable signal at training-set sizes typical of applied measurement.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That difficulty is sufficiently determined by surface and inferential features in the item wording alone, independent of the specific student population or test context; this enters when the authors treat the held-out test set performance as evidence of wording-derivable signal without population-specific calibration data.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"Fine-tuned transformers with multi-task learning recover substantial wording-derived signal for item difficulty at small sample sizes typical in applied testing.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"Fine-tuned transformers predict multiple-choice item difficulty directly from wording, with multi-task learning aiding small-sample cases.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"82549ec004795e0672177f21a889998efde881de0e5bb068b44473751b85c3cf"},"source":{"id":"2605.16991","kind":"arxiv","version":1},"verdict":{"id":"fac931e3-f110-48e5-acfe-e40c74dbb212","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-19T20:18:13.999786Z","strongest_claim":"the multi-task variant delivers significant paired improvements in the smallest-sample regime. Transformer fine-tuning, especially if regularised by a suitable auxiliary task, recovers a substantial share of the wording-derivable signal at training-set sizes typical of applied measurement.","one_line_summary":"Fine-tuned transformers with multi-task learning recover substantial wording-derived signal for item difficulty at small sample sizes typical in applied testing.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That difficulty is sufficiently determined by surface and inferential features in the item wording alone, independent of the specific student population or test context; this enters when the authors treat the held-out test set performance as evidence of wording-derivable signal without population-specific calibration data.","pith_extraction_headline":"Fine-tuned transformers predict multiple-choice item difficulty directly from wording, with multi-task learning aiding small-sample cases."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2605.16991/integrity.json","findings":[],"available":true,"detectors_run":[{"name":"doi_compliance","ran_at":"2026-05-19T20:31:32.954680Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"doi_title_agreement","ran_at":"2026-05-19T20:31:19.029725Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"citation_quote_validity","ran_at":"2026-05-19T19:50:03.340245Z","status":"skipped","version":"0.1.0","findings_count":0},{"name":"cited_work_retraction","ran_at":"2026-05-19T19:23:44.187858Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"claim_evidence","ran_at":"2026-05-19T18:41:56.206764Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"ai_meta_artifact","ran_at":"2026-05-19T18:33:26.295375Z","status":"skipped","version":"1.0.0","findings_count":0}],"snapshot_sha256":"4451a3563adb1bed7dd8d1567c212a2c1abcc783fed1cb072561fa577230bff8"},"references":{"count":190,"sample":[{"doi":"10.1007/s40593-023-00362-1","year":null,"title":"and Tamma, Valentina , date =","work_id":"70e37938-d271-4b69-9be8-54c251622088","ref_index":1,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":null,"title":"Belov, Dmitry and Lüdtke, Oliver and Ulitzsch, Esther , date =. A. OSF , doi =","work_id":"a5b64c6e-d568-48d9-90bf-1cca342c6d3d","ref_index":2,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":null,"title":"A quantitative study of","work_id":"9962e341-7da2-4e56-a5bc-a8f4eb52a228","ref_index":3,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":null,"title":"arXiv , doi =","work_id":"7be4bbe3-30e7-4ed5-b5b4-477b0745a45a","ref_index":5,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"10.1023/a:1007379606734","year":null,"title":"1997 , pages =","work_id":"4cb0ce88-187f-4820-8e5c-d015cb2dc503","ref_index":6,"cited_arxiv_id":"","is_internal_anchor":false}],"resolved_work":190,"snapshot_sha256":"4463b06485be078bb3e436bdc64620d6a82af3b9dcbdcf3ea2be76288707246f","internal_anchors":9},"formal_canon":{"evidence_count":1,"snapshot_sha256":"bb80a12177299a9826dfe845727f23aa4391d90a1a64addede8a1827cad84e5e"},"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":"2605.16991","created_at":"2026-05-20T00:03:34.868908+00:00"},{"alias_kind":"arxiv_version","alias_value":"2605.16991v1","created_at":"2026-05-20T00:03:34.868908+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.16991","created_at":"2026-05-20T00:03:34.868908+00:00"},{"alias_kind":"pith_short_12","alias_value":"HIOEVHICW4D3","created_at":"2026-05-20T00:03:34.868908+00:00"},{"alias_kind":"pith_short_16","alias_value":"HIOEVHICW4D3AJO7","created_at":"2026-05-20T00:03:34.868908+00:00"},{"alias_kind":"pith_short_8","alias_value":"HIOEVHIC","created_at":"2026-05-20T00:03:34.868908+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":0,"internal_anchor_count":0,"sample":[]},"formal_canon":{"evidence_count":1,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/HIOEVHICW4D3AJO7WCLPMVFBOD","json":"https://pith.science/pith/HIOEVHICW4D3AJO7WCLPMVFBOD.json","graph_json":"https://pith.science/api/pith-number/HIOEVHICW4D3AJO7WCLPMVFBOD/graph.json","events_json":"https://pith.science/api/pith-number/HIOEVHICW4D3AJO7WCLPMVFBOD/events.json","paper":"https://pith.science/paper/HIOEVHIC"},"agent_actions":{"view_html":"https://pith.science/pith/HIOEVHICW4D3AJO7WCLPMVFBOD","download_json":"https://pith.science/pith/HIOEVHICW4D3AJO7WCLPMVFBOD.json","view_paper":"https://pith.science/paper/HIOEVHIC","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2605.16991&json=true","fetch_graph":"https://pith.science/api/pith-number/HIOEVHICW4D3AJO7WCLPMVFBOD/graph.json","fetch_events":"https://pith.science/api/pith-number/HIOEVHICW4D3AJO7WCLPMVFBOD/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/HIOEVHICW4D3AJO7WCLPMVFBOD/action/timestamp_anchor","attest_storage":"https://pith.science/pith/HIOEVHICW4D3AJO7WCLPMVFBOD/action/storage_attestation","attest_author":"https://pith.science/pith/HIOEVHICW4D3AJO7WCLPMVFBOD/action/author_attestation","sign_citation":"https://pith.science/pith/HIOEVHICW4D3AJO7WCLPMVFBOD/action/citation_signature","submit_replication":"https://pith.science/pith/HIOEVHICW4D3AJO7WCLPMVFBOD/action/replication_record"}},"created_at":"2026-05-20T00:03:34.868908+00:00","updated_at":"2026-05-20T00:03:34.868908+00:00"}