{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:HF5GGUVKKOEKIOYTODYD6J6DHJ","short_pith_number":"pith:HF5GGUVK","schema_version":"1.0","canonical_sha256":"397a6352aa5388a43b1370f03f27c33a4fcb9799f688ff351e40f82233a00b29","source":{"kind":"arxiv","id":"2607.09438","version":1},"attestation_state":"computed","paper":{"title":"Test-Time Scaling for Small VLMs on Multilingual Visual MCQ","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CL","authors_text":"Peng-Jian Yang, Spiros Baxevanakis","submitted_at":"2026-07-10T14:09:04Z","abstract_excerpt":"Test-time scaling (TTS) reliably improves reasoning in large language models, but whether it transfers to small open vision-language models remains unclear. We examine this on EXAMS-V, a multilingual visual multiple-choice benchmark, comparing self-consistency, describe-then-reason with PRM-guided beam search, and two post-hoc selectors across Qwen2.5-VL-7B-Instruct and Qwen3.5-4B. What matters is the conditions under which TTS runs, not the search or verification machinery. The largest factor is parseability: an early prompt format left many chains reasoning correctly yet never committing to "},"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":"2607.09438","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2026-07-10T14:09:04Z","cross_cats_sorted":["cs.AI","cs.LG"],"title_canon_sha256":"61fb98deab1efa4bb6751c33581e3c454deada15e3e54ca863ad584ca192b4a1","abstract_canon_sha256":"a65820af756f834f0b69b442848fc7a8a89d57983aab2954a7ee40ca8eb76c6d"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-13T01:19:15.121613Z","signature_b64":"Ja80evED9pVGQPJ/qgbXctWgna5+i9mGjqJnml4ADLd0UKY0mWTruISP+1O7neJRJ7Sw7zZ0GP4ky9UE7dQnAQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"397a6352aa5388a43b1370f03f27c33a4fcb9799f688ff351e40f82233a00b29","last_reissued_at":"2026-07-13T01:19:15.120648Z","signature_status":"signed_v1","first_computed_at":"2026-07-13T01:19:15.120648Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Test-Time Scaling for Small VLMs on Multilingual Visual MCQ","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CL","authors_text":"Peng-Jian Yang, Spiros Baxevanakis","submitted_at":"2026-07-10T14:09:04Z","abstract_excerpt":"Test-time scaling (TTS) reliably improves reasoning in large language models, but whether it transfers to small open vision-language models remains unclear. We examine this on EXAMS-V, a multilingual visual multiple-choice benchmark, comparing self-consistency, describe-then-reason with PRM-guided beam search, and two post-hoc selectors across Qwen2.5-VL-7B-Instruct and Qwen3.5-4B. What matters is the conditions under which TTS runs, not the search or verification machinery. The largest factor is parseability: an early prompt format left many chains reasoning correctly yet never committing to "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2607.09438","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/2607.09438/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":"2607.09438","created_at":"2026-07-13T01:19:15.121116+00:00"},{"alias_kind":"arxiv_version","alias_value":"2607.09438v1","created_at":"2026-07-13T01:19:15.121116+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2607.09438","created_at":"2026-07-13T01:19:15.121116+00:00"},{"alias_kind":"pith_short_12","alias_value":"HF5GGUVKKOEK","created_at":"2026-07-13T01:19:15.121116+00:00"},{"alias_kind":"pith_short_16","alias_value":"HF5GGUVKKOEKIOYT","created_at":"2026-07-13T01:19:15.121116+00:00"},{"alias_kind":"pith_short_8","alias_value":"HF5GGUVK","created_at":"2026-07-13T01:19:15.121116+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2608.01207","citing_title":"It's the Decoding Format, Not the Perturbation: Auditing Consistency-Based Selection for Vision-Language Test-Time Scaling","ref_index":2026,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/HF5GGUVKKOEKIOYTODYD6J6DHJ","json":"https://pith.science/pith/HF5GGUVKKOEKIOYTODYD6J6DHJ.json","graph_json":"https://pith.science/api/pith-number/HF5GGUVKKOEKIOYTODYD6J6DHJ/graph.json","events_json":"https://pith.science/api/pith-number/HF5GGUVKKOEKIOYTODYD6J6DHJ/events.json","paper":"https://pith.science/paper/HF5GGUVK"},"agent_actions":{"view_html":"https://pith.science/pith/HF5GGUVKKOEKIOYTODYD6J6DHJ","download_json":"https://pith.science/pith/HF5GGUVKKOEKIOYTODYD6J6DHJ.json","view_paper":"https://pith.science/paper/HF5GGUVK","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2607.09438&json=true","fetch_graph":"https://pith.science/api/pith-number/HF5GGUVKKOEKIOYTODYD6J6DHJ/graph.json","fetch_events":"https://pith.science/api/pith-number/HF5GGUVKKOEKIOYTODYD6J6DHJ/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/HF5GGUVKKOEKIOYTODYD6J6DHJ/action/timestamp_anchor","attest_storage":"https://pith.science/pith/HF5GGUVKKOEKIOYTODYD6J6DHJ/action/storage_attestation","attest_author":"https://pith.science/pith/HF5GGUVKKOEKIOYTODYD6J6DHJ/action/author_attestation","sign_citation":"https://pith.science/pith/HF5GGUVKKOEKIOYTODYD6J6DHJ/action/citation_signature","submit_replication":"https://pith.science/pith/HF5GGUVKKOEKIOYTODYD6J6DHJ/action/replication_record"}},"created_at":"2026-07-13T01:19:15.121116+00:00","updated_at":"2026-07-13T01:19:15.121116+00:00"}