[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-82348-en":3,"doc-seo-82348-105":29,"detail-sidebar-cat-0-en-105":82},{"code":4,"msg":5,"data":6},0,"success",{"doc_id":7,"user_id":8,"nickname":9,"user_avatar":10,"doc_module":4,"category_id":11,"category_name":12,"doc_title":13,"doc_description":14,"doc_content":15,"file_id":16,"file_url":17,"file_type":18,"file_size":19,"view_count":4,"is_deleted":4,"is_public":20,"is_downloadable":20,"audit_status":20,"page_count":21,"language":22,"language_code":23,"site_id":24,"html_lang":23,"table_of_contents":25,"faqs":26,"seo_title":13,"seo_description":14,"update_tm":27,"read_time":28},82348,1374391974585,"Genevieve","https://ap-avatar.wpscdn.com/davatar_276721f389ce27ea32af1340a28f341c",8,"Research & Report","Test-Time Scaling for Small VLMs on Multilingual Visual MCQ","Test-time scaling (TTS) can improve reasoning in large language models, but its effectiveness for small open vision-language models remains uncertain. The study evaluates TTS on EXAMS-V, a multilingual visual multiple-choice benchmark, comparing self-consistency, describe-then-reason with PRM-guided beam search, and multiple post-hoc selectors using Qwen2.5-VL-7B-Instruct and Qwen3.5-4B. Results show parseability and sufficient decoding budget dominate gains, while elaborate selection and generative critics add limited benefit once chains can finish; best configuration reaches 84.1% on ImageCLEF 2026 test.","Nika at ImageCLEF 2026 Task on Multimodal Reasoning: More Tokens, Fewer Trees —Test-Time Scaling for Small VLMs on Multilingual Visual MCQ  \nImageCLEF Lab at CLEF 2026  \nSpiros Baxevanakis1,†, Peng-Jian Yang1,†  \n1 University of Amsterdam, Science Park 904, 1098 XH Amsterdam, The Netherlands  \nAbstract  \nTest-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 multiplechoice 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 an answer letter, which a standard answer cue and a guided repair step largely remove. A larger decoding budget removes the rest: raising the per-chain token limit from 1k to 2k recovers 3.7 pp, whereas sampling more chains (8 to 16) adds only 0.15 pp. Once chains have room to finish, elaborate methods contribute little: PRM-guided beam search trails plain self-consistency by 0.39 ppat over eight times the cost, and neither a training-free generative critic nor a trained multimodal PRM beats majority vote across both policies. The largest gain comes instead from the policy model itself (+11 .4 pp) . Our best configuration reaches 84.1% on the held-out ImageCLEF 2026 test split, ranking first on the Visual MCQ leaderboard.  \nKeywords  \ntest-time scaling, vision-language models, multimodal reasoning, self-consistency, process reward models,  \n1. Introduction  \nMultimodal reasoning remains a challenge for vision–language models (VLMs) whenever questions combine structured visual content with multi-step inference across languages. Real exam questions area natural stress test of this combination, and the EXAMS-V benchmark [1] spans thirteen languages and twenty school subjects rendered as text-in-image panels. The 2026 ImageCLEF multimodal reasoning shared task [2, 3] fixes the practical constraints: a single A40 GPU, open-weight policies with at most 7B parameters, and a reproducibility mandate.  \nThe natural question is whether the inference-time techniques that advanced text-only language models [4, 5, 6] transfer to open VLMs under this budget. The evidence is cautionary: self-refinement degrades open-source VLMs at this scale [7], long sequential “thinking” often underperforms parallel sampling on small policies [8, 9], and discriminative reward models trained on mathematics fail to transfer to humanities and non-English content [10, 11] . A viable pipeline must externalise selection, prefer parallel sampling to long single-chain reasoning, and avoid iterative self-refinement loops.  \nWe investigate whether describe-then-reason with PRM-guided search beats flat self-consistency; how accuracy scales along two axes (chain count and per-chain token budget) under a single-A40 envelope [6]; where TTS helps and fails across subjects and languages [7]; and whether a training-free generative critic can beat majority vote on low-agreement questions, with a cross-model PRM [12] controlling for sycophantic self-scoring [13] . The latter extends generative-verifier findings [14, 15, 16] to a multimodal, multilingual setting under a ≤7B budget not previously studied to our knowledge.  \nCLEF 2026 Working Notes, 21 – 24 September 2026, Jena, Germany †  \nThese authors contributed equally.  \n$ [spiros.baxevanakis@student.uva.nl](spiros.baxevanakis@student.uva.nl) (S. Baxevanakis); [lesterpjy@gmail.com](lesterpjy@gmail.com) (P. Yang)  \n © 2026 Copyright for this paper by its authors. Use permitted under Creative Commons License Attribution 4.0 International (CC BY 4.0) .  \nSummary of findings. Across all four dimensions, the policy substrate and engineering","cbCaiqwxyScWGMk4","https://ap.wps.com/l/cbCaiqwxyScWGMk4","pdf",902743,1,14,"English","en",105,"# Introduction\n# Related Work\n# Method","[{\"question\":\"How does decoding budget affect accuracy compared with sampling more chains?\",\"answer\":\"Increasing per-chain token budget from 1k to 2k recovers about 3.7 percentage points, while increasing the number of chains from 8 to 16 adds only about 0.15 percentage points. Once chains have room to finish, more complex methods contribute little.\"}]",1784179807,35,{"code":4,"msg":30,"data":31},"ok",{"site_id":24,"language":23,"slug":32,"title":13,"keywords":33,"description":14,"schema_data":34,"social_meta":77,"head_meta":79,"extra_data":81,"updated_unix":27},"test-time-scaling-for-small-vlms-on-multilingual-visual-mcq","",{"@graph":35,"@context":76},[36,53,67],{"@type":37,"itemListElement":38},"BreadcrumbList",[39,43,47,50],{"item":40,"name":41,"@type":42,"position":20},"https://docshare.wps.com","Home","ListItem",{"item":44,"name":45,"@type":42,"position":46},"https://docshare.wps.com/document/","Document",2,{"item":48,"name":12,"@type":42,"position":49},"https://docshare.wps.com/document/research-report/",3,{"item":51,"name":13,"@type":42,"position":52},"https://docshare.wps.com/document/test-time-scaling-for-small-vlms-on-multilingual-visual-mcq/82348/",4,{"url":51,"name":13,"@type":54,"author":55,"headline":13,"publisher":57,"fileFormat":60,"inLanguage":23,"description":14,"dateModified":61,"datePublished":61,"encodingFormat":60,"isAccessibleForFree":62,"interactionStatistic":63},"DigitalDocument",{"name":9,"@type":56},"Person",{"url":40,"name":58,"@type":59},"DocShare","Organization","application/pdf","2026-07-16",true,{"@type":64,"interactionType":65,"userInteractionCount":4},"InteractionCounter",{"@type":66},"ViewAction",{"@type":68,"mainEntity":69},"FAQPage",[70],{"name":71,"@type":72,"acceptedAnswer":73},"How does decoding budget affect accuracy compared with sampling more chains?","Question",{"text":74,"@type":75},"Increasing per-chain token budget from 1k to 2k recovers about 3.7 percentage points, while increasing the number of chains from 8 to 16 adds only about 0.15 percentage points. Once chains have room to finish, more complex methods contribute little.","Answer","https://schema.org",{"og:url":51,"og:type":78,"og:title":13,"og:site_name":58,"og:description":14},"article",{"robots":80,"canonical":51},"index,follow",{"doc_id":7,"site_id":24},{"code":4,"msg":5,"data":83},[84,88,92,96,101,106,111,114,119,122,126],{"id":20,"doc_module":4,"doc_module_name":45,"category_name":85,"show_sort_weight":86,"slug":87},"Story & Novel",90,"story-novel",{"id":46,"doc_module":4,"doc_module_name":45,"category_name":89,"show_sort_weight":90,"slug":91},"Literature",80,"literature",{"id":52,"doc_module":4,"doc_module_name":45,"category_name":93,"show_sort_weight":94,"slug":95},"Exam",70,"exam",{"id":97,"doc_module":4,"doc_module_name":45,"category_name":98,"show_sort_weight":99,"slug":100},5,"Comic",60,"comic",{"id":102,"doc_module":4,"doc_module_name":45,"category_name":103,"show_sort_weight":104,"slug":105},6,"Technology",50,"technology",{"id":107,"doc_module":4,"doc_module_name":45,"category_name":108,"show_sort_weight":109,"slug":110},7,"Healthcare",40,"healthcare",{"id":11,"doc_module":4,"doc_module_name":45,"category_name":12,"show_sort_weight":112,"slug":113},30,"research-report",{"id":115,"doc_module":4,"doc_module_name":45,"category_name":116,"show_sort_weight":117,"slug":118},9,"Religion & Spirituality",20,"religion-spirituality",{"id":117,"doc_module":4,"doc_module_name":45,"category_name":120,"show_sort_weight":117,"slug":121},"World Cup","world-cup",{"id":123,"doc_module":4,"doc_module_name":45,"category_name":124,"show_sort_weight":123,"slug":125},10,"Lifestyle","lifestyle",{"id":127,"doc_module":4,"doc_module_name":45,"category_name":128,"show_sort_weight":97,"slug":129},19,"General","general"]