Newest papers

RACE: Learning Multimodal Search through Referential Abstraction and Compositional Evidence

Eric Tao, Minhao Sun, Liyan Liu, Zhijing Wu, Xiaohan Yu, Lang Mei, Li Huang, Chao Feng, Jinchao Ma, Sichen Kang, Siyu Mo, Chong Chen · 2026 · arXiv (Cornell University)

DOI: 10.48550/arxiv.2610.28382 · cited by 0 works (OpenAlex) · open-access PDF

Hindi

प्रशिक्षण मल्टीमॉडल खोज एजेंटों को डेटा की आवश्यकता होती है जो जवाब निर्धारित करने के लिए दृश्य-आधारित खोज आवश्यक बनाते हैं। टेक्स्टल पहचान संकेत और अप्रयुक्त लक्षित सबूत संश्लेषित प्रश्नों में तारों की अनुमति देते हैं, दृश्य खोज और सबूत एकीकरण के लिए पर्यवेक्षण को कमजोर करते हुए। हम RACE, एक छवि-आधारित कार्य संश्लेषण ढांचे की पेशकश करते हैं जो छवियों से लक्षित करने के लिए और लक्षित विशिष्ट सबूतों से उत्तरों तक पर निर्भरताओं का निर्माण करता है। संदर्भ अवलोकन (आरए) विशिष्ट दृश्य संदर्भों या चित्रित उम्मीदवारों के बीच खोज द्वारा हल किए गए अस्पष्ट रिश्तेदार संदर्भों के साथ पहचान-

Machine translation (Meta M2M100); the English original follows.

Abstract

Training multimodal search agents requires tasks that make visually grounded search necessary for determining the answer. Textual identity clues and unused target evidence allow shortcuts in synthetic questions, weakening supervision for visual search and evidence integration. We propose RACE, an image-conditioned task synthesis framework that constructs dependencies from images to targets and from target-specific evidence to answers. Referential Abstraction (RA) replaces identity-revealing text with specific visual references or ambiguous relational references resolved by searching among depicted candidates. Compositional Evidence (CE) links retrieved facts about multiple entities in one image or entities from different images into joint questions whose answers require every designated contribution. Both operations also apply to images found during search, allowing new visual observations to determine subsequent search targets or supply answer evidence. Fine-tuning Qwen3.5-27B on only 12K teacher-generated trajectories from RACE raises average accuracy across ten multimodal benchmarks from 53.5 to 56.0, a relative improvement of approximately 4.7% over the matched base agent. Our code, data, and model will be publicly available.

Read in: English · Spanish · Chinese · Hindi · Arabic · Portuguese · French · German · Japanese · Russian · Korean · Indonesian · Vietnamese

Topics: Abstraction, Computer science, Task (project management), Artificial intelligence, Visual search, Natural language processing

Sources: OpenAlex (CC0), arXiv, Crossref. Agent Exchange