{"slug": "sage-sense-aligned-multimodal-grounding-evaluation-with-gloss-augmented", "title": "SAGE: Sense-Aligned Multimodal Grounding Evaluation with Gloss-Augmented Reasoning", "summary": "A new arXiv paper (2610.31838v1) introduces SAGE, a unified 10-candidate image-selection benchmark that reformulates VWSD, VerSe and WikiDiverse into a shared term-centric protocol covering noun-, verb- and entity-centered grounding, and proposes Gloss-Augmented Reasoning (GAR), a training-free framework combining definition grounding with constraint-guided semantic verification. With Gemma3-12B, combining the two signals achieved the best macro-average performance, raising Hits@1 from 65.90 to 71.22, while definition augmentation improved all three subsets and constraint-guided reasoning produced particularly strong gains on VerSe but was less effective on WikiDiverse. The authors report that strong VWSD performance does not transfer uniformly across grounding settings, positioning SAGE as a tool for unified evaluation and fine-grained analysis of explicit semantic reasoning.", "body_md": "arXiv:2610.31838v1 Announce Type: new \nAbstract: This paper introduces Sense-Aligned Grounding Evaluation (SAGE), a unified 10-candidate image-selection benchmark that reformulates VWSD, VerSe, and WikiDiverse into a shared term-centric protocol covering noun-, verb-, and entity-centered grounding. SAGE includes confusable hard negatives and a context-swapped VerSe diagnostic to probe target-term grounding beyond direct caption--image compatibility. We further propose Gloss-Augmented Reasoning (GAR), a training-free framework combining definition grounding with constraint-guided semantic verification. Experiments show that strong VWSD performance does not transfer uniformly across grounding settings. With Gemma3-12B, definition augmentation improves all three subsets, while constraint-guided reasoning improves VWSD and yields particularly strong gains on VerSe, but is less effective on WikiDiverse. Combining the two signals achieves the best macro-average performance, improving Hits@1 from 65.90 to 71.22. These results show that SAGE supports both unified evaluation and fine-grained analysis of explicit semantic reasoning across grounding settings.", "url": "https://wpnews.pro/news/sage-sense-aligned-multimodal-grounding-evaluation-with-gloss-augmented", "canonical_source": "https://arxiv.org/abs/2610.31838", "published_at": "2026-10-26 04:00:00+00:00", "updated_at": "2026-10-26 04:13:15.738579+00:00", "lang": "en", "topics": ["ai-research", "computer-vision", "natural-language-processing", "machine-learning"], "entities": ["SAGE", "Gloss-Augmented Reasoning", "VWSD", "VerSe", "WikiDiverse", "Gemma3-12B", "arXiv"], "also_reported_by": [], "alternates": {"html": "https://wpnews.pro/news/sage-sense-aligned-multimodal-grounding-evaluation-with-gloss-augmented", "markdown": "https://wpnews.pro/news/sage-sense-aligned-multimodal-grounding-evaluation-with-gloss-augmented.md", "text": "https://wpnews.pro/news/sage-sense-aligned-multimodal-grounding-evaluation-with-gloss-augmented.txt", "jsonld": "https://wpnews.pro/news/sage-sense-aligned-multimodal-grounding-evaluation-with-gloss-augmented.jsonld"}}