Work

V-Rubrics: Visual Faithfulness via Rubric-Based Reinforcement Learning

Year: 2026

Type: conference-paper

Abstract: Vision-language models can produce fluent answers that are insufficiently grounded in the visual evidence: a single unsupported object, chart value, or intermediate inference can undermine an otherwise plausible response.We argue that this is a credit-assignment failure in multimodal post-training.Scalar outcome rewards indicate whether an answer is acceptable, but do not identify which visual facts are grounded, which reasoning steps are valid, or which instruction constraints are missed.We introduce Visual Rubrics-Based Reinforcement Learning, which decomposes reference responses into atomic propositions and scores generated answers along Visual Faithfulness (VF), Reasoning Consistency (RC), and Instruction Following (IF).The resulting rubric items provide structured partial credit and localize rubric credit when supporting evidence spans are available.We first obtain an SFT checkpoint by fine-tuning Qwen3-VL-8B-Instruct on the public OpenMMReasoner-SFT-874K corpus, adapting OpenMMReasoner's cold-start data recipe.We construct V-Rubrics 50K, a 50,248-example training set from 17 visually grounded sources, by applying rule-based filters before deriving example difficulty from rejection-sampling scores and then annotating every example with Gemini-3-Pro under the same structured prompt and protocol.We train our model based on the same SFT checkpoint using component-wise, prefix-localized rubric credit.Experiments show that our rubricbased GRPO improves over both the shared SFT baseline and answer-only GRPO, with the largest gains on knowledge-oriented and visually grounded reasoning benchmarks.The results show rubrics as a useful reward abstraction for visual post-training.Q: What is represented in this image?Visual grounding Reasoning & answer 1 Identifying Objects Identify Sun (A), Earth (C), and Moon (D/G) VF Weight 5 Essential 2 Visual Alignment Recognize Sun, Earth, and Moon are in a straight line.VF Weight 5 Essential VF Important Weight 4 B and F point to the ocean tidal bulges.3 Tidal Bulge Obs. RC Important Alignment Effect 4Weight 4Explain that aligned gravity creates the highest high tides. Select Spring Tide IFWeight 5 EssentialAnswer is "spring tide" / option (3) / option (D). RCWeight 3 ImportantNeap tide would require a 90 ° Sun-Moon angle.Distinguish Neap Tide

Source: Conference on Empirical Methods in Natural Language Processing (EMNLP)

DOI: 10.18653/2026.emnlp-main.1558