SD-MAR: Multi-image Analytical Reasoning via Synthetic Data and Reinforcement Learning
SD-MAR is a framework for training and evaluating vision-language models (VLMs) on multi-image analytical reasoning tasks. It constructs paired visual scenarios through controlled perturbations and generates reasoning tasks spanning semantic change attribution and quantitative comparison. Using GRPO-lite with Backward Discounted Allocation (BDA), a reinforcement learning approach that removes KL regularization, fine-tuning on SD-MAR improves in-domain accuracy by up to 36.95% on Qwen2.5-VL-7B and InternVL3-8B. Qwen2.5-VL-7B outperforms GPT-4.1 on the SD-MAR benchmark. Out-of-domain generalization is preserved or improved, with performance within 1% on MME, MMMU-Pro, MathVista and up to 4% improvement on MMBench. LLM-as-judge evaluation shows consistent improvements in logical coherence and explanation quality.
-->
[Submitted on 15 Jul 2026]
Title:SD-MAR: Multi-image Analytical Reasoning via Synthetic Data and Reinforcement Learning
View a PDF of the paper titled SD-MAR: Multi-image Analytical Reasoning via Synthetic Data and Reinforcement Learning, by Shiyu Yuan and 5 other authors
View PDF HTML (experimental)
Abstract:Vision Language Models (VLMs) demonstrate strong perceptual abilities but remain limited in tasks requiring analytical reasoning across multiple visual states, such as multi-image comparison, change detection, and multi-step visual inference. These capabilities are critical for real-world multimodal applications where reasoning must be grounded in systematic differences between visual contexts. However, existing benchmarks rarely require both explicit visual comparison and analytical reasoning, leaving this capability underexplored. To address this gap, we introduce SD-MAR (Synthetic Data for Multi-image Analytical Reasoning), a framework for training and evaluating VLMs on multi-image analytical reasoning. SD-MAR constructs paired visual scenarios through controlled perturbations and generates reasoning tasks spanning semantic change attribution and quantitative comparison. We further train VLMs using GRPO-lite with Backward Discounted Allocation (BDA), a reinforcement learning approach that removes KL regularization to encourage stronger policy optimization while allocating greater credit to the later reasoning steps where analytical conclusions are formed. Experiments on Qwen2.5-VL-7B and InternVL3-8B show that GRPO-lite fine-tuning on SD-MAR improves in-domain accuracy by up to 36.95%, with Qwen2.5-VL-7B outperforming GPT-4.1 on the SD-MAR benchmark. Importantly, out-of-domain generalization is preserved or improved: performance remains within 1% on MME, MMMU-Pro, and MathVista, while improving by up to 4% on MMBench. LLM-as-judge evaluation further demonstrates consistent improvements in logical coherence and explanation quality across both models.
Subjects:
Computer Vision and Pattern Recognition (cs.CV); Computation and Language (cs.CL)
Cite as: arXiv:2607.14333 [cs.CV]
(or arXiv:2607.14333v1 [cs.CV] for this version)
https://doi.org/10.48550/arXiv.2607.14333
arXiv-issued DOI via DataCite (pending registration)
Submission history
From: Shiyu Yuan [view email] [v1] Wed, 15 Jul 2026 19:55:01 UTC (42,498 KB)
Full-text links:
Access Paper:
View a PDF of the paper titled SD-MAR: Multi-image Analytical Reasoning via Synthetic Data and Reinforcement Learning, by Shiyu Yuan and 5 other authors
View PDF
HTML (experimental)
TeX Source
view license
Current browse context:
cs.CV
new | recent | 2026-07
Change to browse by:
cs cs.CL
References & Citations
NASA ADS
Google Scholar
Semantic Scholar
Loading...
Data provided by:
Bibliographic Tools
Bibliographic and Citation Tools
Bibliographic Explorer Toggle
Bibliographic Explorer (What is the Explorer?)
Connected Papers Toggle
Connected Papers (What is Connected Papers?)
Litmaps Toggle
Litmaps (What is Litmaps?)
scite.ai Toggle
scite Smart Citations (What are Smart Citations?)
Code, Data, Media
Code, Data and Media Associated with this Article
alphaXiv Toggle
alphaXiv (What is alphaXiv?)
Links to Code Toggle
CatalyzeX Code Finder for Papers (What is CatalyzeX?)
DagsHub Toggle
DagsHub (What is DagsHub?)
GotitPub Toggle
Gotit.pub (What is GotitPub?)
Huggingface Toggle
Hugging Face (What is Huggingface?)
ScienceCast Toggle
ScienceCast (What is ScienceCast?)
Demos
Demos
Replicate Toggle
Replicate (What is Replicate?)
Spaces Toggle
Hugging Face Spaces (What is Spaces?)
Spaces Toggle
TXYZ.AI (What is TXYZ.AI?)
Related Papers
Recommenders and Search Tools
Link to Influence Flower
Influence Flower (What are Influence Flowers?)
Core recommender toggle
CORE Recommender (What is CORE?)
Author
Venue
Institution
Topic
About arXivLabs
arXivLabs: experimental projects with community collaborators
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Which authors of this paper are endorsers? | Disable MathJax (What is MathJax?)