mlsanityData to 14 Oct 2026

Ovis2.5 Technical Report

arXiv 2508.11737 ↗First posted 15 Aug 2025cs.CV cs.AI cs.CLcited by 108

On the map: LLM reasoning → Multimodal reasoning

Abstract (arXiv)

We present Ovis2.5, a successor to Ovis2 designed for native-resolution visual perception and strong multimodal reasoning. Ovis2.5 integrates a native-resolution vision transformer that processes images at their native, variable resolutions, avoiding the degradation from fixed-resolution tiling and preserving both fine detail and global layout -- crucial for visually dense content like complex charts. To strengthen reasoning, we train the model to move beyond linear chain-of-thought and perform reflection -- including self-checking and revision. This advanced capability is exposed as an optional "thinking mode" at inference time, allowing users to trade latency for enhanced accuracy on difficult inputs. The model is trained via a comprehensive five-phase curriculum that progressively builds its skills. The process begins with foundational visual and multimodal pretraining, advances through large-scale instruction tuning, and culminates in alignment and reasoning enhancement using DPO and GRPO. To scale these upgrades efficiently, we employ multimodal data packing and hybrid parallelism, yielding a significant end-to-end speedup. We release two open-source models: Ovis2.5-9B and Ovis2.5-2B. The latter continues the "small model, big performance" philosophy of Ovis2, making it ideal for resource-constrained, on-device scenarios. On the OpenCompass multimodal leaderboard, Ovis2.5-9B averages 78.3, marking a substantial improvement over its predecessor, Ovis2-8B, and achieving state-of-the-art results among open-source MLLMs in the sub-40B parameter range; Ovis2.5-2B scores 73.9, establishing SOTA for its size. Beyond aggregate scores, Ovis2.5 achieves leading results on STEM benchmarks, exhibits strong capabilities on grounding and video tasks, and achieves open-source SOTA at its scale for complex chart analysis.

Key facts. Cited by 108 papers in the map (data through 14 Oct 2026). Placed in Multimodal reasoning, one of 17 directions in LLM reasoning. Builds on 38 of its 48 references that are in the map. First posted on arXiv 15 Aug 2025.

Builds on

38 of 48 references.

Closest ancestors · 7

Same direction, newest first

  1. 2024DynaMath: A Dynamic Visual Benchmark for Evaluating Mathematical Reasoning Robustness of Vision Language Models29 Oct
  2. 2024LogicVista: Multimodal LLM Logical Reasoning Benchmark in Visual Contexts6 Jul
  3. 2024We-Math: Does Your Large Multimodal Model Achieve Human-like Mathematical Reasoning?1 Jul
  4. 2024Measuring Multimodal Mathematical Reasoning with MATH-Vision Dataset22 Feb
  5. 2023MathVista: Evaluating Mathematical Reasoning of Foundation Models in Visual Contexts3 Oct
  6. 2023MMBench: Is Your Multi-modal Model an All-around Player?12 Jul
  7. 2022ChartQA: A Benchmark for Question Answering about Charts with Visual and Logical Reasoning19 Mar

Other references · 31

  1. 2024DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Modelscited by 9,379
  2. 2025Qwen3 Technical Reportcited by 8,820
  3. 2023Segment Anythingcited by 8,234
  4. 2023Direct Preference Optimization: Your Language Model is Secretly a Reward Modelcited by 8,065
  5. 2025Qwen2.5-VL Technical Reportcited by 5,587
  6. 2024Qwen2.5 Technical Reportcited by 4,942
  7. 2021RoFormer: Enhanced Transformer with Rotary Position Embeddingcited by 4,703
  8. 2022LAION-5B: An open large-scale dataset for training next generation image-text modelscited by 4,152
  9. 2025Gemini 2.5: Pushing the Frontier with Advanced Reasoning, Multimodality, Long Context, and Next Generation Agentic Capabilitiescited by 3,796
  10. 2023MMMU: A Massive Multi-discipline Multimodal Understanding and Reasoning Benchmark for Expert AGIcited by 2,080
  11. 2019Megatron-LM: Training Multi-Billion Parameter Language Models Using Model Parallelismcited by 1,868
  12. 2025InternVL3: Exploring Advanced Training and Test-Time Recipes for Open-Source Multimodal Modelscited by 1,834
19 more
  1. 2019Towards VQA Models That Can Readcited by 1,718
  2. 2024Video-MME: The First-Ever Comprehensive Evaluation Benchmark of Multi-modal LLMs in Video Analysiscited by 1,413
  3. 2020DocVQA: A Dataset for VQA on Document Imagescited by 1,317
  4. 2023MM-Vet: Evaluating Large Multimodal Models for Integrated Capabilitiescited by 1,239
  5. 2023MVBench: A Comprehensive Multi-modal Video Understanding Benchmarkcited by 1,109
  6. 2024Are We on the Right Way for Evaluating Large Vision-Language Models?cited by 914
  7. 2015Generation and Comprehension of Unambiguous Object Descriptionscited by 904
  8. 2016A Diagram Is Worth A Dozen Imagescited by 804
  9. 2023DataComp: In search of the next generation of multimodal datasetscited by 644
  10. 2024MathVerse: Does Your Multi-modal LLM Truly See the Diagrams in Visual Math Problems?cited by 586
  11. 2023HallusionBench: An Advanced Diagnostic Suite for Entangled Language Hallucination and Visual Illusion in Large Vision-Language Modelscited by 583
  12. 2024BLINK: Multimodal Large Language Models Can See but Not Perceivecited by 550
  13. 2024MMMU-Pro: A More Robust Multi-discipline Multimodal Understanding Benchmarkcited by 512
  14. 2023OCRBench: On the Hidden Mystery of OCR in Large Multimodal Modelscited by 476
  15. 2024TempCompass: Do Video LLMs Really Understand Videos?cited by 363
  16. 2024MMT-Bench: A Comprehensive Multimodal Benchmark for Evaluating Large Vision-Language Models Towards Multitask AGIcited by 193
  17. 2024OCRBench v2: An Improved Benchmark for Evaluating Large Multimodal Models on Visual Text Localization and Reasoningcited by 173
  18. 2022Wukong: A 100 Million Large-scale Chinese Cross-modal Pre-training Benchmarkcited by 114
  19. 2025ChartQAPro: A More Diverse and Challenging Benchmark for Chart Question Answeringcited by 108

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2025: 34; 2026: 74.

108 papers.

Sort:

By area:

  1. 2025MiniCPM-V 4.5: Cooking Efficient MLLMs via Architecture, Data, and Training Recipecited by 186
  2. 2025Detect Anything via Next Point Predictioncited by 73
  3. 2025Don't Blind Your VLA: Aligning Visual Representations for OOD Generalizationcited by 35
  4. 2025MM-CoT:A Benchmark for Probing Visual Chain-of-Thought Reasoning in Multimodal Modelscited by 34
  5. 2026LocateAnything: Fast and High-Quality Vision-Language Grounding with Parallel Box Decodingcited by 28
  6. 2025From Perception to Cognition: A Survey of Vision-Language Interactive Reasoning in Multimodal Large Language Modelscited by 26
  7. 2025ENACT: Evaluating Embodied Cognition with World Modeling of Egocentric Interactioncited by 15
  8. 2025SAIL-VL2 Technical Reportcited by 14
  9. 2026HunyuanOCR-1.5: Making Lightweight OCR VLMs Faster and Bettercited by 14
  10. 2025Visual Multi-Agent System: Mitigating Hallucination Snowballing via Visual Flowcited by 13
  11. 2025FG-CLIP 2: A Bilingual Fine-grained Vision-Language Alignment Modelcited by 13
  12. 2026MMR-Bench: A Comprehensive Benchmark for Multimodal LLM Routingcited by 13
38 more
  1. 2025Efficient Test-Time Scaling for Small Vision-Language Modelscited by 12
  2. 2025OralGPT-Omni: A Versatile Dental Multimodal Large Language Modelcited by 11
  3. 2025PhysToolBench: Benchmarking Physical Tool Understanding for MLLMscited by 9
  4. 2026Triplets Better Than Pairs: Towards Stable and Effective Self-Play Fine-Tuning for LLMscited by 9
  5. 2026Refer-Agent: A Collaborative Multi-Agent System with Reasoning and Reflection for Referring Video Object Segmentationcited by 9
  6. 2025GSM8K-V: Can Vision Language Models Solve Grade School Math Word Problems in Visual Contextscited by 8
  7. 2025Video-LMM Post-Training: A Deep Dive into Video Reasoning with Large Multimodal Modelscited by 8
  8. 2025Small Drafts, Big Verdict: Information-Intensive Visual Reasoning via Speculationcited by 8
  9. 2025Ovis-Image Technical Reportcited by 8
  10. 2025Deep But Reliable: Advancing Multi-turn Reasoning for Thinking with Imagescited by 8
  11. 2026VISTA-Bench: Do Vision-Language Models Really Understand Visualized Text as Well as Pure Text?cited by 8
  12. 2025SPACE: Noise Contrastive Estimation Stabilizes Self-Play Fine-Tuning for Large Language Modelscited by 7
  13. 2025From Generated Human Videos to Physically Plausible Robot Trajectoriescited by 6
  14. 2025Evaluating Large Language Models on Multimodal Chemistry Olympiad Examscited by 6
  15. 2026Q-Zoom: Query-Aware Adaptive Perception for Efficient Multimodal Large Language Modelscited by 6
  16. 2026Self in Space: Benchmarking Self-Awareness and Spatial Cognition in UAV Embodied Intelligencecited by 6
  17. 2025ManipShield: A Unified Framework for Image Manipulation Detection, Localization and Explanationcited by 5
  18. 2025The Unseen Bias: How Norm Discrepancy in Pre-Norm MLLMs Leads to Visual Information Losscited by 5
  19. 2025FysicsWorld: A Unified Full-Modality Benchmark for Any-to-Any Understanding, Generation, and Reasoningcited by 5
  20. 2025A Benchmark for Omni-Modal Reasoning in Long Videoscited by 5
  21. 2026OmniFysics: Towards Physical Intelligence Evolution via Omni-Modal Signal Processing and Network Optimizationcited by 5
  22. 2026Wan-Weaver: Interleaved Multi-modal Generation via Decoupled Trainingcited by 5
  23. 2026Walk the Talk: Bridging the Reasoning-Action Gap for Thinking with Images via Multimodal Agentic Policy Optimizationcited by 5
  24. 2025PRISMM-Bench: A Benchmark of Peer-Review Grounded Multimodal Inconsistenciescited by 4
  25. 2025HERBench: A Benchmark for Multi-Evidence Integration in Video Question Answeringcited by 4
  26. 2026EditHF-1M: A Million-Scale Rich Human Preference Feedback for Image Editingcited by 4
  27. 2026TVI-CoT: Text-Visual Interleaved Chain-of-Thought Reasoning for Multimodal Understandingcited by 4
  28. 2025FRIEDA: Benchmarking Multi-Step Cartographic Reasoning in Vision-Language Modelscited by 3
  29. 2026MapTab: A Diagnostic Benchmark for Long-Horizon Multi-Criteria Multimodal Reasoning on Heterogeneous Topological Graphscited by 3
  30. 2026Probing Visual Concepts in Lightweight Vision-Language Models for Automated Drivingcited by 3
  31. 2026CodePercept: Code-Grounded Visual STEM Perception for MLLMscited by 3
  32. 2026EgoMind: Activating Spatial Cognition through Linguistic Reasoning in MLLMscited by 3
  33. 2026AgentRx: A Benchmark Study of LLM Agents for Multimodal Clinical Prediction Taskscited by 3
  34. 2026SPACENUM: Revisiting Spatial Numerical Understanding in VLMscited by 3
  35. 2026OralAgent: Integrating Reasoning, Tools, and Knowledge for Interactive Dental Image Analysiscited by 3
  36. 2026Zone of Proximal Policy Optimization: Teacher in Prompts, Not Gradientscited by 3
  37. 2025Scope: Selective Cross-modal Orchestration of Visual Perception Expertscited by 2
  38. 2025GRAN-TED: Generating Robust, Aligned, and Nuanced Text Embedding for Diffusion Modelscited by 2

50 most cited of 108 All 108 (JSON)

Similar, not cited

Close in wording, no citation either way

  1. 2026Vero: An Open RL Recipe for General Visual Reasoning
  2. 2025BLINK-Twice: You see, but do you observe? A Reasoning Benchmark on Visual Perception
  3. 2025Image Tiling for High-Resolution Reasoning: Balancing Local Detail with Global Context
  4. 2025Simple o3: Towards Interleaved Vision-Language Reasoning
  5. 2025VERIFY: A Benchmark of Visual Explanation and Reasoning for Investigating Multimodal Reasoning Fidelity
  6. 2025LlamaV-o1: Rethinking Step-by-step Visual Reasoning in LLMs
  7. 2025Diagnosing Visual Reasoning: Challenges, Insights, and a Path Forward
  8. 2026VIEW2SPACE: Studying Multi-View Visual Reasoning from Sparse Observations