| Qwen3.5: Towards Native Multimodal Agents | ✓ Link | 88.6 | Qwen3.5-397B-A17B | 2026-02-16 |
| dots3-note Preview: A Small but Mighty Step Toward Long-Horizon Agency in Real Life | ✓ Link | 87.66 | dots3-note Preview | 2026-08-13 |
| Kimi K2.6 | ✓ Link | 87.4 | Kimi K2.6 | 2026-04-20 |
| Qwen3.5: Towards Native Multimodal Agents | ✓ Link | 86.2 | Qwen3.5-122B-A10B | 2026-02-16 |
| Qwen3.5: Towards Native Multimodal Agents | ✓ Link | 86.0 | Qwen3.5-27B | 2026-02-16 |
| Gemma 4 Technical Report | ✓ Link | 85.6 | Gemma 4 31B | 2026-07-02 |
| Kimi K2.5: Visual Agentic Intelligence | ✓ Link | 84.2 | Kimi K2.5 | 2026-02-02 |
| Qwen3.5: Towards Native Multimodal Agents | ✓ Link | 83.9 | Qwen3.5-35B-A3B | 2026-02-16 |
| HopChain: Multi-Hop Data Synthesis for Generalizable Vision-Language Reasoning | | 83.71 | Qwen3.5-397B-A17B (RLVR w/ Multi-Hop) | 2026-03-17 |
| Gemma 4 Technical Report | ✓ Link | 82.4 | Gemma 4 26B-A4B | 2026-07-02 |
| Gemma 4 Technical Report | ✓ Link | 79.7 | Gemma 4 12B Unified | 2026-07-02 |
| SenseNova-U1: Unifying Multimodal Understanding and Generation with NEO-unify Architecture | ✓ Link | 79.63 | SenseNova-U1-A3B-MoT | 2026-05-12 |
| Vero: Open Reinforcement Learning Recipes for Visual Reasoning | | 79.0 | Vero-Qwen35-9B | 2026-09-08 |
| Qwen3.5: Towards Native Multimodal Agents | ✓ Link | 78.9 | Qwen3.5-9B | 2026-02-16 |
| Priming Base Models for Agentic Post-Training: An Open Midtraining Recipe | | 78.16 | NVIDIA Nemotron 3.5 Super VL 120B-A12B BF16 | 2026-10-15 |
| STEP3-VL-10B Technical Report | ✓ Link | 75.95 | Step3-VL-10B (PaCoRe) | 2026-01-14 |
| EXAONE 4.5 Technical Report | ✓ Link | 75.2 | EXAONE 4.5 (Reasoning) | 2026-04-09 |
| Qwen3-VL Technical Report | ✓ Link | 74.6 | Qwen3-VL-235B-A22B-Thinking | 2025-11-26 |
| ERNIE 5.0 Technical Report | | 74.34 | ERNIE 5.0 | 2026-02-04 |
| SenseNova-U1.5: Towards Native Unified Visual Intelligence | ✓ Link | 73.55 | SenseNova-U1.5 8B-Think | 2026-09-10 |
| Qwen3.5-Omni Technical Report | | 73.0 | Qwen3.5-Omni-Plus | 2026-04-21 |
| Self-Supervised Visual On-Policy Distillation | | 72.66 | S^2VOPD (Qwen3.5-4B, FineVision-12K) (extra data) | 2026-08-14 |
| DiffusionGemma Technical Report | ✓ Link | 70.5 | DiffusionGemma 26B A4B | 2026-07-31 |
| CodePercept: Code-Grounded Visual STEM Perception for MLLMs | ✓ Link | 69.96 | CodePercept (32B, S1) | 2026-03-11 |
| LongCat-Next: Lexicalizing Modalities as Discrete Tokens | ✓ Link | 64.7 | LongCat-Next (A3B, 68.5B) | 2026-03-29 |
| See Less, See Right: Bi-directional Perceptual Shaping For Multimodal Reasoning | ✓ Link | 63.9 | BiPS (Qwen3-VL-8B-Thinking) | 2025-12-26 |
| Qwen3-Omni Technical Report | ✓ Link | 63.8 | Qwen3-Omni-Flash-Thinking | 2025-09-22 |
| GLM-4.1V-Thinking: Towards Versatile Multimodal Reasoning with Scalable Reinforcement Learning | ✓ Link | 63.5 | GLM-4.6V | 2025-07-01 |
| Vero: An Open RL Recipe for General Visual Reasoning | ✓ Link | 63.5 | Vero-Qwen3T-8B | 2026-04-06 |
| V-Rubrics: Visual Faithfulness via Rubric-Based Reinforcement Learning | ✓ Link | 58.88 | V-Rubrics (GRPO w/ rubrics) (extra data) | 2026-08-26 |
| Co-Evolving Policy Distillation | | 57.88 | CoPD | 2026-04-29 |
| Learning Consistency in Reward Modeling for Multi-Modal Reasoning | | 57.2 | CRM-RL-8B (Ours, soft-reward) | 2026-09-08 |
| Kimi-VL Technical Report | ✓ Link | 56.9 | Kimi-VL-A3B-Thinking-2506 | 2025-04-10 |
| Guidance Contrastive Token Credit Assignment for Discrete Policy Optimization | ✓ Link | 56.9 | Qwen3-VL-Instruct-8B + GCPO | 2026-05-28 |
| Near-Future Policy Optimization | | 55.72 | AutoNPO | 2026-04-22 |
| Beyond Simply Environment Scaling: Designing Effective Environment Distributions for Multimodal Agent Learning | ✓ Link | 54.9 | AES+HDC (Qwen3-VL-8B-Instruct) | 2026-08-06 |
| Can We Trust the Teacher? Decoupled Credit Direction-Magnitude for Self-Distillation | | 54.11 | DCSD (Qwen3-VL-8B-Instruct) | 2026-09-28 |
| Ovis2.5 Technical Report | ✓ Link | 53.9 | Ovis2.5-9B | 2025-08-15 |
| Odysseus: Scaling VLMs to 100+ Turn Decision-Making in Games via Reinforcement Learning | | 53.52 | Odysseus (RL on SFT) | 2026-05-01 |
| OpenVLThinkerV2: A Generalist Multimodal Reasoning Model for Multi-domain Visual Tasks | ✓ Link | 53.4 | OpenVLThinkerV2 (Ours) | 2026-04-09 |
| Self-Distilled RLVR | | 52.73 | RLSD (Qwen3-VL-8B-Instruct) | 2026-04-03 |
| Context-Aware RL for Agentic and Multimodal LLMs | ✓ Link | 52.0 | ContextRL (trained from Qwen3-VL-8B) | 2026-06-15 |
| Learning Adaptive Reasoning Paths for Efficient Visual Reasoning | ✓ Link | 49.7 | AVR-8B | 2026-04-16 |
| Insight-V++: Towards Advanced Long-Chain Visual Reasoning with Multimodal Large Language Models | | 48.6 | Insight-V++ | 2026-03-18 |
| Kwai Keye-VL 1.5 Technical Report | ✓ Link | 46.8 | Keye-VL-1.5-8B | 2025-09-01 |
| Kwai Keye-VL Technical Report | ✓ Link | 46.0 | Keye-VL-8B | 2025-07-02 |
| ReaGEN: Adaptive Generation of Structured Chains-of-Thought for Efficient Multimodal Reasoning | ✓ Link | 44.6 | ReaGEN (Qwen3-VL-4B, 4 iter) | 2026-06-03 |
| LASER: A Corrective Lens for LVLMs via Visual Attention Preservation and Sink Suppression | ✓ Link | 44.6 | LASER | 2026-07-02 |
| Not All Prompts Are Equal: Exploration-Guided Prompt Scaffolding for Multimodal Reinforcement Post-Training | | 44.41 | Qwen3-VL-4B + EPS (MMK12) | 2026-09-14 |
| AnE: Pushing the Reasoning Frontier of Multimodal LLMs via Anchor Evolution | ✓ Link | 43.9 | AnE-3rd | 2026-05-25 |
| Maestro: Reinforcement Learning to Orchestrate Hierarchical Model-Skill Ensembles | ✓ Link | 43.4 | Maestro (4B orchestrator) | 2026-05-21 |
| InternVL3: Exploring Advanced Training and Test-Time Recipes for Open-Source Multimodal Models | ✓ Link | 43.1 | InternVL3-78B | 2025-04-14 |
| Beyond Layer-Wise Merging: Chain-of-Merging for Vision-Language Models | | 40.95 | CoM (Qwen2.5-VL-72B + Qwen2.5-Math-72B) | 2026-06-03 |
| Dynamic Important Example Mining for Reinforcement Finetuning | ✓ Link | 40.3 | Qwen2.5-VL-32B + DIEM | 2026-08-29 |
| LoopVL: Recurrent Visual Intelligence | ✓ Link | 38.49 | LoopVL-1B | 2026-09-29 |
| Qwen2.5-VL Technical Report | ✓ Link | 38.1 | Qwen2.5-VL-72B | 2025-02-19 |
| MetaphorStar: Image Metaphor Understanding and Reasoning with End-to-End Visual Reinforcement Learning | ✓ Link | 38.1 | MetaphorStar-32B | 2026-02-11 |
| From Narrow to Panoramic Vision: Attention-Guided Cold-Start Reshapes Multimodal Reasoning | ✓ Link | 37.4 | AVAR-Thinker | 2026-03-04 |
| Intern Lumina U2: A Multi-Codebook Diffusion Large Language Model for Omni-Visual Understanding and Image Generation | ✓ Link | 36.5 | Intern Lumina U2 (Ascend) | 2026-10-14 |
| Co-RL: Unsupervised Reasoning Emerges from Diverse Cohort in Multi-agent RL | ✓ Link | 35.3 | Co-RL (InternVL-3.5-8B, Different family) | 2026-08-19 |
| Thinking in Uncertainty: Mitigating Hallucinations in MLRMs with Latent Entropy-Aware Decoding | ✓ Link | 33.1 | VL-Rethinker-7B with LEAD | 2026-03-09 |
| Beyond Where to Look: Trajectory-Guided Reinforcement Learning for Multimodal RLVR | | 31.26 | TGRL-DAPO (Qwen2.5-VL-7B, 39K) | 2026-03-27 |
| Puzzle Curriculum GRPO for Vision-Centric Reasoning | ✓ Link | 30.92 | Qwen2.5-VL-7B (Jigsaw) | 2025-12-16 |
| Improving Vision-language Models with Perception-centric Process Reward Models | ✓ Link | 30.92 | Qwen2.5-VL (7B) with Perceval | 2026-04-27 |
| When Models Judge Themselves: Unsupervised Self-Evolution for Multimodal Reasoning | ✓ Link | 30.9 | Qwen2.5-VL-7B + SelfJudge (Geo3K) | 2026-03-22 |
| Unlocking Token Rewards via Training-Free Reward Attribution | | 30.9 | Qwen2.5-VL-7B-Instruct + GRPO (P2T token reward) (extra data) | 2026-06-03 |
| CPPO: Contrastive Perception for Vision Language Policy Optimization | ✓ Link | 29.9 | CPPO (Qwen2.5-VL-7B) | 2026-01-01 |
| VideoAuto-R1: Video Auto Reasoning via Thinking Once, Answering Twice | ✓ Link | 29.6 | VideoAuto-R1 (Qwen2.5-VL-7B) | 2026-01-08 |
| PointThinker: Point-Incentivized Parallel Thinking for Multimodal Large Language Model | | 28.1 | PointThinker-7B (extra data) | 2026-06-03 |
| Length-Unbiased Sequence Policy Optimization: Revealing and Controlling Response Length Variation in RLVR | ✓ Link | 28.0 | Qwen2.5-VL-7B-Instruct (LUSPO) | 2026-02-05 |
| Process Rewards with Learned Reliability | ✓ Link | 26.65 | ACA with BetaPRM (Qwen2.5-VL-7B) | 2026-05-15 |
| VideoLLaMA 3: Frontier Multimodal Foundation Models for Image and Video Understanding | ✓ Link | 26.2 | VideoLLaMA3 7B | 2025-01-22 |
| Qwen2-VL: Enhancing Vision-Language Model's Perception of the World at Any Resolution | ✓ Link | 25.9 | Qwen2-VL-72B | 2024-09-18 |
| ARM-Thinker: Reinforcing Multimodal Generative Reward Models with Agentic Tool Use and Visual Reasoning | ✓ Link | 25.9 | ARM-Thinker-7B | 2025-12-04 |
| Decouple to Generalize: Context-First Self-Evolving Learning for Data-Scarce Vision-Language Reasoning | ✓ Link | 25.2 | DoGe-7B (Iter3) | 2025-12-07 |
| Qwen2.5-VL Technical Report | ✓ Link | 25.1 | Qwen2.5-VL-7B | 2025-02-19 |
| Kimi-VL Technical Report | ✓ Link | 21.4 | Kimi-VL-A3B-Instruct | 2025-04-10 |
| SVR-R1: Bootstrapping Multi-modal Reasoning with Self-verification in Reinforcement Learning | | 19.1 | SVR-R1 (7B ThinkLite) | 2026-07-13 |
| DiffusionVL: Translating Any Autoregressive Models into Diffusion Vision Language Models | ✓ Link | 19.08 | DiffusionVL | 2025-12-17 |
| Qwen2-VL: Enhancing Vision-Language Model's Perception of the World at Any Resolution | ✓ Link | 16.3 | Qwen2-VL-7B | 2024-09-18 |
| VideoLLaMA 3: Frontier Multimodal Foundation Models for Image and Video Understanding | ✓ Link | 15.5 | VideoLLaMA3 2B | 2025-01-22 |
| Qwen2-VL: Enhancing Vision-Language Model's Perception of the World at Any Resolution | ✓ Link | 12.4 | Qwen2-VL-2B | 2024-09-18 |