mlsanityData to 7 Oct 2026

COSMO-RL: Towards Trustworthy LMRMs via Joint Safety and Stability

arXiv 2510.04196 ↗First posted 5 Oct 2025cs.AI cs.LGcited by 0

On the map: LLM reasoning → Reasoning for agents, planning and code

also in 1 more
Abstract (arXiv)

Large Multimodal Reasoning Models (LMRMs) are moving into real applications, where they must be both useful and safe. Safety is especially challenging in multimodal settings: images and text can be combined to bypass guardrails, and single objective training can cause policy drift that yields over-refusal on benign inputs or unsafe compliance on risky ones. We present COSMO-RL, a mixed reinforcement learning framework that trains reasoning oriented LMRMs under multimodal, multitask, and multiobjective signals, and we release the resulting model, COSMO-R1. Our approach aims to let safety and capability grow together in one stable pipeline rather than competing during alignment. In experiments, COSMO-R1 improves safety while maintaining-and often improving multimodal reasoning and instruction following, shows stronger robustness to multimodal jailbreaks, and reduces unnecessary refusals. The framework also transfers across backbones with consistent gains. Ablations support the design choices, indicating a simple path to advancing safety and general capability together in LMRMs.

Key facts. No citing papers in the map yet (data through 7 Oct 2026). Placed in Reasoning for agents, planning and code, one of 17 directions in LLM reasoning. Builds on 23 of its 27 references that are in the map. First posted on arXiv 5 Oct 2025.

Builds on

23 of 27 references.

Closest ancestors · 5

Same direction, newest first

  1. 2025RealSafe-R1: Safety-Aligned DeepSeek-R1 without Compromising Reasoning Capability14 Apr
  2. 2025SafeMLRM: Demystifying Safety in Multi-modal Large Reasoning Models9 Apr
  3. 2025Safety Tax: Safety Alignment Makes Your Large Reasoning Models Less Reasonable1 Mar
  4. 2025The Hidden Risks of Large Reasoning Models: A Safety Assessment of R118 Feb
  5. 2025SafeChain: Safety of Language Models with Long Chain-of-Thought Reasoning Capabilities17 Feb

Other references · 18

  1. 2025DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learningcited by 6,146
  2. 2023GPQA: A Graduate-Level Google-Proof Q&A Benchmarkcited by 2,748
  3. 2023MMMU: A Massive Multi-discipline Multimodal Understanding and Reasoning Benchmark for Expert AGIcited by 2,044
  4. 2023MathVista: Evaluating Mathematical Reasoning of Foundation Models in Visual Contextscited by 1,582
  5. 2024OlympiadBench: A Challenging Benchmark for Promoting AGI with Olympiad-Level Bilingual Multimodal Scientific Problemscited by 1,099
  6. 2023Aligning Large Multimodal Models with Factually Augmented RLHFcited by 703
  7. 2025VLM-R1: A Stable and Generalizable R1-style Large Vision-Language Modelcited by 515
  8. 2023XSTest: A Test Suite for Identifying Exaggerated Safety Behaviours in Large Language Modelscited by 477
  9. 2023RLHF-V: Towards Trustworthy MLLMs via Behavior Alignment from Fine-grained Correctional Human Feedbackcited by 404
  10. 2023MM-SafetyBench: A Benchmark for Safety Evaluation of Multimodal Large Language Modelscited by 255
  11. 2024JailBreakV: A Benchmark for Assessing the Robustness of MultiModal Large Language Models against Jailbreak Attackscited by 208
  12. 2024LLaVA-Critic: Learning to Evaluate Multimodal Modelscited by 132
6 more
  1. 2024mDPO: Conditional Preference Optimization for Multimodal Large Language Modelscited by 92
  2. 2024Tuning Large Multimodal Models for Videos using Reinforcement Learning from AI Feedbackcited by 36
  3. 2023Flames: Benchmarking Value Alignment of LLMs in Chinesecited by 35
  4. 2024Safe Inputs but Unsafe Output: Benchmarking Cross-modality Safety Alignment of Large Vision-Language Modelcited by 24
  5. 2025CPGD: Toward Stable Rule-based Reinforcement Learning for Language Modelscited by 11
  6. 2024MM-MoralBench: A MultiModal Moral Evaluation Benchmark for Large Vision-Language Modelscited by 9

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2025SaFeR-VLM: Toward Safety-aware Fine-grained Reasoning in Multimodal Models
  2. 2025SafeGRPO: Self-Rewarded Multimodal Safety Alignment via Rule-Governed Policy Optimization
  3. 2025When Safe Unimodal Inputs Collide: Optimizing Reasoning Chains for Cross-Modal Safety in Multimodal Large Language Models
  4. 2025SafeWork-R1: Coevolving Safety and Intelligence under the AI-45circ Law
  5. 2025Think in Safety: Unveiling and Mitigating Safety Alignment Collapse in Multimodal Large Reasoning Model
  6. 2025Safe RLHF-V: Safe Reinforcement Learning from Multi-modal Human Feedback
  7. 2025Beyond SFT: Reinforcement Learning for Safer Large Reasoning Models with Better Reasoning Ability
  8. 2025Breaking the Safety-Capability Tradeoff: Reinforcement Learning with Verifiable Rewards Maintains Safety Guardrails in LLMs