mlsanityData to 7 Oct 2026

Stabilizing Unsupervised Self-Evolution of MLLMs via Continuous Softened Retracing reSampling

arXiv 2604.03647 ↗First posted 4 Apr 2026cs.CV cs.AIcited by 3

On the map: Multimodal LLMs / vision-language models → Embodied and 3D vision-language models

Abstract (arXiv)

In the unsupervised self-evolution of Multimodal Large Language Models, the quality of feedback signals during post-training is pivotal for stable and effective learning. However, existing self-evolution methods predominantly rely on majority voting to select the most frequent output as the pseudo-golden answer, which may stem from the model's intrinsic biases rather than guaranteeing the objective correctness of the reasoning paths. To counteract the degradation, we propose Continuous Softened Retracing reSampling (CSRS) in MLLM self-evolution. Specifically, we introduce a Retracing Re-inference Mechanism (RRM) that the model re-inferences from anchor points to expand the exploration of long-tail reasoning paths. Simultaneously, we propose Softened Frequency Reward (SFR), which replaces binary rewards with continuous signals, calibrating reward based on the answers' frequency across sampled reasoning sets. Furthermore, incorporated with Visual Semantic Perturbation (VSP), CSRS ensures the model prioritizes mathematical logic over visual superficiality. Experimental results demonstrate that CSRS significantly enhances the reasoning performance of Qwen2.5-VL-7B on benchmarks such as MathVision. We achieve state-of-the-art (SOTA) results in unsupervised self-evolution on geometric tasks. Our code is avaible at https://github.com/yyy195/CSRS.

Key facts. Cited by 3 papers in the map (data through 7 Oct 2026). Placed in Embodied and 3D vision-language models, one of 16 directions in Multimodal LLMs / vision-language models. Builds on 24 of its 31 references that are in the map. First posted on arXiv 4 Apr 2026.

Builds on

24 of 31 references.

Closest ancestors · 3

Same direction, newest first

  1. 2025VisPlay: Self-Evolving Vision-Language Models from Images19 Nov
  2. 2025Enhancing the Outcome Reward-based RL Training of MLLMs with Self-Consistency Sampling13 Nov
  3. 2025First SFT, Second RL, Third UPT: Continual Improving Multi-Modal LLM Reasoning via Unsupervised Post-Training28 May

Other references · 21

  1. 2017Proximal Policy Optimization Algorithmscited by 14,441
  2. 2023Direct Preference Optimization: Your Language Model is Secretly a Reward Modelcited by 7,916
  3. 2023MathVista: Evaluating Mathematical Reasoning of Foundation Models in Visual Contextscited by 1,582
  4. 2024Measuring Multimodal Mathematical Reasoning with MATH-Vision Datasetcited by 516
  5. 2021Inter-GPS: Interpretable Geometry Problem Solving with Formal Language and Symbolic Reasoningcited by 376
  6. 2023The Curse of Recursion: Training on Generated Data Makes Models Forgetcited by 343
  7. 2025Absolute Zero: Reinforced Self-play Reasoning with Zero Datacited by 323
  8. 2024We-Math: Does Your Large Multimodal Model Achieve Human-like Mathematical Reasoning?cited by 281
  9. 2021GeoQA: A Geometric Question Answering Benchmark Towards Multimodal Numerical Reasoningcited by 224
  10. 2024Meta-Rewarding Language Models: Self-Improving Alignment with LLM-as-a-Meta-Judgecited by 202
  11. 2025SPICE: Self-Play In Corpus Environments Improves Reasoningcited by 78
  12. 2025Can Large Reasoning Models Self-Train?cited by 66
9 more
  1. 2025Multi-Agent Evolve: LLM Self-Improve through Co-evolutioncited by 58
  2. 2025Evolving Language Models without Labels: Majority Drives Selection, Novelty Promotes Variationcited by 44
  3. 2025EvoLMM: Self-Evolving Large Multimodal Models with Continuous Rewardscited by 24
  4. 2025Bootstrapping Task Spaces for Self-Improvementcited by 10
  5. 2025GPO: Learning from Critical Steps to Improve LLM Reasoningcited by 10
  6. 2025OpenSIR: Open-Ended Self-Improving Reasonercited by 7
  7. 2026R3G: A Reasoning-Retrieval-Reranking Framework for Vision-Centric Answer Generationcited by 4
  8. 2025Beyond RLHF: A Unified Theoretical Framework of Alignmentcited by 3
  9. 2025MathSE: Improving Multimodal Mathematical Reasoning via Self-Evolving Iterative Reflection and Reward-Guided Fine-Tuningcited by 2

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2026: 3.

3 papers.

  1. 2026Beyond Visual Similarity: Entity-Aligned Retrieval for Knowledge-Based Visual Question Answeringcited by 0
  2. 2026Unsupervised Post-Training of Foundation Models: A Surveycited by 0
  3. 2026A Visual Dependence-Aware Framework for Multimodal Unsupervised Continual Post-Trainingcited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2026When Models Judge Themselves: Unsupervised Self-Evolution for Multimodal Reasoning
  2. 2026EVE: Verifiable Self-Evolution of MLLMs via Executable Visual Transformations
  3. 2026On the Generalization Gap in Self-Evolving Language Model Reasoning
  4. 2026M2Note: Continual Evolution of Vision Language Models via Mistake Notebook Learning
  5. 2024Efficient Self-Improvement in Multimodal Large Language Models: A Model-Level Judge-Free Approach
  6. 2026Paying More Attention to Visual Tokens in Self-Evolving Large Multimodal Models
  7. 2026RISE: Reliable Improvement in Self-Evolving Vision-Language Models
  8. 2024Diving into Self-Evolving Training for Multimodal Reasoning