First SFT, Second RL, Third UPT: Continual Improving Multi-Modal LLM Reasoning via Unsupervised Post-Training

• URL: http://arxiv.org/abs/2505.22453v2
• Date: Mon, 27 Oct 2025 09:06:32 GMT
• Title: First SFT, Second RL, Third UPT: Continual Improving Multi-Modal LLM Reasoning via Unsupervised Post-Training
• Authors: Lai Wei, Yuting Li, Chen Wang, Yue Wang, Linghe Kong, Weiran Huang, Lichao Sun,
• Abstract summary: We propose MM-UPT, a simple yet effective framework for unsupervised post-training of MLLMs.<n>Our experiments demonstrate that such training method effectively improves the reasoning ability of Qwen2.5-VL-7B.<n>We extend our framework to a data self-generation setting, designing two strategies that prompt the MLLM to synthesize new training samples.
• Score: 37.80193099472551
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Improving Multi-modal Large Language Models (MLLMs) in the post-training stage typically relies on supervised fine-tuning (SFT) or reinforcement learning (RL), which require expensive and manually annotated multi-modal data--an ultimately unsustainable resource. This limitation has motivated a growing interest in unsupervised paradigms as a third stage of post-training after SFT and RL. While recent efforts have explored this direction, their methods are complex and difficult to iterate. To address this, we propose MM-UPT, a simple yet effective framework for unsupervised post-training of MLLMs, enabling continual self-improvement without any external supervision. The training method of MM-UPT builds upon GRPO, replacing traditional reward signals with a self-rewarding mechanism based on majority voting over multiple sampled responses. Our experiments demonstrate that such training method effectively improves the reasoning ability of Qwen2.5-VL-7B (e.g., 66.3\%$\rightarrow$72.9\% on MathVista, 62.9\%$\rightarrow$68.7\% on We-Math), using standard dataset without ground truth labels. To further explore scalability, we extend our framework to a data self-generation setting, designing two strategies that prompt the MLLM to synthesize new training samples on its own. Additional experiments show that combining these synthetic data with the unsupervised training method can also boost performance, highlighting a promising approach for scalable self-improvement. Overall, MM-UPT offers a new paradigm for autonomous enhancement of MLLMs, serving as a critical third step after initial SFT and RL in the absence of external supervision. Our code is available at https://github.com/waltonfuture/MM-UPT.

Related papers

• S3-CoT: Self-Sampled Succinct Reasoning Enables Efficient Chain-of-Thought LLMs [48.80914119283909]
  Large language models equipped with chain-of-thought (CoT) achieve strong performance and offer a window into behavior.<n>Recent evidence suggests that improvements in CoT capabilities often come with redundant reasoning processes.<n>Our study presents a self-sampling framework based on activation steering for efficient CoT learning.
  arXiv  Detail & Related papers  (2026-02-02T11:37:36Z)
• Agentic Reinforced Policy Optimization [66.96989268893932]
  Large-scale reinforcement learning with verifiable rewards (RLVR) has demonstrated its effectiveness in harnessing the potential of large language models (LLMs) for single-turn reasoning tasks.<n>Current RL algorithms inadequately balance the models' intrinsic long-horizon reasoning capabilities and their proficiency in multi-turn tool interactions.<n>We propose Agentic Reinforced Policy Optimization (ARPO), a novel agentic RL algorithm tailored for training multi-turn LLM-based agents.
  arXiv  Detail & Related papers  (2025-07-26T07:53:11Z)
• Small LLMs Do Not Learn a Generalizable Theory of Mind via Reinforcement Learning [1.6114012813668932]
  Small language models (LLMs) struggle to develop a generic Theory of Mind (ToM) capability.<n> prolonged RL training leads to models hacking'' the statistical patterns of the training datasets.<n>This suggests the learned behavior is a form of narrow overfitting rather than the acquisition of a true, abstract ToM capability.
  arXiv  Detail & Related papers  (2025-07-21T16:47:59Z)
• High-Resolution Visual Reasoning via Multi-Turn Grounding-Based Reinforcement Learning [43.8114307203968]
  State-of-the-art large multi-modal models (LMMs) face challenges when processing high-resolution images.<n>In this paper, we propose Multi-turn Grounding-based Policy Optimization (MGPO)<n>MGPO enables LMMs to iteratively focus on key visual regions by automatically cropping sub-images.
  arXiv  Detail & Related papers  (2025-07-08T12:05:05Z)
• How to Train Your LLM Web Agent: A Statistical Diagnosis [102.04125085041473]
  We present the first statistically grounded study on compute allocation for LLM web-agent post-training.<n>Our approach uses a two-stage pipeline, training a Llama 3.1 8B student to imitate a Llama 3.3 70B teacher via supervised fine-tuning (SFT) and on-policy reinforcement learning.<n>Our results show that combining SFT with on-policy RL consistently outperforms either approach alone on both WorkArena and MiniWob++.
  arXiv  Detail & Related papers  (2025-07-05T17:12:33Z)
• Heterogeneous Group-Based Reinforcement Learning for LLM-based Multi-Agent Systems [25.882461853973897]
  We propose Multi-Agent Heterogeneous Group Policy Optimization (MHGPO), which guides policy updates by estimating relative reward advantages.<n>MHGPO eliminates the need for Critic networks, enhancing stability and reducing computational overhead.<n>We also introduce three group rollout sampling strategies that trade off between efficiency and effectiveness.
  arXiv  Detail & Related papers  (2025-06-03T10:17:19Z)
• Advancing Multimodal Reasoning via Reinforcement Learning with Cold Start [24.244577648817188]
  "aha moment" patterns are often attributed to emergent properties from reinforcement learning (RL)<n>We present a comprehensive study on enhancing multimodal reasoning through a two-stage approach.<n>Our experiments show that this combined approach consistently outperforms both SFT-only and RL-only methods.
  arXiv  Detail & Related papers  (2025-05-28T13:21:38Z)
• InternVL3: Exploring Advanced Training and Test-Time Recipes for Open-Source Multimodal Models [139.19991097260115]
  We introduce InternVL3, a significant advancement in the InternVL series featuring a native multimodal pre-training paradigm.<n>In particular, InternVL3-78B achieves a score of 72.2 on the MMMU benchmark, setting a new state-of-the-art among open-source MLLMs.<n>In pursuit of open-science principles, we will publicly release both the training data and model weights to foster further research and development in next-generation MLLMs.
  arXiv  Detail & Related papers  (2025-04-14T17:59:25Z)
• Will Pre-Training Ever End? A First Step Toward Next-Generation Foundation MLLMs via Self-Improving Systematic Cognition [89.50068130832635]
  Self-Improving cognition (SIcog) is a self-learning framework for constructing next-generation foundation MLLMs by multimodal knowledge.<n>We propose Chain-of-Description for step-by-step visual understanding and integrate structured Chain-of-Thought (CoT) reasoning to support in-depth multimodal reasoning.<n>Experiments demonstrate SIcog's effectiveness in developing MLLMs with enhanced multimodal cognition.
  arXiv  Detail & Related papers  (2025-03-16T00:25:13Z)
• MM-Eureka: Exploring the Frontiers of Multimodal Reasoning with Rule-based Reinforcement Learning [55.82649731348012]
  We introduce the MMK12 dataset and MM-EUREKA with 7B and 32B parameters.<n>The former is a high-quality multimodal mathematics reasoning dataset featuring diverse knowledge domains with human-verified answers and solution processes.<n>The latter is a multimodal model employing rule-based reinforcement learning utilizing online filtering and two-stage training strategy to enhance training stability.
  arXiv  Detail & Related papers  (2025-03-10T14:23:12Z)
• Satori: Reinforcement Learning with Chain-of-Action-Thought Enhances LLM Reasoning via Autoregressive Search [57.28671084993782]
  Large language models (LLMs) have demonstrated remarkable reasoning capabilities across diverse domains.<n>Recent studies have shown that increasing test-time computation enhances LLMs' reasoning capabilities.<n>We propose a two-stage training paradigm: 1) a small-scale format tuning stage to internalize the COAT reasoning format and 2) a large-scale self-improvement stage leveraging reinforcement learning.
  arXiv  Detail & Related papers  (2025-02-04T17:26:58Z)
• Large Language Model-Enhanced Multi-Armed Bandits [43.34246396804588]
  Large language models (LLMs) have been adopted to solve sequential decision-making tasks such as multi-armed bandits (MAB)<n>We propose an alternative approach which combines the strengths of classical MAB and LLMs.<n>We conduct empirical evaluations using both synthetic MAB tasks and experiments designed using real-world text datasets.
  arXiv  Detail & Related papers  (2025-02-03T07:19:05Z)
• URSA: Understanding and Verifying Chain-of-thought Reasoning in Multimodal Mathematics [23.80647785460245]
  Process Reward Models (PRMs) have shown promise in enhancing the mathematical reasoning capabilities of Large Language Models.<n>We take the first step toward unlocking the potential of PRMs in multimodal mathematical reasoning.<n>We introduce URSA, a three-stage Unfolding multimodal Process-Supervision Aided training framework.
  arXiv  Detail & Related papers  (2025-01-08T18:49:41Z)
• 7B Fully Open Source Moxin-LLM -- From Pretraining to GRPO-based Reinforcement Learning Enhancement [42.10844666788254]
  Moxin 7B is a fully open-source Large Language Models (LLMs) developed adhering to principles of open science, open source, open data, and open access.<n>We release the pre-training code and configurations, training and fine-tuning datasets, and intermediate and final checkpoints.<n> Experiments show that our models achieve superior performance in various evaluations such as zero-shot evaluation, few-shot evaluation, and CoT evaluation.
  arXiv  Detail & Related papers  (2024-12-08T02:01:46Z)
• MALT: Improving Reasoning with Multi-Agent LLM Training [66.9481561915524]
  MALT (Multi-Agent LLM Training) is a novel post-training strategy that divides the reasoning process into generation, verification, and refinement steps.<n>On MATH, GSM8K, and CSQA, MALT surpasses the same baseline LLM with a relative improvement of 15.66%, 7.42%, and 9.40% respectively.
  arXiv  Detail & Related papers  (2024-12-02T19:30:36Z)
• Language Models are Hidden Reasoners: Unlocking Latent Reasoning Capabilities via Self-Rewarding [74.31981011985681]
  Large language models (LLMs) have shown impressive capabilities, but still struggle with complex reasoning tasks requiring multiple steps. We introduce LaTent Reasoning Optimization (LaTRO), a principled framework that formulates reasoning as sampling from a latent distribution. We validate LaTRO through experiments on GSM8K and ARC-Challenge datasets using multiple model architectures.
  arXiv  Detail & Related papers  (2024-11-06T22:02:30Z)
• Mono-InternVL: Pushing the Boundaries of Monolithic Multimodal Large Language Models with Endogenous Visual Pre-training [48.455597568212944]
  We present Mono-InternVL, a novel monolithic MLLM that seamlessly integrates a set of visual experts via a multimodal mixture-of-experts structure.<n>In particular, EViP is designed as a progressive learning process for visual experts, which aims to fully exploit the visual knowledge from noisy data to high-quality data.
  arXiv  Detail & Related papers  (2024-10-10T17:59:22Z)
• Self-Play Fine-Tuning Converts Weak Language Models to Strong Language Models [52.98743860365194]
  We propose a new fine-tuning method called Self-Play fIne-tuNing (SPIN) At the heart of SPIN lies a self-play mechanism, where the LLM refines its capability by playing against instances of itself. This sheds light on the promise of self-play, enabling the achievement of human-level performance in LLMs without the need for expert opponents.
  arXiv  Detail & Related papers  (2024-01-02T18:53:13Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.