Towards Robust Endogenous Reasoning: Unifying Drift Adaptation in Non-Stationary Tuning

• URL: http://arxiv.org/abs/2604.15705v1
• Date: Fri, 17 Apr 2026 05:24:04 GMT
• Title: Towards Robust Endogenous Reasoning: Unifying Drift Adaptation in Non-Stationary Tuning
• Abstract summary: This paper proposes a comprehensive framework adapted to the multi-modal concept drift.<n>It integrates counterfactual reasoning with domain knowledge to execute controlled perturbations across thinking and perception.<n>It achieves superior performance in reasoning coherence, decision-making precision, and inherent robustness against extreme interference.
• Score: 20.73438231053258
• License: http://creativecommons.org/licenses/by-nc-nd/4.0/
• Abstract: Reinforcement Fine-Tuning (RFT) has established itself as a critical paradigm for the alignment of Multi-modal Large Language Models (MLLMs) with complex human values and domain-specific requirements. Nevertheless, current research primarily focuses on mitigating exogenous distribution shifts arising from data-centric factors, the non-stationarity inherent in the endogenous reasoning remains largely unexplored. In this work, a critical vulnerability is revealed within MLLMs: they are highly susceptible to endogenous reasoning drift, across both thinking and perception perspectives. It manifests as unpredictable distribution changes that emerge spontaneously during the autoregressive generation process, independent of external environmental perturbations. To adapt it, we first theoretically define endogenous reasoning drift within the RFT of MLLMs as the multi-modal concept drift. In this context, this paper proposes Counterfactual Preference Optimization ++ (CPO++), a comprehensive and autonomous framework adapted to the multi-modal concept drift. It integrates counterfactual reasoning with domain knowledge to execute controlled perturbations across thinking and perception, employing preference optimization to disentangle spurious correlations. Extensive empirical evaluations across two highly dynamic and safety-critical domains: medical diagnosis and autonomous driving. They demonstrate that the proposed framework achieves superior performance in reasoning coherence, decision-making precision, and inherent robustness against extreme interference. The methodology also exhibits exceptional zero-shot cross-domain generalization, providing a principled foundation for reliable multi-modal reasoning in safety-critical applications.

Related papers

• AnE: Pushing the Reasoning Frontier of Multimodal LLMs via Anchor Evolution [61.593935260052795]
  Post-training viaSupervised Fine-Tuning (SFT) and Reinforcement Learning (RL) is crucial for enhancing reasoning in Multimodal Large Language Models (MLLMs)<n>Existing paradigms often reach a performance bottleneck due to the limitations of static data.<n>We propose Anchor Evolution (AnE), a new paradigm that integrates truth-anchored data curation and model evolution.
  arXiv  Detail & Related papers  (2026-05-25T08:26:34Z)
• Breakthrough the Suboptimal Stable Point in Value-Factorization-Based Multi-Agent Reinforcement Learning [45.92643622013714]
  We introduce the stable point, which characterizes the potential convergence of value factorization in general cases.<n>We show that non-optimal stable points are the primary cause of poor performance.<n>We propose a novel Multi-Round Value Factorization (MRVF) framework.
  arXiv  Detail & Related papers  (2026-04-07T01:05:13Z)
• Hit-RAG: Learning to Reason with Long Contexts via Preference Alignment [97.55382322103986]
  Hit-RAG is a multi-stage preference alignment framework designed to resolve cognitive bottlenecks.<n>Our approach systematically refines the utilization of external evidence via three distinct stages.
  arXiv  Detail & Related papers  (2026-03-07T04:05:27Z)
• Multimodal Fact-Level Attribution for Verifiable Reasoning [80.60864342985748]
  Multimodal large language models (MLLMs) are increasingly used for real-world tasks involving multi-step reasoning and long-form generation.<n>Existing multimodal grounding benchmarks and evaluation methods fail to assess attribution in complex multimodal reasoning.<n>We introduce MuRGAt, a benchmark for evaluating fact-level multimodal attribution in settings that require reasoning beyond direct observation.
  arXiv  Detail & Related papers  (2026-02-12T03:10:02Z)
• On the Plasticity and Stability for Post-Training Large Language Models [54.757672540381236]
  We identify a root cause as the conflict between plasticity and stability gradients.<n>We propose Probabilistic Conflict Resolution (PCR), a framework that models gradients as random variables.<n>PCR significantly smooths the training trajectory and achieves superior performance in various reasoning tasks.
  arXiv  Detail & Related papers  (2026-02-06T07:31:26Z)
• EvoQRE: Modeling Bounded Rationality in Safety-Critical Traffic Simulation via Evolutionary Quantal Response Equilibrium [0.0]
  EvoQRE is a principled framework for modeling safety-critical traffic interactions.<n>It integrates a pre-trained generative world model with entropy-regularized replicator dynamics.<n>It achieves state-of-the-art realism, improved safety metrics, and controllable generation of diverse safety-critical scenarios.
  arXiv  Detail & Related papers  (2026-01-09T09:21:36Z)
• The Reasoning-Creativity Trade-off: Toward Creativity-Driven Problem Solving [57.652356955571065]
  State-of-the-art large language model (LLM) pipelines rely on bootstrapped reasoning loops.<n>We analyze how this design choice is sensitive to the collapse of the model's distribution over reasoning paths.<n>We introduce Distributional Creative Reasoning (DCR), a unified variational objective that casts training as gradient flow through probability measures on solution traces.
  arXiv  Detail & Related papers  (2026-01-02T17:10:31Z)
• The Silent Scholar Problem: A Probabilistic Framework for Breaking Epistemic Asymmetry in LLM Agents [0.6117371161379209]
  We propose a formal probabilistic framework that provides agents with a non-altruistic motive for bidirectional knowledge exchange.<n>We show how these accumulated belief states serve as verifiable reward signals for Reinforcement Learning from Human Feedback (RLHF) and high-quality data filters for Supervised Fine-Tuning (SFT)<n> Simulation results validate that this uncertainty-driven strategy significantly outperforms random baselines in heterogeneous environments.
  arXiv  Detail & Related papers  (2025-12-24T02:02:25Z)
• Drift No More? Context Equilibria in Multi-Turn LLM Interactions [58.69551510148673]
  contexts drift is the gradual divergence of a model's outputs from goal-consistent behavior across turns.<n>Unlike single-turn errors, drift unfolds temporally and is poorly captured by static evaluation metrics.<n>We show that multi-turn drift can be understood as a controllable equilibrium phenomenon rather than as inevitable decay.
  arXiv  Detail & Related papers  (2025-10-09T04:48:49Z)
• Coherent Multimodal Reasoning with Iterative Self-Evaluation for Vision-Language Models [4.064135211977999]
  Large language models (LLMs) and vision-language models (LVLMs) struggle with complex, multi-step, cross-modal common sense reasoning tasks.<n>We propose the Coherent Multimodal Reasoning Framework (CMRF), a novel approach that enhances LVLMs' common sense reasoning capabilities.<n>CMRF mimics human problem-solving by decomposing complex queries, generating step-by-step inferences, and self-correcting errors.
  arXiv  Detail & Related papers  (2025-08-04T20:33:58Z)
• Walking the Tightrope: Disentangling Beneficial and Detrimental Drifts in Non-Stationary Custom-Tuning [16.97188816362991]
  This paper uncovers a critical yet overlooked phenomenon in multi-modal large language models (MLLMs)<n>We are pioneers in establishing the theoretical bridge between concept drift theory and RFT processes.<n>We propose a novel counterfact-aware RFT that systematically decouples beneficial distribution adaptation from harmful concept drift.
  arXiv  Detail & Related papers  (2025-05-19T13:13:38Z)
• Trustworthy Multimodal Regression with Mixture of Normal-inverse Gamma Distributions [91.63716984911278]
  We introduce a novel Mixture of Normal-Inverse Gamma distributions (MoNIG) algorithm, which efficiently estimates uncertainty in principle for adaptive integration of different modalities and produces a trustworthy regression result. Experimental results on both synthetic and different real-world data demonstrate the effectiveness and trustworthiness of our method on various multimodal regression tasks.
  arXiv  Detail & Related papers  (2021-11-11T14:28:12Z)

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.