Walking the Tightrope: Disentangling Beneficial and Detrimental Drifts in Non-Stationary Custom-Tuning

This paper uncovers a critical yet overlooked phenomenon in multi-modal large language models (MLLMs): detrimental concept drift within chain-of-thought (CoT) reasoning during non-stationary reinforcement fine-tuning (RFT), where reasoning token distributions evolve unpredictably, thereby introducing significant biases in final predictions. To address this, we are pioneers in establishing the theoretical bridge between concept drift theory and RFT processes by formalizing CoT's autoregressive token streams as non-stationary distributions undergoing arbitrary temporal shifts. Leveraging this framework, we propose a novel counterfact-aware RFT that systematically decouples beneficial distribution adaptation from harmful concept drift through concept graph-empowered LLM experts generating counterfactual reasoning trajectories. Our solution, Counterfactual Preference Optimization (CPO), enables stable RFT in non-stationary environments, particularly within the medical domain, through custom-tuning of counterfactual-aware preference alignment. Extensive experiments demonstrate our superior performance of robustness, generalization and coordination within RFT. Besides, we also contributed a large-scale dataset CXR-CounterFact (CCF), comprising 320,416 meticulously curated counterfactual reasoning trajectories derived from MIMIC-CXR. Our code and data are public.

Learning Robust Spectral Dynamics for Temporal Domain Generalization

Modern machine learning models struggle to maintain performance in dynamic environments where temporal distribution shifts, \emph{i.e., concept drift}, are prevalent. Temporal Domain Generalization (TDG) seeks to enable model generalization across evolving domains, yet existing approaches typically assume smooth incremental changes, struggling with complex real-world drifts involving long-term structure (incremental evolution/periodicity) and local uncertainties. To overcome these limitations, we introduce FreKoo, which tackles these challenges via a novel frequency-domain analysis of parameter trajectories. It leverages the Fourier transform to disentangle parameter evolution into distinct spectral bands. Specifically, low-frequency component with dominant dynamics are learned and extrapolated using the Koopman operator, robustly capturing diverse drift patterns including both incremental and periodicity. Simultaneously, potentially disruptive high-frequency variations are smoothed via targeted temporal regularization, preventing overfitting to transient noise and domain uncertainties. In addition, this dual spectral strategy is rigorously grounded through theoretical analysis, providing stability guarantees for the Koopman prediction, a principled Bayesian justification for the high-frequency regularization, and culminating in a multiscale generalization bound connecting spectral dynamics to improved generalization. Extensive experiments demonstrate FreKoo's significant superiority over SOTA TDG approaches, particularly excelling in real-world streaming scenarios with complex drifts and uncertainties.

The Tenth NTIRE 2025 Image Denoising Challenge Report

This paper presents an overview of the NTIRE 2025 Image Denoising Challenge ({\sigma} = 50), highlighting the proposed methodologies and corresponding results. The primary objective is to develop a network architecture capable of achieving high-quality denoising performance, quantitatively evaluated using PSNR, without constraints on computational complexity or model size. The task assumes independent additive white Gaussian noise (AWGN) with a fixed noise level of 50. A total of 290 participants registered for the challenge, with 20 teams successfully submitting valid results, providing insights into the current state-of-the-art in image denoising.

Causal-Informed Contrastive Learning: Towards Bias-Resilient Pre-training under Concept Drift

The evolution of large-scale contrastive pre-training propelled by top-tier datasets has reached a transition point in the scaling law. Consequently, sustaining and enhancing a model's pre-training capabilities in drift environments have surfaced as a notable challenge. In this paper, we initially uncover that contrastive pre-training methods are significantly impacted by concept drift wherein distributions change unpredictably, resulting in notable biases in the feature space of the pre-trained model. Empowered by causal inference, we construct a structural causal graph to analyze the impact of concept drift to contrastive pre-training systemically, and propose the causal interventional contrastive objective. Upon achieving this, we devise a resilient contrastive pre-training approach to accommodate the data stream of concept drift, with simple and scalable implementation. Extensive experiments on various downstream tasks demonstrate our resilient contrastive pre-training effectively mitigates the bias stemming from the concept drift data stream. Codes are available at https://anonymous.4open.science/r/ResilientCL/.

Early Concept Drift Detection via Prediction Uncertainty

Concept drift, characterized by unpredictable changes in data distribution over time, poses significant challenges to machine learning models in streaming data scenarios. Although error rate-based concept drift detectors are widely used, they often fail to identify drift in the early stages when the data distribution changes but error rates remain constant. This paper introduces the Prediction Uncertainty Index (PU-index), derived from the prediction uncertainty of the classifier, as a superior alternative to the error rate for drift detection. Our theoretical analysis demonstrates that: (1) The PU-index can detect drift even when error rates remain stable. (2) Any change in the error rate will lead to a corresponding change in the PU-index. These properties make the PU-index a more sensitive and robust indicator for drift detection compared to existing methods. We also propose a PU-index-based Drift Detector (PUDD) that employs a novel Adaptive PU-index Bucketing algorithm for detecting drift. Empirical evaluations on both synthetic and real-world datasets demonstrate PUDD's efficacy in detecting drift in structured and image data.

Functional Stochastic Gradient MCMC for Bayesian Neural Networks

Classical variational inference for Bayesian neural networks (BNNs) in parameter space usually suffers from unresolved prior issues such as knowledge encoding intractability and pathological behaviors in deep networks, which could lead to an improper posterior inference. Hence, functional variational inference has been proposed recently to resolve these issues via stochastic process priors. Beyond variational inference, stochastic gradient Markov Chain Monte Carlo (SGMCMC) is another scalable and effective inference method for BNNs to asymptotically generate samples from true posterior by simulating a continuous dynamic. However, the existing SGMCMC methods only work in parametric space, which has the same issues of parameter-space variational inference, and extending the parameter-space dynamics to function-space dynamics is not a trivial undertaking. In this paper, we introduce a new functional SGMCMC scheme via newly designed diffusion dynamics, which can incorporate more informative functional priors. Moreover, we prove that the stationary distribution of these functional dynamics is the target posterior distribution over functions. We demonstrate better performance in both accuracy and uncertainty quantification of our functional SGMCMC on several tasks compared with naive SGMCMC and functional variational inference methods.

Sharpness-Aware Cross-Domain Recommendation to Cold-Start Users

Cross-Domain Recommendation (CDR) is a promising paradigm inspired by transfer learning to solve the cold-start problem in recommender systems. Existing state-of-the-art CDR methods train an explicit mapping function to transfer the cold-start users from a data-rich source domain to a target domain. However, a limitation of these methods is that the mapping function is trained on overlapping users across domains, while only a small number of overlapping users are available for training. By visualizing the loss landscape of the existing CDR model, we find that training on a small number of overlapping users causes the model to converge to sharp minima, leading to poor generalization. Based on this observation, we leverage loss-geometry-based machine learning approach and propose a novel CDR method called Sharpness-Aware CDR (SCDR). Our proposed method simultaneously optimizes recommendation loss and loss sharpness, leading to better generalization with theoretical guarantees. Empirical studies on real-world datasets demonstrate that SCDR significantly outperforms the other CDR models for cold-start recommendation tasks, while concurrently enhancing the model's robustness to adversarial attacks.

LSTM Autoencoder-based Deep Neural Networks for Barley Genotype-to-Phenotype Prediction

Artificial Intelligence (AI) has emerged as a key driver of precision agriculture, facilitating enhanced crop productivity, optimized resource use, farm sustainability, and informed decision-making. Also, the expansion of genome sequencing technology has greatly increased crop genomic resources, deepening our understanding of genetic variation and enhancing desirable crop traits to optimize performance in various environments. There is increasing interest in using machine learning (ML) and deep learning (DL) algorithms for genotype-to-phenotype prediction due to their excellence in capturing complex interactions within large, high-dimensional datasets. In this work, we propose a new LSTM autoencoder-based model for barley genotype-to-phenotype prediction, specifically for flowering time and grain yield estimation, which could potentially help optimize yields and management practices. Our model outperformed the other baseline methods, demonstrating its potential in handling complex high-dimensional agricultural datasets and enhancing crop phenotype prediction performance.

Methodology and Real-World Applications of Dynamic Uncertain Causality Graph for Clinical Diagnosis with Explainability and Invariance

AI-aided clinical diagnosis is desired in medical care. Existing deep learning models lack explainability and mainly focus on image analysis. The recently developed Dynamic Uncertain Causality Graph (DUCG) approach is causality-driven, explainable, and invariant across different application scenarios, without problems of data collection, labeling, fitting, privacy, bias, generalization, high cost and high energy consumption. Through close collaboration between clinical experts and DUCG technicians, 46 DUCG models covering 54 chief complaints were constructed. Over 1,000 diseases can be diagnosed without triage. Before being applied in real-world, the 46 DUCG models were retrospectively verified by third-party hospitals. The verified diagnostic precisions were no less than 95%, in which the diagnostic precision for every disease including uncommon ones was no less than 80%. After verifications, the 46 DUCG models were applied in the real-world in China. Over one million real diagnosis cases have been performed, with only 17 incorrect diagnoses identified. Due to DUCG's transparency, the mistakes causing the incorrect diagnoses were found and corrected. The diagnostic abilities of the clinicians who applied DUCG frequently were improved significantly. Following the introduction to the earlier presented DUCG methodology, the recommendation algorithm for potential medical checks is presented and the key idea of DUCG is extracted.

Analyzing Social Biases in Japanese Large Language Models

With the development of Large Language Models (LLMs), social biases in the LLMs have become a crucial issue. While various benchmarks for social biases have been provided across languages, the extent to which Japanese LLMs exhibit social biases has not been fully investigated. In this study, we construct the Japanese Bias Benchmark dataset for Question Answering (JBBQ) based on the English bias benchmark BBQ, and analyze social biases in Japanese LLMs. The results show that while current Japanese LLMs improve their accuracies on JBBQ by instruction-tuning, their bias scores become larger. In addition, augmenting their prompts with warning about social biases reduces the effect of biases in some models.