Machine Learning · Cross-listed from cs.SE
SWE-Tester: Training Open-Source LLMs for Issue Reproduction in Real-World Repositories
Aditya Bharat Soni, Rajat Ghosh, Vaishnavi Bhargava et al. (+2 more)
Software testing is crucial for ensuring the correctness and reliability of software systems. Automated generation of issue reproduction tests from natural
language issue descriptions enhances developer productivity by simplifying root cause analysis, promotes test-driven development – "test first, write code later", and can be used for improving the effectiveness of automated issue resolution systems like coding
agents. Existing methods proposed for this task predominantly rely on closed-source LLMs, with limited exploration of open models. To address this, we propose SWE-Tester – a novel pipeline for training open-source LLMs to generate issue reproduction tests. First, we curate a high-quality training
dataset of 41K instances from 2.6K open-source GitHub repositories and use it to train LLMs of varying sizes and families. The fine-tuned models achieve absolute improvements of up to 10% in success rate and 21% in change coverage on SWT-Bench Verified. Further analysis shows consistent improvements with increased inference-time compute, more data, and larger models. These results highlight the effectiveness of our framework for advancing open-source LLMs in this domain.
Read the full abstract ▾
All authors Aditya Bharat Soni, Rajat Ghosh, Vaishnavi Bhargava, Valerie Chen, Debojyoti Dutta
Machine Learning
Who Benefits From Sinus Surgery? Comparing Generative AI and Supervised Machine Learning for Predicting Surgical Outcomes in Chronic Rhinosinusitis
Sayeed Shafayet Chowdhury, Snehasis Mukhopadhyay, Shiaofen Fang et al. (+1 more)
Artificial intelligence has reshaped medical imaging, yet the use of AI on clinical data for prospective decision support remains limited. We study pre-operative prediction of clinically meaningful improvement in chronic rhinosinusitis (CRS), defining success as a more than 8.9-point reduction in SNOT-22 at 6 months (MCID). In a prospectively collected cohort where all patients underwent surgery, we ask whether models using only pre-operative clinical data could have identified those who would have poor outcomes, i.e. those who should have avoided surgery. We
benchmark supervised ML (logistic regression, tree ensembles, and an in-house MLP) against generative AI (ChatGPT, Claude, Gemini, Perplexity), giving each the same structured inputs and constraining outputs to binary recommendations with confidence. Our best ML model (MLP) achieves 85 % accuracy with superior calibration and decision-curve net benefit. GenAI models underperform on discrimination and calibration across zero-shot setting. Notably, GenAI justifications align with clinician heuristics and the MLP's feature importance, repeatedly highlighting baseline SNOT-22, CT/endoscopy severity, polyp phenotype, and physchology/pain comorbidities. We provide a reproducible tabular-to-GenAI evaluation protocol and subgroup analyses. Findings support an ML-first, GenAI- augmented workflow: deploy calibrated ML for primary triage of
surgical candidacy, with GenAI as an explainer to enhance transparency and shared decision-making.
Read the full abstract ▾
All authors Sayeed Shafayet Chowdhury, Snehasis Mukhopadhyay, Shiaofen Fang, Vijay R. Ramakrishnan
Machine Learning · Cross-listed from quant-ph
Generative Adversarial Networks for Resource State Generation
Shahbaz Shaik, Sourav Chatterjee, Sayantan Pramanik et al. (+1 more)
We introduce a physics-informed Generative
Adversarial Network framework that recasts
quantum resource-state generation as an inverse-design task. By
embedding task-specific utility functions into training, the model learns to generate valid two-
qubit states optimized for teleportation and
entanglement broadcasting. Comparing decomposition-based and direct-generation architectures reveals that structural enforcement of Hermiticity, trace-one, and positivity yields higher fidelity and training stability than loss-only approaches. The framework reproduces theoretical resource boundaries for Werner-like and Bell-diagonal states with fidelities exceeding ~98%, establishing
adversarial learning as a lightweight yet effective method for constraint-driven
quantum-state discovery. This approach provides a scalable foundation for automated design of tailored
quantum resources for information-processing applications, exemplified with teleportation and broadcasting of
entanglement, and it opens up the possibility of using such states in efficient
quantum network design.
Read the full abstract ▾
All authors Shahbaz Shaik, Sourav Chatterjee, Sayantan Pramanik, Indranil Chakrabarty
Machine Learning · Cross-listed from cs.CV
Attention-space Contrastive Guidance for Efficient Hallucination Mitigation in LVLMs
Yujin Jo, Sangyoon Bae and Taesup Kim
Hallucinations in large
vision–
language models (LVLMs) often arise when
language priors dominate over visual evidence, leading to object misidentification and visually inconsistent descriptions. We address this problem by framing
hallucination mitigation as contrastive guidance that steers generation toward visually grounded and semantically faithful text. We propose
Attention-space Contrastive Guidance (ACG), a training-free, single-pass method that operates directly in self-
attention layers, where
hallucination-inducing cross-modal biases emerge. ACG constructs both image-conditioned and approximate text-only
attention paths within a single forward pass, enabling efficient guidance before errors accumulate at the output layer. Because this masking-based surrogate can introduce approximation bias, we further apply a lightweight orthogonal projection that suppresses components aligned with the text-only path, yielding a more visually grounded correction. Experiments on CHAIR and POPE show that ACG improves faithfulness over existing training-free baselines while maintaining caption quality, reducing latency by up to $2\times$ compared to multi-pass contrastive decoding methods.
Read the full abstract ▾
Machine Learning · Cross-listed from cs.SD
Performance and Complexity Trade-off Optimization of Speech Models During Training
Esteban Gómez and Tom Bäckström
In
speech machine learning, neural network models are typically designed by choosing an architecture with fixed layer sizes and structure. These models are then trained to maximize performance on metrics aligned with the task's objective. While the overall architecture is usually guided by prior knowledge of the task, the sizes of individual layers are often chosen heuristically. However, this approach does not guarantee an optimal trade-off between performance and computational complexity; consequently, post hoc methods such as weight quantization or model
pruning are typically employed to reduce computational cost. This occurs because
stochastic gradient descent (SGD) methods can only optimize differentiable functions, while factors influencing computational complexity, such as layer sizes and floating-point operations per second (FLOP/s), are non-differentiable and require modifying the model structure during training. We propose a reparameterization technique based on feature noise injection that enables joint optimization of performance and computational complexity during training using SGD-based methods. Unlike traditional
pruning methods, our approach allows the model size to be dynamically optimized for a target performance-complexity trade-off, without relying on heuristic criteria to select which weights or structures to remove. We demonstrate the effectiveness of our method through three case studies, including a synthetic example and two practical real-world applications: voice activity detection and audio anti-spoofing. The code related to our work is publicly available to encourage further research.
Read the full abstract ▾
Machine Learning
Does Privacy Always Harm Fairness? Data-Dependent Trade-offs via Chernoff Information Neural Estimation
Arjun Nichani, Hsiang Hsu, Chun-Fu Chen et al. (+1 more)
Fairness and
privacy are two vital pillars of trustworthy machine learning. Despite extensive research on these individual topics, their relationship has received significantly less
attention. In this paper, we utilize an information-theoretic measure Chernoff Information to characterize the fundamental trade-off between
fairness,
privacy, and accuracy, as induced by the input data distribution. We first propose Chernoff Difference, a notion of data
fairness, along with its noisy variant, Noisy Chernoff Difference, which allows us to analyze both
fairness and
privacy simultaneously. Through simple Gaussian examples, we show that Noisy Chernoff Difference exhibits three qualitatively distinct behaviors depending on the underlying data distribution. To extend this analysis beyond synthetic settings, we develop the Chernoff Information Neural
Estimator (CINE), the first neural network-based
estimator of Chernoff Information for unknown distributions. We apply CINE to analyze the Noisy Chernoff Difference on real-world
datasets. Together, this work fills a critical gap in the literature by providing a principled, data-dependent characterization of the
fairness-
privacy interaction.
Read the full abstract ▾
All authors Arjun Nichani, Hsiang Hsu, Chun-Fu Chen, Haewon Jeong
Machine Learning · Cross-listed from cs.CL
Uncertainty-Aware Gradient Signal-to-Noise Data Selection for Instruction Tuning
Zhihang Yuan, Chengyu Yue, Long Huang et al. (+2 more)
Instruction tuning is a standard paradigm for adapting large
language models (LLMs), but modern instruction
datasets are large, noisy, and redundant, making full-data fine-tuning costly and often unnecessary. Existing data selection methods either build expensive gradient datastores or assign static scores from a weak proxy, largely ignoring evolving uncertainty, and thus missing a key source of LLM interpretability. We propose GRADFILTERING, an objective-agnostic, uncertainty-aware data selection framework that utilizes a small GPT-2 proxy with a LoRA ensemble and aggregates per-example gradients into a Gradient Signal-to-Noise Ratio (G-SNR) utility. Our method matches or surpasses random subsets and strong baselines in most LLM-as-a-judge evaluations as well as in human assessment. Moreover, GRADFILTERING-selected subsets converge faster than competitive filters under the same compute budget, reflecting the benefit of uncertainty-aware scoring.
Read the full abstract ▾
All authors Zhihang Yuan, Chengyu Yue, Long Huang, Litu Ou, Lei Shi
Machine Learning
Autoregressive deep learning for real-time simulation of soft tissue dynamics during virtual neurosurgery
Fabian Greifeneder, Wolfgang Fenz, Benedikt Alkin et al. (+3 more)
Accurate simulation of
brain deformation is a key component for developing realistic, interactive neurosurgical simulators, as complex nonlinear deformations must be captured to ensure realistic tool-tissue interactions. However, traditional numerical solvers often fall short in meeting real-time performance requirements. To overcome this, we introduce a deep learning-based surrogate model that efficiently simulates transient
brain deformation caused by continuous interactions between
surgical instruments and the virtual
brain geometry. Building on Universal Physics
Transformers, our approach operates directly on large-scale mesh data and is trained on an extensive
dataset generated from nonlinear finite element simulations, covering a broad spectrum of temporal instrument-tissue interaction scenarios. To reduce the accumulation of errors in autoregressive inference, we propose a
stochastic teacher forcing strategy applied during model training. Specifically, training consists of short
stochastic rollouts in which the proportion of ground truth inputs is gradually decreased in favor of model-generated predictions. Our results show that the proposed surrogate model achieves accurate and efficient predictions across a range of transient
brain deformation scenarios, scaling to meshes with up to 150,000 nodes. The introduced
stochastic teacher forcing technique substantially improves long-term rollout stability, reducing the maximum prediction error from 6.7 mm to 3.5 mm. We further integrate the trained surrogate model into an interactive neurosurgical simulation environment, achieving runtimes below 10 ms per simulation step on consumer-grade inference hardware. Our proposed deep learning framework enables rapid, smooth and accurate biomechanical simulations of dynamic
brain tissue deformation, laying the foundation for realistic
surgical training environments.
Read the full abstract ▾
All authors Fabian Greifeneder, Wolfgang Fenz, Benedikt Alkin, Johannes Brandstetter, Michael Giretzlehner, Philipp Moser
Machine Learning · Cross-listed from cs.NI
Reinforcement Learning for Opportunistic Routing in Software-Defined LEO-Terrestrial Systems
Sivaram Krishnan, Zhouyou Gu, Jihong Park et al. (+2 more)
The proliferation of large-scale low Earth orbit (LEO) satellite constellations is driving the need for intelligent routing strategies that can effectively deliver data to terrestrial networks under rapidly time-varying
topologies and intermittent gateway visibility. Leveraging the global control capabilities of a geostationary (GEO)-resident software-defined networking (SDN) controller, we introduce opportunistic routing, which aims to minimize delivery delay by forwarding packets to any currently available ground gateways rather than fixed destinations. This makes it a promising approach for achieving low-latency and robust data delivery in highly dynamic LEO networks. Specifically, we formulate a constrained
stochastic optimization problem and employ a residual reinforcement learning framework to optimize opportunistic routing for reducing transmission delay. Simulation results over multiple days of orbital data demonstrate that our method achieves significant improvements in queue length reduction compared to classical backpressure and other well-known queueing
algorithms.
Read the full abstract ▾
All authors Sivaram Krishnan, Zhouyou Gu, Jihong Park, Sung-Min Oh, Jinho Choi
Machine Learning · Cross-listed from cs.RO
Communication-Free Collective Navigation for a Swarm of UAVs via LiDAR-Based Deep Reinforcement Learning
Myong-Yol Choi, Hankyoul Ko, Hanse Cho et al. (+4 more)
This paper presents a deep reinforcement learning (DRL) based controller for collective
navigation of unmanned aerial vehicle (
UAV) swarms in communication-denied environments, enabling robust operation in complex, obstacle-rich environments. Inspired by biological swarms where informed individuals guide groups without explicit communication, we employ an implicit leader-follower framework. In this paradigm, only the leader possesses goal information, while follower
UAVs learn robust policies using only onboard
LiDAR sensing, without requiring any inter-
agent communication or leader identification. Our system utilizes
LiDAR point clustering and an extended Kalman filter for stable neighbor tracking, providing reliable perception independent of external positioning systems. The core of our approach is a DRL controller, trained in GPU-accelerated Nvidia Isaac Sim, that enables followers to learn complex emergent behaviors - balancing flocking and obstacle avoidance - using only local perception. This allows the swarm to implicitly follow the leader while robustly addressing perceptual challenges such as occlusion and limited field-of-view. The robustness and sim-to-real transfer of our approach are confirmed through extensive simulations and challenging real-world experiments with a swarm of five
UAVs, which successfully demonstrated collective
navigation across diverse indoor and outdoor environments without any communication or external localization.
Read the full abstract ▾
All authors Myong-Yol Choi, Hankyoul Ko, Hanse Cho, Changseung Kim, Seunghwan Kim, Jaemin Seo, Hyondong Oh
Machine Learning
TimeART: Towards Agentic Time Series Reasoning via Tool-Augmentation
Xingjian Wu, Junkai Lu, Zhengyu Li et al. (+5 more)
Time series data widely exist in real-world cyber-physical systems. Though analyzing and interpreting them contributes to significant values, e.g, disaster prediction and financial risk control, current workflows mainly rely on human data scientists, which requires significant labor costs and lacks automation. To tackle this, we introduce TimeART, a framework fusing the analytical capability of strong out-of-the-box tools and the
reasoning capability of Large
Language Models (LLMs), which serves as a fully agentic data scientist for Time Series Question Answering (TSQA). To teach the LLM-based Time Series
Reasoning Models (TSRMs) strategic tool-use, we also collect a 100k expert trajectory corpus called TimeToolBench. To enhance TSRMs' generalization capability, we then devise a four-stage training strategy, which boosts TSRMs through learning from their own early experiences and self-reflections. Experimentally, we train an 8B TSRM on TimeToolBench and equip it with the TimeART framework, and it achieves consistent state-of-the-art performance on multiple TSQA tasks, which pioneers a novel approach towards agentic time series
reasoning.
Read the full abstract ▾
All authors Xingjian Wu, Junkai Lu, Zhengyu Li, Xiangfei Qiu, Jilin Hu, Chenjuan Guo, Christian S. Jensen, Bin Yang
Machine Learning
Quadratic Upper Bound for Boosting Robustness
Euijin You and Hyang-Won Lee
Fast
adversarial training (FAT) aims to enhance the robustness of models against
adversarial attacks with reduced training time, however, FAT often suffers from compromised robustness due to insufficient exploration of
adversarial space. In this paper, we develop a loss function to mitigate the problem of degraded robustness under FAT. Specifically, we derive a quadratic upper bound (QUB) on the
adversarial training (AT) loss function and propose to utilize the bound with existing FAT methods. Our experimental results show that applying QUB loss to the existing methods yields significant improvement of robustness. Furthermore, using various metrics, we demonstrate that this improvement is likely to result from the smoothened loss landscape of the resulting model.
Read the full abstract ▾
Machine Learning · Cross-listed from cs.CL
Towards Token-Level Text Anomaly Detection
Yang Cao, Bicheng Yu, Sikun Yang et al. (+2 more)
Despite significant progress in text anomaly detection for web applications such as spam filtering and fake news detection, existing methods are fundamentally limited to document-level analysis, unable to identify which specific parts of a text are anomalous. We introduce
token-level anomaly detection, a novel paradigm that enables fine-grained localization of anomalies within text. We formally define text anomalies at both document and
token-levels, and propose a unified detection framework that operates across multiple levels. To facilitate research in this direction, we collect and annotate three
benchmark datasets spanning spam, reviews and grammar errors with
token-level labels. Experimental results demonstrate that our framework get better performance than other 6 baselines, opening new possibilities for precise anomaly localization in text. All the codes and data are publicly available on https://github.com/charles-cao/TokenCore.
Read the full abstract ▾
All authors Yang Cao, Bicheng Yu, Sikun Yang, Ming Liu, Yujiu Yang
Machine Learning · Cross-listed from stat.ML
Sample Complexity of Average-Reward Q-Learning: From Single-agent to Federated Reinforcement Learning
Yuchen Jiao, Jiin Woo, Gen Li et al. (+2 more)
Average-reward reinforcement learning offers a principled framework for long-term decision-making by maximizing the mean reward per time step. Although Q-learning is a widely used model-free
algorithm with established sample complexity in discounted and finite-horizon Markov decision processes (MDPs), its theoretical guarantees for average-reward settings remain limited. This work studies a simple but effective Q-learning
algorithm for average-reward MDPs with finite state and action spaces under the weakly communicating assumption, covering both single-
agent and
federated scenarios. For the single-
agent case, we show that Q-learning with carefully chosen parameters achieves sample complexity $\widetilde{O}\left(\frac{|\mathcal{S}||\mathcal{A}|\|h^{\star}\|_{\mathsf{sp}}^3}{\varepsilon^3}\right)$, where $\|h^{\star}\|_{\mathsf{sp}}$ is the span norm of the bias function, improving previous results by at least a factor of $\frac{\|h^{\star}\|_{\mathsf{sp}}^2}{\varepsilon^2}$. In the
federated setting with $M$
agents, we prove that collaboration reduces the per-
agent sample complexity to $\widetilde{O}\left(\frac{|\mathcal{S}||\mathcal{A}|\|h^{\star}\|_{\mathsf{sp}}^3}{M\varepsilon^3}\right)$, with only $\widetilde{O}\left(\frac{\|h^{\star}\|_{\mathsf{sp}}}{\varepsilon}\right)$ communication rounds required. These results establish the first
federated Q-learning
algorithm for average-reward MDPs, with provable efficiency in both sample and communication complexity.
Read the full abstract ▾
All authors Yuchen Jiao, Jiin Woo, Gen Li, Gauri Joshi, Yuejie Chi
Machine Learning · Cross-listed from cs.CL
CauScientist: Teaching LLMs to Respect Data for Causal Discovery
Bo Peng, Sirui Chen, Lei Xu et al. (+1 more)
Causal discovery is fundamental to scientific understanding and reliable decision-making. Existing approaches face critical limitations: purely data-driven methods suffer from statistical indistinguishability and modeling assumptions, while recent LLM-based methods either ignore statistical evidence or incorporate unverified priors that can mislead result. To this end, we propose CauScientist, a collaborative framework that synergizes LLMs as hypothesis-generating "data scientists" with probabilistic statistics as rigorous "verifiers". CauScientist employs hybrid initialization to select superior starting graphs, iteratively refines structures through LLM-proposed modifications validated by statistical criteria, and maintains error
memory to guide efficient search space. Experiments demonstrate that CauScientist substantially outperforms purely data-driven baselines, achieving up to 53.8% F1 score improvement and enhancing recall from 35.0% to 100.0%. Notably, while standalone LLM performance degrades with graph complexity, CauScientist reduces structural hamming distance (SHD) by 44.0% compared to Qwen3-32B on 37-node graphs. Our project page is at https://github.com/OpenCausaLab/CauScientist.
Read the full abstract ▾
All authors Bo Peng, Sirui Chen, Lei Xu, Chaochao Lu
Machine Learning
Fisher-Informed Parameterwise Aggregation for Federated Learning with Heterogeneous Data
Zhipeng Chang, Ting He and Wenrui Hao
Federated learning aggregates model updates from distributed clients, but standard first order methods such as FedAvg apply the same scalar weight to all parameters from each client. Under non-IID data, these uniformly weighted updates can be strongly misaligned across clients, causing client drift and degrading the global model. Here we propose Fisher-Informed Parameterwise Aggregation (FIPA), a second-order aggregation method that replaces client-level scalar weights with parameter-specific Fisher Information
Matrix (FIM) weights, enabling true parameter-level scaling that captures how each client's data uniquely influences different parameters. With low-rank approximation, FIPA remains communication- and computation-efficient. Across nonlinear function regression, PDE learning, and image classification, FIPA consistently improves over averaging-based aggregation, and can be effectively combined with state-of-the-art client-side optimization
algorithms to further improve image classification accuracy. These results highlight the benefits of FIPA for
federated learning under heterogeneous data distributions.
Read the full abstract ▾
Machine Learning · Cross-listed from math.NA
Optimizing Parallel Schemes with Lyapunov Exponents and kNN-LLE Estimation
Mudassir Shams, Andrei Velichko and Bruno Carpentieri
Inverse parallel schemes remain indispensable tools for computing the roots of nonlinear systems, yet their dynamical behavior can be unexpectedly rich, ranging from strong contraction to oscillatory or chaotic transients depending on the choice of algorithmic parameters and initial states. A unified analytical-data-driven methodology for identifying, measuring, and reducing such instabilities in a family of uni-parametric inverse parallel solvers is presented in this study. On the theoretical side, we derive stability and bifurcation characterizations of the underlying iterative maps, identifying parameter regions associated with periodic or chaotic behavior. On the computational side, we introduce a micro-series pipeline based on kNN-driven estimation of the local largest Lyapunov exponent (LLE), applied to scalar time series derived from solver trajectories. The resulting sliding-window Lyapunov profiles provide fine-grained, real-time diagnostics of contractive or unstable phases and reveal transient behaviors not captured by coarse linearized analysis. Leveraging this correspondence, we introduce a Lyapunov-informed parameter selection strategy that identifies solver settings associated with stable behavior, particularly when the estimated LLE indicates persistent instability. Comprehensive experiments on ensembles of perturbed initial guesses demonstrate close agreement between the theoretical stability diagrams and empirical Lyapunov profiles, and show that the proposed adaptive mechanism significantly improves robustness. The study establishes micro-series Lyapunov analysis as a practical, interpretable tool for constructing self-stabilizing root-finding schemes and opens avenues for extending such diagnostics to higher-dimensional or noise-contaminated problems.
Read the full abstract ▾
Machine Learning · Cross-listed from cs.IT
A Gaussian Perspective for Distributional Discrepancy in Generative Diffusion Models
Qiang Sun, H. Vincent Poor and Wenyi Zhang
This paper introduces an analytical approach to quantifying and optimizing the distributional discrepancy in generative
diffusion models. For a multivariate Gaussian source, we explicitly derive the closed-form
evolution trajectory and the resulting Kullback-Leibler (KL) divergence between the distributions of the source data and the reversely sampled data. Asymptotic analysis via the Euler-Maclaurin expansion characterizes the convergence behavior of this KL divergence, extracting its dominant term as an explicit functional of the noise schedule. Minimizing this dominant term via the calculus of variations yields a noise schedule described by a tangent law, inherently determined by the source covariance spectrum. We further prove that the Gaussian source exhibits an extremal property for the KL divergence among general source distributions with a given covariance. We also utilize the analytical KL divergence as a principled metric to identify efficient time discretization strategies for pretrained
diffusion models, and demonstrate via experiments over diverse
datasets that the identified strategies consistently outperform established baselines, particularly under constrained function evaluation budgets.
Read the full abstract ▾
Machine Learning
Diffusion In Diffusion: Reclaiming Global Coherence in Semi-Autoregressive Diffusion
Linrui Ma, Yufei Cui, Kai Han et al. (+1 more)
One of the most compelling features of global discrete
diffusion language models is their global bidirectional contextual capability. However, existing block-based
diffusion studies tend to introduce autoregressive priors, which, while offering benefits, can cause models to lose this global coherence at the macro level. To regain global contextual understanding while preserving the advantages of the semi-autoregressive paradigm, we propose
Diffusion in
Diffusion, a 'draft-then-refine' framework designed to overcome the irreversibility and myopia problems inherent in block
diffusion models. Our approach first employs block
diffusion to generate rapid drafts using small blocks, then refines these drafts through global bidirectional
diffusion with a larger bidirectional receptive field. We utilize snapshot confidence remasking to identify the most critical
tokens that require modification, and apply mix-scale training to expand the block
diffusion model's global capabilities. Empirical results demonstrate that our approach sets a new
benchmark for discrete
diffusion models on the OpenWebText
dataset. Using only 26% of the fine-tuning budget of baseline models, we reduce generative perplexity from 25.7 to 21.9, significantly narrowing the performance gap with autoregressive models.
Read the full abstract ▾
All authors Linrui Ma, Yufei Cui, Kai Han, Yunhe Wang
Machine Learning
Machine learning based radiative parameterization scheme and its performance in operational reforecast experiments
Hao Jing, Sa Xiao, Haoyu Li et al. (+2 more)
Radiation is typically the most time-consuming physical process in numerical models. One solution is to use machine learning methods to simulate the radiation process to improve computational efficiency. From an operational standpoint, this study investigates critical limitations inherent to hybrid
forecasting frameworks that embed deep neural networks into numerical prediction models, with a specific focus on two fundamental bottlenecks: coupling compatibility and long-term integration stability. A residual convolutional neural network is employed to approximate the Rapid Radiative Transfer Model for General Circulation Models (RRTMG) within the global operational system of China Meteorological Administration. We adopted an offline training and online coupling approach. First, a comprehensive
dataset is generated through model simulations, encompassing all atmospheric columns both with and without cloud cover. To ensure the stability of the hybrid model, the
dataset is enhanced via experience replay, and additional output constraints based on physical significance are imposed. Meanwhile, a LibTorch-based coupling method is utilized, which is more suitable for real-time operational computations. The hybrid model is capable of performing ten-day integrated forecasts as required. A two-month operational reforecast experiment demonstrates that the machine learning emulator achieves accuracy comparable to that of the traditional physical scheme, while accelerating the computation speed by approximately eightfold.
Read the full abstract ▾
All authors Hao Jing, Sa Xiao, Haoyu Li, Huadong Xiao, Wei Xue
Machine Learning
Neural Organ Transplantation (NOT): Checkpoint-Based Modular Adaptation for Transformer Models
Ahmad Al-Zuraiqi
We introduce Neural Organ Transplantation (NOT), a modular adaptation framework that enables trained
transformer layers to function as reusable transferable checkpoints for domain adaptation. Unlike conventional fine-tuning approaches that tightly couple trained parameters to specific model instances and training data, NOT extracts contiguous layer subsets ("donor organs") from pre-trained models, trains them independently on domain-specific data, and saves them as standalone checkpoint files that can be transplanted into compatible recipient models without access to the original training data. Through experiments on three decoder-only
transformer architectures spanning 124M to 20B parameters (GPT-2, TinyLlama, and GPT-OSS), we demonstrate that donor transplantation substantially outperforms existing adaptation methods, achieving an order-of-magnitude improvement in perplexity over LoRA while training significantly faster. The method exhibits position dependence, with early insertion positions yielding optimal results. Cross-domain transfer at billion-parameter scale reveals unexpected regularization benefits. These findings demonstrate that
transformer middle layers can support efficient modular transfer for decoder-only architectures, enabling
privacy-preserving expertise sharing through checkpoint distribution. We note that this approach is currently limited to decoder-only models; preliminary experiments on encoder-based architectures show reduced effectiveness.
Read the full abstract ▾
Machine Learning
FG-OrIU: Towards Better Forgetting via Feature-Gradient Orthogonality for Incremental Unlearning
Qian Feng, JiaHang Tu, Mintong Kang et al. (+3 more)
Incremental unlearning (IU) is critical for pre-trained models to comply with sequential data deletion requests, yet existing methods primarily suppress parameters or confuse knowledge without explicit constraints on both feature and gradient level, resulting in superficial forgetting where residual information remains recoverable. This incomplete forgetting risks security breaches and disrupts retention balance, especially in IU scenarios. We propose FG-OrIU (Feature-Gradient Orthogonality for Incremental Unlearning), the first framework unifying orthogonal constraints on both features and gradients level to achieve deep forgetting, where the forgetting effect is irreversible. FG-OrIU decomposes feature spaces via Singular Value Decomposition (SVD), separating forgetting and remaining class features into distinct subspaces. It then enforces dual constraints: feature orthogonal projection on both forgetting and remaining classes, while gradient orthogonal projection prevents the reintroduction of forgotten knowledge and disruption to remaining classes during updates. Additionally, dynamic subspace adaptation merges newly forgetting subspaces and contracts remaining subspaces, ensuring a stable balance between removal and retention across sequential unlearning tasks. Extensive experiments demonstrate the effectiveness of our method.
Read the full abstract ▾
All authors Qian Feng, JiaHang Tu, Mintong Kang, Hanbin Zhao, Chao Zhang, Hui Qian
Machine Learning
Behavior Knowledge Merge in Reinforced Agentic Models
Xiangchi Yuan, Dachuan Shi, Chunhui Zhang et al. (+4 more)
Reinforcement learning (RL) is central to post-training, particularly for agentic models that require specialized
reasoning behaviors. In this setting, model merging offers a practical mechanism for integrating multiple RL-trained
agents from different tasks into a single generalist model. However, existing merging methods are designed for supervised fine-tuning (SFT), and they are suboptimal to preserve task-specific capabilities on RL-trained agentic models. The root is a task-vector mismatch between RL and SFT: on-policy RL induces task vectors that are highly sparse and heterogeneous, whereas SFT-style merging implicitly assumes dense and globally comparable task vectors. When standard global averaging is applied under this mismatch, RL's non-overlapping task vectors that encode critical task-specific behaviors are reduced and parameter updates are diluted. To address this issue, we propose Reinforced
Agent Merging (RAM), a distribution-aware merging framework explicitly designed for RL-trained agentic models. RAM disentangles shared and task-specific unique parameter updates, averaging shared components while selectively preserving and rescaling unique ones to counteract parameter update dilution. Experiments across multiple
agent domains and model architectures demonstrate that RAM not only surpasses merging baselines, but also unlocks synergistic potential among
agents to achieve performance superior to that of specialized
agents in their domains.
Read the full abstract ▾
All authors Xiangchi Yuan, Dachuan Shi, Chunhui Zhang, Zheyuan Liu, Shenglong Yao, Soroush Vosoughi, Wenke Lee
Machine Learning
GeoDynamics: A Geometric State-Space Neural Network for Understanding Brain Dynamics on Riemannian Manifolds
Tingting Dan, Jiaqi Ding and Guorong Wu
State-space models (SSMs) have become a cornerstone for unraveling
brain dynamics, revealing how latent neural states evolve over time and give rise to observed signals. By combining the flexibility of deep learning with the principled dynamical structure of SSMs, recent studies have achieved powerful fits to functional neuroimaging data. However, most existing approaches still view the
brain as a set of loosely connected regions or impose oversimplified network priors, falling short of a truly holistic and self-organized dynamical system perspective.
Brain functional connectivity (FC) at each time point naturally forms a symmetric positive definite (SPD)
matrix, which resides on a curved Riemannian
manifold rather than in Euclidean space. Capturing the trajectories of these SPD
matrices is key to understanding how coordinated networks support cognition and behavior. To this end, we introduce GeoDynamics, a
geometric state-space neural network that tracks latent
brain-state trajectories directly on the high-dimensional SPD
manifold. GeoDynamics embeds each connectivity
matrix into a
manifold-aware recurrent framework, learning smooth and geometry-respecting transitions that reveal task-driven state changes and early markers of Alzheimer's disease, Parkinson's disease, and autism. Beyond neuroscience, we validate GeoDynamics on human action recognition
benchmarks (UTKinect, Florence, HDM05), demonstrating its scalability and robustness in modeling complex spatiotemporal dynamics across diverse domains.
Read the full abstract ▾
Machine Learning
DRGW: Learning Disentangled Representations for Robust Graph Watermarking
Jiasen Li, Yanwei Liu, Zhuoyi Shang et al. (+2 more)
Graph-structured data is foundational to numerous web applications, and watermarking is crucial for protecting their intellectual property and ensuring data provenance. Existing watermarking methods primarily operate on graph structures or entangled graph representations, which compromise the transparency and robustness of watermarks due to the information coupling in representing graphs and uncontrollable discretization in transforming continuous numerical representations into graph structures. This motivates us to propose DRGW, the first graph watermarking framework that addresses these issues through disentangled representation learning. Specifically, we design an adversarially trained encoder that learns an invariant structural representation against diverse perturbations and derives a statistically independent watermark carrier, ensuring both robustness and transparency of watermarks. Meanwhile, we devise a graph-aware invertible neural network to provide a lossless channel for watermark
embedding and extraction, guaranteeing high detectability and transparency of watermarks. Additionally, we develop a structure-aware editor that resolves the issue of latent modifications into discrete graph edits, ensuring robustness against structural perturbations. Experiments on diverse
benchmark datasets demonstrate the superior effectiveness of DRGW.
Read the full abstract ▾
All authors Jiasen Li, Yanwei Liu, Zhuoyi Shang, Xiaoyan Gu, Weiping Wang
Machine Learning
Self-Improvement as Coherence Optimization: A Theoretical Account
Tianyi Qiu, Ahmed Hani Ismail, Zhonghao He et al. (+1 more)
Can
language models improve their accuracy without external supervision? Methods such as debate, bootstrap, and internal coherence maximization achieve this surprising feat, even matching golden finetuning performance. Yet why they work remains theoretically unclear. We show that they can all be understood as coherence optimization, the search for a context-to-behavior mapping that is most compressible and jointly predictable, with debate an exact instance and bootstrap and internal coherence maximization closely related to it. We prove that coherence optimization is equivalent to description-length regularization, and that among all such regularization schemes, coherence regularization with a prior derived from a pretrained model optimizes a lower bound of worst-case accuracy for semi-supervised learning. Our theory, supported by preliminary experiments, explains why feedback-free self-improvement works and predicts when it should succeed or fail.
Read the full abstract ▾
All authors Tianyi Qiu, Ahmed Hani Ismail, Zhonghao He, Shi Feng
Machine Learning
Multi-objective fluorescent molecule design with a data-physics dual-driven generative framework
Yanheng Li, Zhichen Pu, Lijiang Yang et al. (+2 more)
Designing fluorescent small molecules with tailored optical and physicochemical properties requires navigating vast, underexplored chemical space while satisfying multiple objectives and constraints. Conventional generate-score-screen approaches become impractical under such realistic design specifications, owing to their low search efficiency, unreliable generalizability of machine-learning prediction, and the prohibitive cost of
quantum chemical calculation. Here we present LUMOS, a data-and-physics driven framework for inverse design of fluorescent molecules. LUMOS couples generator and predictor within a shared latent representation, enabling direct specification-to-molecule design and efficient exploration. Moreover, LUMOS combines neural networks with a fast time-dependent density functional theory (TD-DFT) calculation workflow to build a suite of complementary predictors spanning different trade-offs in speed, accuracy, and generalizability, enabling reliable property prediction across diverse scenarios. Finally, LUMOS employs a property-guided
diffusion model integrated with multi-objective evolutionary
algorithms, enabling de novo design and molecular optimization under multiple objectives and constraints. Across comprehensive
benchmarks, LUMOS consistently outperforms baseline models in terms of accuracy, generalizability and physical plausibility for fluorescence property prediction, and demonstrates superior performance in multi-objective scaffold- and fragment-level molecular optimization. Further validation using TD-DFT and molecular dynamics (MD) simulations demonstrates that LUMOS can generate valid fluorophores that meet various target specifications. Overall, these results establish LUMOS as a data-physics dual-driven framework for general fluorophore inverse design.
Read the full abstract ▾
All authors Yanheng Li, Zhichen Pu, Lijiang Yang, Zehao Zhou, Yi Qin Gao
Machine Learning
ButterflyMoE: Compression-Scalable Ternary Experts via Structured Butterfly Orbits
Aryan Karmore
In current Mixture of Experts (MoE) architectures, linear
memory scaling is present, the
memory grows as the number of experts increases. $N$ independent expert weight
matrices require $\mathcal{O}(N \cdot d^2)$
memory which exceeds the
memory budget of edge devices. Current compression methods like quantization,
pruning, and low-rank factorization reduce constant factors, but the scaling bottleneck is still unresolved. We introduce ButterflyMoE, a method which treats experts not as independent
matrices but as
geometric reorientations of a shared quantized substrate. Diversity amongst the experts arises from viewing different angles of the shared capacity and not from redundant storage. Learned rotations are applied to a shared ternary prototype. With this, each expert yields $\mathcal{O}(d^2 + N \cdot d \log d)$
memory-reducing per-expert cost from $\mathcal{O}(d^2)$ to $\mathcal{O}(d \log d)$. The key insight is that training these rotations with quantization reduces activation outliers and stabilizes extreme low-bit training where other static methods collapse. Across
language modeling
benchmarks, ButterflyMoE achieves 80$\times$
memory reduction at 8 experts with a highly favorable
memory-accuracy tradeoff.At this 80x compression ButterflyMoE outperforms an equal
memory dense baseline, showing that orbital parameterization extracts fundamentally more utility per byte. When scaled up to 256 experts, ButterflyMoE asymptotically compresses the
memory by 150 $\times$. ButterflyMoE reduces the constant factor of linear scaling with compression ratio growing with the expert count.
Read the full abstract ▾
Machine Learning · Cross-listed from cs.AI
Zhiguang Liu and Yi Shang
ARC tasks can be formulated as visual transformation problems, where a model must infer a high-level transformation rule from a small set of demonstrations and apply it consistently across a grid. While dense
Vision Transformers (ViTs) perform well in this setting, they entangle local visual state and global transformation logic within a shared
token representation, which can hinder consistent
reasoning. In this paper, we introduce RoSeViT, a role-separated
Vision Transformer that explicitly decouples high-level instruction inference from visual workspace computation. RoSeViT uses controller
tokens to aggregate transformation-level information and workspace
tokens to represent the visual grid, with a structured
attention pass in which nonlocal workspace interactions are routed through the controller. This design is motivated by our formal analysis showing that enforcing a shared instruction representation can reduce the effective complexity of visual transformation modeling. RoSeViT outperforms dense ViT baselines under the standard VARC training and evaluation protocol for ARC. In a controlled backbone replacement setting, it improves ARC-1 accuracy from 54.5 to 56.6, and with recurrent refinement reaches 58.8. When integrated into the full system, RoSeViT-Ensemble improves performance from 60.4 to 62.6 on ARC-1. These results demonstrate that explicitly separating instruction and workspace representations leads to more effective visual
reasoning on ARC tasks. GitHub repository: https://github.com/lz7fd/RoSeViT.
Read the full abstract ▾
Machine Learning
Patterning: The Dual of Interpretability
George Wang and Daniel Murfet
Mechanistic interpretability aims to understand how neural networks generalize beyond their training data by reverse-engineering their internal structures. We introduce patterning as the dual problem: given a desired form of generalization, determine what training data produces it. Our approach is based on susceptibilities, which measure how posterior expectation values of observables respond to infinitesimal shifts in the data distribution. Inverting this linear response relationship yields the data intervention that steers the model toward a target internal configuration. We demonstrate patterning in a small
language model, showing that re-weighting training data along principal susceptibility directions can accelerate or delay the formation of structure, such as the induction
circuit. In a synthetic parentheses balancing task where multiple
algorithms achieve perfect training accuracy, we show that patterning can select which
algorithm the model learns by targeting the local learning coefficient of each solution. These results establish that the same mathematical framework used to read internal structure can be inverted to write it.
Read the full abstract ▾