Machine Learning · Cross-listed from cs.AI
SUPERNOVA: Eliciting General Reasoning in LLMs with Reinforcement Learning on Natural Instructions
Ashima Suvarna, Kendrick Phan, Mehrab Beikzadeh et al. (+2 more)
Reinforcement Learning with Verifiable Rewards (RLVR) has substantially improved
reasoning in formal domains such as mathematics and code, but extending these gains beyond STEM remains challenging. Extending RLVR beyond STEM is fundamentally constrained by the lack of high-quality verifiable training data. In this work, we introduce
SUPERNOVA, a framework for curating RLVR data from natural instruction
datasets, which are a rich source of expert-annotated data but are underexplored for RLVR training. Through 100+ controlled RL experiments, we systematically study how to utilize these
dataset for RLVR and how data curation decisions affect downstream
reasoning performance . In particular, we investigate three data designs: (a) source task selection, (b) task mixing, and (c) synthetic interventions. Our analysis reveals that source task selection has a significant impact on downstream
reasoning performance. Moreover, selecting tasks based on their performance for individual target tasks outperforms strategies based on overall average performance and synthetic interventions do not improve
reasoning. Guided by these insights, we construct
SUPERNOVA, a high-quality RLVR
dataset of 25K instances curated from natural instruction
datasets. We show that training Qwen3-0.6B on
SUPERNOVA outperforms the base Qwen3-0.6B, yielding a relative gain of 64.4pp on BigBench Extra Hard (BBEH), a challenging
benchmark comprising 23 complex
reasoning tasks. Importantly, we find that gains from
SUPERNOVA generalize to unseen
benchmarks, larger model scales, and newer model families. Overall, our findings provide practical insights for curating human-annotated resources to extend RLVR to general
reasoning. Models, Data, Code at https://github.com/asuvarna31/
supernova.
Read the full abstract ▾
All authors Ashima Suvarna, Kendrick Phan, Mehrab Beikzadeh, Hritik Bansal, Saadia Gabriel
Machine Learning
Quantization Impact on the Accuracy and Communication Efficiency Trade-off in Federated Learning for Aerospace Predictive Maintenance
Abdelkarim Loukili
Federated learning (FL) enables
privacy-preserving predictive maintenance across distributed aerospace fleets, but gradient communication overhead constrains deployment on bandwidth-limited IoT nodes. This paper investigates the impact of symmetric uniform quantization ($b \in \{32,8,4,2\}$ bits) on the accuracy–efficiency trade-off of a custom-designed lightweight 1-D convolutional model (AeroConv1D, 9\,697 parameters) trained via FL on the NASA C-MAPSS
benchmark under a realistic Non-IID client partition. Using a rigorous multi-seed evaluation ($N=10$ seeds), we show that INT4 achieves accuracy statistically indistinguishable from FP32 on both FD001 ($p=0.341$) and FD002 ($p=0.264$ MAE, $p=0.534$ NASA score) while delivering an $8\times$ reduction in gradient communication cost (37.88~KiB $\to$ 4.73~KiB per round). A key methodological finding is that naïve IID client partitioning artificially suppresses variance; correct Non-IID evaluation reveals the true operational instability of extreme quantization, demonstrated via a direct empirical IID vs.\ Non-IID comparison. INT2 is empirically characterized as unsuitable: while it achieves lower MAE on FD002 through extreme quantization-induced over-regularization, this apparent gain is accompanied by catastrophic NASA score instability (CV\,=\,45.8% vs.\ 22.3% for FP32), confirming non-reproducibility under heterogeneous operating conditions. Analytical FPGA resource projections on the Xilinx ZCU102 confirm that INT4 fits within hardware constraints (85.5% DSP utilization), potentially enabling a complete FL pipeline on a single SoC. The full simulation codebase and FPGA estimation scripts are publicly available at https://github.com/therealdeadbeef/aerospace-fl-quantization.
Read the full abstract ▾
Machine Learning
Persistence-Augmented Neural Networks
Elena Xinyi Wang, Arnur Nigmetov and Dmitriy Morozov
Topological Data Analysis (TDA) provides tools to describe the shape of data, but integrating topological features into deep learning pipelines remains challenging, especially when preserving local
geometric structure rather than summarizing it globally. We propose a persistence-based data augmentation framework that encodes local gradient flow regions and their hierarchical
evolution using the Morse-Smale complex. This representation, compatible with both convolutional and graph neural networks, retains spatially localized topological information across multiple scales. Importantly, the augmentation procedure itself is efficient, with computational complexity $O(n \log n)$, making it practical for large
datasets. We evaluate our method on histopathology image classification and 3D porous material regression, where it consistently outperforms baselines and global TDA descriptors such as persistence images and landscapes. We also show that
pruning the base level of the hierarchy reduces
memory usage while maintaining competitive performance. These results highlight the potential of local, structured topological augmentation for scalable and interpretable learning across data modalities.
Read the full abstract ▾
Machine Learning
TTVS: Boosting Self-Exploring Reinforcement Learning via Test-time Variational Synthesis
Sikai Bai, Haoxi Li, Jie Zhang et al. (+2 more)
Despite significant advances in Large
Reasoning Models (LRMs) driven by reinforcement learning with verifiable rewards (RLVR), this paradigm is fundamentally limited in specialized or novel domains where such supervision is prohibitively expensive or unavailable, posing a key challenge for test-time adaptation. While existing test-time methods offer a potential solution, they are constrained by learning from static query sets, risking overfitting to textual patterns. To address this gap, we introduce Test-Time Variational Synthesis (TTVS), a novel framework that enables LRMs to self-evolve by dynamically augmenting the training stream from unlabeled test queries. TTVS comprises two synergistic modules: (1) Online Variational Synthesis, which transforms static test queries into a dynamic stream of diverse, semantically-equivalent variations, enforcing the model to learn underlying problem logic rather than superficial patterns; (2) Test-time Hybrid Exploration, which balances accuracy-driven exploitation with consistency-driven exploration across synthetic variants. Extensive experiments show TTVS yields superior performance across eight model architectures. Notably, using only unlabeled test-time data, TTVS not only surpasses other test-time adaptation methods but also outperforms state-of-the-art supervised RL-based techniques trained on vast, high-quality labeled data.
Read the full abstract ▾
All authors Sikai Bai, Haoxi Li, Jie Zhang, Yongjiang Liu, Song Guo
Machine Learning
A Machine Learning Framework for Turbofan Health Estimation via Inverse Problem Formulation
Milad Leyli-Abadi, Lucas Thil, Sebastien Razakarivony et al. (+2 more)
Estimating the health state of turbofan engines is a challenging ill-posed inverse problem, hindered by sparse sensing and complex nonlinear thermodynamics. Research in this area remains fragmented, with comparisons limited by the use of unrealistic
datasets and insufficient exploration of the exploitation of temporal information. This work investigates how to recover component-level health indicators from operational sensor data under realistic degradation and maintenance patterns. To support this study, we introduce a new
dataset that incorporates industry-oriented complexities such as maintenance events and usage changes. Using this
dataset, we establish an initial
benchmark that compares steady-state and nonstationary data-driven models, and
Bayesian filters, classic families of methods used to solve this problem. In addition to this
benchmark, we introduce self-supervised learning (SSL) approaches that learn latent representations without access to true health labels, a scenario reflective of real-world operational constraints. By comparing the downstream estimation performance of these unsupervised representations against the direct prediction baselines, we establish a practical lower bound on the difficulty of solving this inverse problem. Our results reveal that traditional filters remain strong baselines, while SSL methods reveal the intrinsic complexity of health estimation and highlight the need for more advanced and interpretable inference strategies. For reproducibility, both the generated
dataset and the implementation used in this work are made accessible.
Read the full abstract ▾
All authors Milad Leyli-Abadi, Lucas Thil, Sebastien Razakarivony, Guillaume Doquet, Jesse Read
Machine Learning
Less Data Approximates More: Earning Faithful Confidence in High-Stakes Domains
Haokai Ma, Lee Yan Zhen, Gang Yang et al. (+4 more)
Large
language models are increasingly deployed in high-stakes domains, where confident yet incorrect inferences may cause severe real-world harm, bringing the long-overlooked issue of confidence faithfulness to the forefront. A promising solution jointly optimizes unsupervised Reinforcement Learning from Internal Feedback (RLIF) with
reasoning-trace-guided
Reasoning Distillation (RD), yet it faces three persistent challenges, namely the scarcity of high-quality training corpora, factually unwarranted overconfidence, and erroneous updates amplified by indiscriminate fusion. Inspired by how human confidence accumulates from uncertainty to certainty, we propose Progressive
Reasoning Gain (PRG) to measure whether
reasoning steps progressively strengthen confidence in the final answer. Building on PRG, we introduce HyTuning, a hybrid post-training framework that adaptively reweights RD and RLIF, using scarce supervised
reasoning traces as a stable anchor while exploiting abundant unlabeled queries for scalability. Experiments on several domain-specific and general
benchmarks demonstrate that HyTuning improves accuracy while achieving confidence faithfulness under limited supervision, supporting a practical “Less Data Approximates More” effect. Our code will be released upon acceptance.
Read the full abstract ▾
All authors Haokai Ma, Lee Yan Zhen, Gang Yang, Yunxiang Chen, Yunshan Ma, Tat-Seng Chua, Ee-Chien Chang
Machine Learning
Adalina: Adaptive Linear Approximation for the Shapley Value and Beyond
Weida Li, Yaoliang Yu and Bryan Kian Hsiang Low
The Shapley value, and its broader family of semi-values, has received much
attention in various attribution problems. A fundamental and long-standing challenge is their efficient approximation, since exact computation generally requires an exponential number of utility queries in the number of players $n$. To meet the challenges of large-scale applications, we explore the limits of efficiently approximating semi-values under a $\Theta(n)$ space constraint. Building upon a vector concentration inequality, we establish a theoretical framework that enables sharper query complexities for existing unbiased randomized
algorithms. Within this framework, we systematically develop a linear-space
algorithm that requires $O(\frac{n}{\epsilon^{2}}\log\frac{1}{\delta})$ utility queries to ensure $P(\|\hat{\boldsymbol\phi}-\boldsymbol\phi\|_{2}\geq\epsilon)\leq \delta$ for all commonly used semi-values. In particular, our framework naturally bridges OFA, unbiased kernelSHAP, SHAP-IQ and the regression-adjusted approach, and definitively characterizes when paired sampling is beneficial. Moreover, our
algorithm allows explicit minimization of the mean squared error $\mathbb{E}[\|\hat{\boldsymbol\phi}-\boldsymbol\phi\|_{2}^{2}]$ for each specific utility function. Accordingly, we introduce the first adaptive, linear-time, linear-space randomized
algorithm, Adalina, that theoretically achieves improved mean squared error. All of our theoretical findings are experimentally validated. Our code is available at https://github.com/watml/adalina.
Read the full abstract ▾
Machine Learning
KV Cache Offloading for Context-Intensive Tasks
Andrey Bocharnikov, Ivan Ermakov, Denis Kuznedelev et al. (+2 more)
With the growing demand for long-context LLMs across a wide range of applications, the key-value (KV) cache has become a critical bottleneck for both latency and
memory usage. Recently, KV-cache offloading has emerged as a promising approach to reduce
memory footprint and inference latency while preserving accuracy. Prior evaluations have largely focused on tasks that do not require extracting large amounts of information from the context. In this work, we study KV-cache offloading on context-intensive tasks: problems where the solution requires looking up a lot of information from the input
prompt. We create and release the Text2JSON
benchmark, a highly context-intensive task that requires extracting structured knowledge from raw text. We evaluate modern KV offloading on Text2JSON and other context-intensive tasks and find significant performance degradation on both Llama 3 and Qwen 3 models. Our analysis identifies two key reasons for poor accuracy: low-rank projection of keys and unreliable landmarks, and proposes a simpler alternative strategy that significantly improves accuracy across multiple LLM families and
benchmarks. These findings highlight the need for a comprehensive and rigorous evaluation of long-context compression techniques.
Read the full abstract ▾
All authors Andrey Bocharnikov, Ivan Ermakov, Denis Kuznedelev, Vyacheslav Zhdanovskiy, Yegor Yershov
Machine Learning · Cross-listed from cs.AI
On-board Telemetry Monitoring in Autonomous Satellites: Challenges and Opportunities
Lorenzo Capelli, Leandro de Souza Rosa, Maurizio De Tommasi et al. (+8 more)
The increasing autonomy of spacecraft demands fault-detection systems that are both reliable and explainable. This work addresses eXplainable Artificial Intelligence for onboard Fault Detection, Isolation and Recovery within the Attitude and Orbit Control Subsystem by introducing a framework that enhances interpretability in neural anomaly detectors. We propose a method to derive low-dimensional, semantically annotated encodings from intermediate neural activations, called peepholes. Applied to a convolutional autoencoder, the framework produces interpretable indicators that enable the identification and localization of anomalies in reaction-wheel telemetry. Peepholes analysis further reveals bias detection and supports fault localization. The proposed framework enables the semantic characterization of detected anomalies while requiring only a marginal increase in computational resources, thus supporting its feasibility for on-board deployment.
Read the full abstract ▾
All authors Lorenzo Capelli, Leandro de Souza Rosa, Maurizio De Tommasi, Livia Manovi, Andriy Enttsel, Mauro Mangia, Riccardo Rovatti, Ilaria Pinci, Carlo Ciancarelli, Eleonora Mariotti, Gianluca Furano
Machine Learning · Cross-listed from cs.CL
Synthetic Data for any Differentiable Target
Tristan Thrush, Sung Min Park, Herman Brunborg et al. (+5 more)
What are the limits of controlling
language models via synthetic training data? We develop a reinforcement learning (RL) primitive, the
Dataset Policy Gradient (DPG), which can precisely optimize synthetic data generators to produce a
dataset of targeted examples. When used for supervised fine-tuning (SFT) of a target model, these examples cause the target model to do well on a differentiable metric of our choice. Our approach achieves this by taking exact data attribution via higher-order gradients and using those scores as policy gradient rewards. We prove that this procedure closely approximates the true, intractable gradient for the synthetic data generator. To illustrate the potential of DPG, we show that, using only SFT on generated examples, we can cause the target model's LM head weights to (1) embed a QR code, (2) embed the pattern $\texttt{67}$, and (3) have lower $\ell^2$ norm. We additionally show that we can cause the generator to (4) rephrase inputs in a new
language and (5) produce a specific UUID, even though neither of these objectives is conveyed in the generator's input
prompts. These findings suggest that DPG is a powerful and flexible technique for shaping model properties using only synthetic training examples.
Read the full abstract ▾
All authors Tristan Thrush, Sung Min Park, Herman Brunborg, Luke Bailey, Marcel Roed, Neil Band, Christopher Potts, Tatsunori Hashimoto
Machine Learning · Cross-listed from cs.GR
What a Comfortable World: Ergonomic Principles Guided Apartment Layout Generation
Piotr Nieciecki, Aleksander Plocharski and Przemyslaw Musialski
Current data-driven floor plan generation methods often reproduce the ergonomic inefficiencies found in real-world training
datasets. To address this, we propose a novel approach that integrates architectural design principles directly into a
transformer-based generative process. We formulate differentiable loss functions based on established architectural standards from literature to optimize room adjacency and proximity. By guiding the model with these ergonomic priors during training, our method produces layouts with significantly improved livability metrics. Comparative evaluations show that our approach outperforms baselines in ergonomic compliance while maintaining high structural validity.
Read the full abstract ▾
Machine Learning
Adversarial Label Invariant Graph Data Augmentations for Out-of-Distribution Generalization
Simon Zhang, Ryan P. DeMilt, Kun Jin et al. (+1 more)
Out-of-distribution (OoD) generalization occurs when representation learning encounters a distribution shift. This occurs frequently in practice when training and testing data come from different environments.
Covariate shift is a type of distribution shift that occurs only in the input data, while the concept distribution stays invariant. We propose RIA - Regularization for Invariance with
Adversarial training, a new method for OoD generalization under convariate shift. Motivated by an analogy to $Q$-learning, it performs an
adversarial exploration for counterfactual data environments. These new environments are induced by
adversarial label invariant data augmentations that prevent a collapse to an in-distribution trained learner. It works with many existing OoD generalization methods for
covariate shift that can be formulated as constrained optimization problems. We develop an alternating gradient descent-ascent
algorithm to solve the problem in the context of causally generated graph data, and perform extensive experiments on OoD graph classification for various kinds of synthetic and natural distribution shifts. We demonstrate that our method can achieve high accuracy compared with OoD baselines.
Read the full abstract ▾
All authors Simon Zhang, Ryan P. DeMilt, Kun Jin, Cathy H. Xia
Machine Learning
Zero-shot Multivariate Time Series Forecasting Using Tabular Prior Fitted Networks
Mayuka Jayawardhana, Nihal Sharma, Kazem Meidani et al. (+3 more)
Tabular foundation models, particularly Prior-data Fitted Networks like TabPFN have emerged as the leading contender in a myriad of tasks ranging from data imputation to label prediction on the tabular data format surpassing the historical successes of tree-based models. This has led to investigations on their applicability to
forecasting time series data which can be formulated as a tabular problem. While recent work to this end has displayed positive results, most works have limited their treatment of multivariate time series problems to several independent univariate time series
forecasting subproblems, thus ignoring any inter-channel interactions. Overcoming this limitation, we introduce a generally applicable framework for multivariate time series
forecasting using tabular foundation models. We achieve this by recasting the multivariate time series
forecasting problem as a series of scalar regression problems which can then be solved zero-shot by any tabular foundation model with regression capabilities. We present results of our method using the TabPFN-TS backbone and compare performance with the current state of the art tabular methods.
Read the full abstract ▾
All authors Mayuka Jayawardhana, Nihal Sharma, Kazem Meidani, Bayan Bruss, Tom Goldstein, Doron Bergman
Machine Learning
ADAPTive Input Training for Many-to-One Pre-Training on Time-Series Classification
Paul Quinlan, Qingguo Li and Xiaodan Zhu
Recent work on time-series models has leveraged self-supervised training to learn meaningful features and patterns in order to improve performance on downstream tasks and generalize to unseen modalities. While these pretraining methods have shown great promise in one-to-many scenarios, where a model is pre-trained on one
dataset and fine-tuned on a downstream
dataset, they have struggled to generalize to new
datasets when more
datasets are added during pre-training. This is a fundamental challenge in building foundation models for time-series data, as it limits the ability to develop models that can learn from a large variety of diverse
datasets available. To address this challenge, we present a new pre-training paradigm for time-series data called ADAPT, which can efficiently align the physical properties of data in the time-series domain, enabling mixed-batch pre-training despite the extreme discrepancies in the input sizes and channel dimensions of pre-training data. We trained on 162 time-series classification
datasets and set new state-of-the-art performance for classification
benchmarks. We successfully train a model within the time-series domain on a wide range of
datasets simultaneously, which is a major building block for building generalist foundation models in time-series domains.
Read the full abstract ▾
Machine Learning
SOLAR: Communication-Efficient Model Adaptation via Subspace-Oriented Latent Adapter Reparametrization
Seyed Mahmoud Sajjadi Mohammadabadi, Xiaolong Ma, Lei Yang et al. (+2 more)
Parameter-efficient fine-tuning (PEFT) methods, such as LoRA, enable scalable adaptation of foundation models by injecting low-rank adapters. However, their communication and storage costs remain a major bottleneck in resource-constrained settings. We propose
SOLAR (Subspace-Oriented Latent Adapter Reparameterization), a post-training compression framework that substantially reduces the communication cost (i.e., the number of parameters to transmit or store) of PEFT adapters.
SOLAR expresses each PEFT update as a linear combination of basis vectors formed from the foundation model's singular vectors with controlled random perturbations. By exploiting the subspace similarity (the
alignment of principal directions) between the foundation model and task-specific fine-tuned updates,
SOLAR decouples the adapter size from PEFT structure and ensures compact yet expressive representations. It is model-agnostic and compatible with existing PEFT methods, including LoRA, AdaLoRA, and other adapter modules. We theoretically establish a bound on the reconstruction error. Experiments on
language and
vision tasks using LLaMA, GPT, and ViT models demonstrate that
SOLAR preserves task performance while significantly reducing model representation sizes, offering an effective and communication-efficient solution for deployment in distributed systems and edge devices.
Read the full abstract ▾
All authors Seyed Mahmoud Sajjadi Mohammadabadi, Xiaolong Ma, Lei Yang, Feng Yan, Junshan Zhang
Machine Learning
Scaling-Aware Data Selection for End-to-End Autonomous Driving Systems
Tolga Dimlioglu, Nadine Chang, Maying Shen et al. (+2 more)
Large-scale deep learning models for physical AI applications depend on diverse training data collection efforts. These models and correspondingly, the training data, must address different evaluation criteria necessary for the models to be deployable in real-world environments. Data selection policies can guide the development of the training set, but current frameworks do not account for the ambiguity in how data points affect different metrics. In this work, we propose Mixture Optimization via Scaling-Aware Iterative Collection (MOSAIC), a general data selection framework that operates by: (i) partitioning the
dataset into domains; (ii) fitting neural scaling laws from each data domain to the evaluation metrics; and (iii) optimizing a data mixture by iteratively adding data from domains that maximize the change in metrics. We apply MOSAIC to autonomous driving (AD), where an End-to-End (E2E) planner model is evaluated on the Extended Predictive Driver Model Score (EPDMS), an aggregate of driving rule compliance metrics. Here, MOSAIC outperforms a diverse set of baselines on EPDMS with up to 80% less data.
Read the full abstract ▾
All authors Tolga Dimlioglu, Nadine Chang, Maying Shen, Rafid Mahmood, Jose M. Alvarez
Machine Learning · Cross-listed from cs.CL
Towards Real-world Human Behavior Simulation: Benchmarking Large Language Models on Long-horizon, Cross-scenario, Heterogeneous Behavior Traces
Jiawei Chen, Ruoxi Xu, Boxi Cao et al. (+12 more)
The emergence of Large
Language Models (LLMs) has illuminated the potential for a general-purpose user simulator. However, existing
benchmarks remain constrained to isolated scenarios, narrow action spaces, or synthetic data, failing to capture the holistic nature of authentic human behavior. To bridge this gap, we introduce OmniBehavior, the first predictive user simulation
benchmark constructed entirely from real-world data, integrating long-horizon, cross-scenario, and heterogeneous behavioral patterns into a unified framework. Based on this
benchmark, we first provide empirical evidence that previous
datasets with isolated scenarios suffer from tunnel
vision, whereas real-world decision-making relies on long-term, cross-scenario causal chains. Extensive evaluations of state-of-the-art LLMs reveal that current models struggle to accurately predict these complex behaviors, with performance plateauing even as context windows expand. Crucially, a systematic comparison between simulated and authentic behaviors uncovers a fundamental structural bias: LLMs tend to converge towards a positive average person, exhibiting hyper-activity, persona homogenization, and a utopian bias. This results in the loss of individual differences and long-tail behaviors, highlighting critical directions for future high-fidelity simulation research.
Read the full abstract ▾
All authors Jiawei Chen, Ruoxi Xu, Boxi Cao, Ruotong Pan, Yunfei Zhang, Yifei Hu, Yong Du, Tingting Gao, Han Li, Yaojie Lu, Yingfei Sun, Xianpei Han, Le Sun, Xiangyu Wu, Hongyu Lin
Machine Learning · Cross-listed from quant-ph
Scalable Neural Decoders for Practical Fault-Tolerant Quantum Computation
Andi Gu, J. Pablo Bonilla Ataides, Mikhail D. Lukin et al. (+1 more)
Quantum error correction (QEC) is essential for scalable
quantum computing. However, it requires classical decoders that are fast and accurate enough to keep pace with
quantum hardware. While
quantum low-density parity-check codes have recently emerged as a promising route to efficient fault tolerance, current decoding
algorithms do not allow one to realize the full potential of these codes in practical settings. Here, we introduce a convolutional neural network decoder that exploits the
geometric structure of QEC codes, and use it to probe a novel "waterfall" regime of error suppression, demonstrating that the logical error rates required for large-scale fault-tolerant
algorithms are attainable with modest code sizes at current physical error rates, and with latencies within the real-time budgets of several leading hardware platforms. For example, for the $[144, 12, 12]$ Gross code, the decoder achieves logical error rates up to $\sim 17$x below existing decoders - reaching logical error rates $\sim 10^{-10}$ at physical error $p=0.1\%$ - with 3-5 orders of magnitude higher throughput. This decoder also produces well-calibrated confidence estimates that can significantly reduce the time overhead of repeat-until-success protocols. Taken together, these results suggest that the space-time costs associated with fault-tolerant
quantum computation may be significantly lower than previously anticipated.
Read the full abstract ▾
All authors Andi Gu, J. Pablo Bonilla Ataides, Mikhail D. Lukin, Susanne F. Yelin
Machine Learning
Bias-Constrained Diffusion Schedules for PDE Emulations: Reconstruction Error Minimization and Efficient Unrolled Training
Constantin Le Cleï, Nils Thuerey and Xiaoxiang Zhu
Conditional
Diffusion Models are powerful surrogates for emulating complex spatiotemporal dynamics, yet they often fail to match the accuracy of deterministic neural emulators for high-precision tasks. In this work, we address two critical limitations of autoregressive PDE
diffusion models: their sub-optimal single-step accuracy and the prohibitive computational cost of unrolled training. First, we characterize the relationship between the noise schedule, the reconstruction error reduction rate and the
diffusion exposure bias, demonstrating that standard schedules lead to suboptimal reconstruction error. Leveraging this insight, we propose an Adaptive Noise Schedule framework that minimizes inference reconstruction error by dynamically constraining the model's exposure bias. We further show that this optimized schedule enables a fast Proxy Unrolled Training method to stabilize long-term rollouts without the cost of full Markov Chain sampling. Both proposed methods enable significant improvements in short-term accuracy and long-term stability over
diffusion and deterministic baselines on diverse
benchmarks, including forced Navier-Stokes, Kuramoto-Sivashinsky and Transonic Flow.
Read the full abstract ▾
Machine Learning
EgoEverything: A Benchmark for Human Behavior Inspired Long Context Egocentric Video Understanding in AR Environment
Qiance Tang, Ziqi Wang, Jieyu Lin et al. (+3 more)
Long context
egocentric video understanding has recently attracted significant research
attention, with augmented reality (AR) highlighted as one of its most important application domains. Nevertheless, the task remains highly challenging due to the need for
reasoning over extended temporal contexts and diverse, unstructured activities. Although several
benchmarks exist, most
egocentric datasets rely on human worn cameras and focus mainly on visual content, with limited consideration of underlying user behavior when forming
video-related queries. EgoEverything is a
benchmark that explicitly considers human behavior by leveraging human
attention signals, abstracted from gaze data, when generating questions. It comprises over 5,000 multiple choice question answer pairs, spanning more than 100 hours of
video. By integrating human
attention signals during question generation, it more faithfully captures natural human behavior and offers a realistic evaluation setting for long-context
egocentric video understanding in AR.
Read the full abstract ▾
All authors Qiance Tang, Ziqi Wang, Jieyu Lin, Ziyun Li, Barbara De Salvo, Sai Qian Zhang
Machine Learning
Leveraging Complementary Embeddings for Replay Selection in Continual Learning with Small Buffers
Danit Yanowsky and Daphna Weinshall
Catastrophic forgetting remains a key challenge in Continual Learning (CL). In replay-based CL with severe
memory constraints, performance critically depends on the sample selection strategy for the replay buffer. Most existing approaches construct
memory buffers using
embeddings learned under supervised objectives. However, class-agnostic, self-supervised representations often encode rich, class-relevant semantics that are overlooked. We propose a new method, Multiple
Embedding Replay Selection, MERS, which replaces the buffer selection module with a graph-based approach that integrates both supervised and self-supervised
embeddings. Empirical results show consistent improvements over SOTA selection strategies across a range of continual learning
algorithms, with particularly strong gains in low-
memory regimes. On CIFAR-100 and TinyImageNet, MERS outperforms single-
embedding baselines without adding model parameters or increasing replay volume, making it a practical, drop-in enhancement for replay-based continual learning.
Read the full abstract ▾
Machine Learning
Dead Weights, Live Signals: Feedforward Graphs of Frozen Language Models
Marcus Armstrong, Navid Ayoobi and Arjun Mukherjee
We present a feedforward graph architecture in which heterogeneous frozen large
language models serve as computational nodes, communicating through a shared continuous latent space via learned linear projections. Building on recent work demonstrating
geometric compatibility between independently trained LLM latent spaces~\cite{armstrong2026thinking}, we extend this finding from static two-model steering to end-to-end trainable multi-node graphs, where projection
matrices are optimized jointly via backpropagation through residual stream injection hooks. Three small frozen models (Llama-3.2-1B, Qwen2.5-1.5B, Gemma-2-2B) encode the input into a shared latent space whose aggregate signal is injected into two larger frozen models (Phi-3-mini, Mistral-7B), whose representations feed a lightweight cross-
attention output node. With only 17.6M trainable parameters against approximately 12B frozen, the architecture achieves 87.3% on ARC-Challenge, 82.8% on OpenBookQA, and 67.2% on MMLU, outperforming the best single constituent model by 11.4, 6.2, and 1.2 percentage points respectively, and outperforming parameter-matched learned classifiers on frozen single models by 9.1, 5.2, and 6.7 points. Gradient flow through multiple frozen model boundaries is empirically verified to be tractable, and the output node develops selective routing behavior across layer-2 nodes without explicit supervision.
Read the full abstract ▾
Machine Learning · Cross-listed from cs.CV
Lost in the Hype: Revealing and Dissecting the Performance Degradation of Medical Multimodal Large Language Models in Image Classification
Xun Zhu, Fanbin Mo, Xi Chen et al. (+6 more)
The rise of
multimodal large
language models (MLLMs) has sparked an unprecedented wave of applications in the field of medical imaging analysis. However, as one of the earliest and most fundamental tasks integrated into this paradigm, medical image classification reveals a sobering reality: state-of-the-art medical MLLMs consistently underperform compared to traditional deep learning models, despite their overwhelming advantages in pre-training data and model parameters. This paradox
prompts a critical rethinking: where exactly does the performance degradation originate? In this paper, we conduct extensive experiments on 14 open-source medical MLLMs across three representative image classification
datasets. Moving beyond superficial performance benchmarking, we employ feature probing to track the information flow of visual features module-by-module and layer-by-layer throughout the entire MLLM pipeline, enabling explicit visualization of where and how classification signals are distorted, diluted, or overridden. As the first attempt to dissect classification performance degradation in medical MLLMs, our findings reveal four failure modes: 1) quality limitation in visual representation, 2) fidelity loss in connector projection, 3) comprehension deficit in LLM
reasoning, and 4) misalignment of semantic mapping. Meanwhile, we introduce quantitative scores that characterize the healthiness of feature
evolution, enabling principled comparisons across diverse MLLMs and
datasets. Furthermore, we provide insightful discussions centered on the critical barriers that prevent current medical MLLMs from fulfilling their promised clinical potential. We hope that our work provokes rethinking within the community-highlighting that the road from high expectations to clinically deployable MLLMs remains long and winding.
Read the full abstract ▾
All authors Xun Zhu, Fanbin Mo, Xi Chen, Kaili Zheng, Shaoshuai Yang, Yiming Shi, Jian Gao, Miao Li, Ji Wu
Machine Learning · Cross-listed from eess.IV
HistDiT: A Structure-Aware Latent Conditional Diffusion Model for High-Fidelity Virtual Staining in Histopathology
Aasim Bin Saleem, Amr Ahmed, Ardhendu Behera et al. (+5 more)
Immunohistochemistry (IHC) is essential for assessing specific immune biomarkers like Human Epidermal growth-factor Receptor 2 (HER2) in breast cancer. However, the traditional protocols of obtaining IHC stains are resource-intensive, time-consuming, and prone to structural damages. Virtual staining has emerged as a scalable alternative, but it faces significant challenges in preserving fine-grained cellular structures while accurately translating biochemical expressions. Current state-of-the-art methods still rely on Generative
Adversarial Networks (GANs) or standard convolutional U-Net
diffusion models that often struggle with "structure and staining trade-offs". The generated samples are either structurally relevant but blurry, or texturally realistic but have artifacts that compromise their diagnostic use. In this paper, we introduce HistDiT, a novel latent conditional
Diffusion Transformer (DiT) architecture that establishes a new
benchmark for visual fidelity in virtual histological staining. The novelty introduced in this work is, a) the Dual-Stream Conditioning strategy that explicitly maintains a balance between spatial constraints via VAE-encoded latents and semantic phenotype guidance via UNI
embeddings; b) the multi-objective loss function that contributes to sharper images with clear morphological structure; and c) the use of the Structural Correlation Metric (SCM) to focus on the core morphological structure for precise assessment of sample quality. Consequently, our model outperforms existing baselines, as demonstrated through rigorous quantitative and qualitative evaluations.
Read the full abstract ▾
All authors Aasim Bin Saleem, Amr Ahmed, Ardhendu Behera, Hafeezullah Amin, Iman Yi Liao, Mahmoud Khattab, Pan Jia Wern, Haslina Makmur
Machine Learning
DMax: Aggressive Parallel Decoding for dLLMs
Zigeng Chen, Gongfan Fang, Xinyin Ma et al. (+2 more)
We present DMax, a new paradigm for efficient
diffusion language models (dLLMs). It mitigates error accumulation in parallel decoding, enabling aggressive decoding parallelism while preserving generation quality. Unlike conventional masked dLLMs that decode through a binary mask-to-
token transition, DMax reformulates decoding as a progressive self-refinement from mask
embeddings to
token embeddings. At the core of our approach is On-Policy Uniform Training, a novel training strategy that efficiently unifies masked and uniform dLLMs, equipping the model to recover clean
tokens from both masked inputs and its own erroneous predictions. Building on this foundation, we further propose Soft Parallel Decoding. We represent each intermediate decoding state as an interpolation between the predicted
token embedding and the mask
embedding, enabling iterative self-revising in
embedding space. Extensive experiments across a variety of
benchmarks demonstrate the effectiveness of DMax. Compared with the original LLaDA-2.0-mini, our method improves TPF on GSM8K from 2.04 to 5.47 while preserving accuracy. On MBPP, it increases TPF from 2.71 to 5.86 while maintaining comparable performance. On two H200 GPUs, our model achieves an average of 1,338 TPS at batch size 1. Code is available at: https://github.com/czg1225/DMax
Read the full abstract ▾
All authors Zigeng Chen, Gongfan Fang, Xinyin Ma, Ruonan Yu, Xinchao Wang
Machine Learning · Cross-listed from quant-ph
QARIMA: A Quantum Approach To Classical Time Series Analysis
Nishikanta Mohanty, Bikash K. Behera, Badshah Mukherjee et al. (+3 more)
We present QARIMA, a
quantum state-similarity-based reconstruction of the classical ARIMA modelling pipeline. Rather than using a
quantum circuit as a standalone forecaster, QARIMA preserves ARIMA's interpretable
forecasting structure while reformulating its core building blocks through analogous
quantum-compatible modules. The framework integrates
quantum differencing assessment, QACF/QPACF lag discovery, compact-swap-test state projection, swap-test/VQC-based AR and MA coefficient estimation, and weak-lag refinement within a single ARIMA
forecasting workflow. QACF and QPACF serve the functional roles of ACF and PACF for MA and AR lag discovery, but construct lag relevance through
quantum measurement geometry rather than direct classical correlation. Given screened candidate orders $(p,d,q)$, AR and MA coefficients are estimated through state-
alignment losses incorporating cosine
alignment,
entropy regularization, phase correction, and norm control. We evaluate QARIMA across environmental, climatic, industrial, and weather time-series
datasets using rolling-origin out-of-sample testing against automated classical ARIMA baselines, with performance assessed through MSE, MAPE, and Diebold–Mariano tests. The results show that
quantum state-similarity modules can produce competitive and, in several cases, improved
forecasting behaviour while preserving ARIMA's transparency, modularity, and interpretability. QARIMA therefore establishes a distinct pathway for
quantum-enhanced statistical
forecasting: its modules remain functionally analogous to classical ARIMA subroutines, but are not algebraic replicas; they serve the same modelling roles through state overlap, projection, and measurement-driven parameter estimation.
Read the full abstract ▾
All authors Nishikanta Mohanty, Bikash K. Behera, Badshah Mukherjee, Pravat Dash, Giuseppe Sergioli, Roberto Giuntini
Machine Learning
An Illusion of Unlearning? Assessing Machine Unlearning Through Internal Representations
Yichen Gao, Altay Unal, Akshay Rangamani et al. (+1 more)
While numerous machine unlearning (MU) methods have recently been developed with promising results in erasing the influence of forgotten data, classes, or concepts, they are also highly vulnerable-for example, simple fine-tuning can inadvertently reintroduce erased concepts. In this paper, we address this contradiction by examining the internal representations of unlearned models, in contrast to prior work that focuses primarily on output-level behavior. Our analysis shows that many state-of-the-art MU methods appear successful mainly due to a misalignment between last-layer features and the classifier, a phenomenon we call feature-classifier misalignment. In fact, hidden features remain highly discriminative, and simple linear probing can recover near-original accuracy. Assuming neural collapse in the original model, we further demonstrate that adjusting only the classifier can achieve negligible forget accuracy while preserving retain accuracy, and we corroborate this with experiments using classifier-only fine-tuning. Motivated by these findings, we propose MU methods based on a class-mean features (CMF) classifier, which explicitly enforces
alignment between features and classifiers. Experiments on standard
benchmarks show that CMF-based unlearning reduces forgotten information in representations while maintaining high retain accuracy, highlighting the need for faithful representation-level evaluation of MU.
Read the full abstract ▾
All authors Yichen Gao, Altay Unal, Akshay Rangamani, Zhihui Zhu
Machine Learning
Introducing Echo Networks for Computational Neuroevolution
Christian Kroos and Fabian Küch
For applications on the extreme edge, minimal networks of only a few dozen artificial
neurons for event detection and classification in discrete time signals would be highly desirable. Feed-forward networks, RNNs, and CNNs evolved through evolutionary
algorithms can all be successful in this respect but pose the problem of allowing little systematicity in mutation and recombination if the standard direct genetic encoding of the weights is used (as for instance in the classic NEAT
algorithm). We therefore introduce Echo Networks, a type of recurrent network that consists of the connection
matrix only, with the source
neurons of the synapses represented as rows, destination
neurons as columns and weights as entries. There are no layers, and connections between
neurons can be bidirectional but are technically all recurrent. Input and output can be arbitrarily assigned to any of the
neurons and only use an additional (optional) function in their computational path, e.g., a sigmoid to obtain a binary classification output. We evaluated Echo Networks successfully on the classification of electrocardiography signals but see the most promising potential in their genome representation as a single
matrix, allowing
matrix computations and factorisations as mutation and recombination operators.
Read the full abstract ▾
Machine Learning
Approximation of the Basset force in the Maxey-Riley-Gatignol equations via universal differential equations
Finn Sommer, Vamika Rathi, Sebastian Goetschel et al. (+1 more)
The Maxey-Riley-Gatignol equations (MaRGE) model the motion of spherical inertial particles in a fluid. They contain the Basset force, an integral term which models history effects due to the formation of wakes and boundary layer effects. This causes the force that acts on a particle to depend on its past trajectory and complicates the numerical solution of MaRGE. Therefore, the Basset force is often neglected, despite substantial evidence that it has both quantitative and qualitative impact on the movement patterns of modelled particles. Using the concept of universal differential equations, we propose an approximation of the history term via neural networks which approximates MaRGE by a system of ordinary differential equations that can be solved with standard numerical solvers like Runge-Kutta methods.
Read the full abstract ▾
All authors Finn Sommer, Vamika Rathi, Sebastian Goetschel, Daniel Ruprecht
Machine Learning
Inside-Out: Measuring Generalization in Vision Transformers Through Inner Workings
Yunxiang Peng, Mengmeng Ma, Ziyu Yao et al. (+1 more)
Reliable generalization metrics are fundamental to the evaluation of machine learning models. Especially in high-stakes applications where labeled target data are scarce, evaluation of models' generalization performance under distribution shift is a pressing need. We focus on two practical scenarios: (1) Before deployment, how to select the best model for unlabeled target data? (2) After deployment, how to monitor model performance under distribution shift? The central need in both cases is a reliable and label-free proxy metric. Yet existing proxy metrics, such as model confidence or accuracy-on-the-line, are often unreliable as they only assess model output while ignoring the internal mechanisms that produce them. We address this limitation by introducing a new perspective: using the inner workings of a model, i.e.,
circuits, as a predictive metric of generalization performance. Leveraging
circuit discovery, we extract the causal interactions between internal representations as a
circuit, from which we derive two metrics tailored to the two practical scenarios. (1) Before deployment, we introduce Dependency Depth Bias, which measures different models' generalization capability on target data. (2) After deployment, we propose
Circuit Shift Score, which predicts a model's generalization under different distribution shifts. Across various tasks, both metrics demonstrate significantly improved correlation with generalization performance, outperforming existing proxies by an average of 13.4% and 34.1%, respectively. Our code is available at https://github.com/deep-real/GenCircuit.
Read the full abstract ▾
All authors Yunxiang Peng, Mengmeng Ma, Ziyu Yao, Xi Peng