Machine Learning
Lost in the Non-convex Loss Landscape: How to Fine-tune the Large Time Series Model?
Xu Zhang, Peang Wang and Wei Wang
Recently, large time series models (LTSMs) have gained increasing
attention due to their similarities to large
language models, including flexible context length, scalability, and task generality, outperforming advanced task-specific models. However, prior studies indicate that pre-trained LTSMs may exhibit a poorly conditioned non-convex loss landscape, leading to limited trainability. As a result, direct fine-tuning tends to cause overfitting and suboptimal performance, sometimes even worse than training from scratch, substantially diminishing the benefits of pre-training. To overcome this limitation, we propose Smoothed Full Fine-tuning (SFF), a novel fine-tuning technology. Specifically, we construct an auxiliary LTSM via random initialization to obtain a smoother loss landscape, and then linearly interpolate its weights with those of the pre-trained model to smooth the original landscape. This process improves trainability while preserving pre-trained knowledge, thereby enabling more effective downstream fine-tuning. From an optimization perspective, SFF perturbs sharp minima without significantly harming flat regions, facilitating escape from poor local basins toward smoother and more generalizable solutions. Extensive experiments on
benchmark datasets demonstrate consistent improvements across eight representative LTSMs, including Timer, TimesFM, MOMENT, UniTS, MOIRAI, Chronos, TTMs, and Sundial, on diverse downstream tasks. The code is available at the link: https://github.com/Meteor-
Stars/SFF.
Read the full abstract ▾
Machine Learning
OrderDP: A Theoretically Guaranteed Lossless Dynamic Data Pruning Framework
Chenhan Jin, Shengze Xu, Qingsong Wang et al. (+3 more)
Data
pruning (DP), as an oft-stated strategy to alleviate heavy training burdens, reduces the volume of training samples according to a well-defined
pruning method while striving for near-lossless performance. However, existing approaches, which commonly select highly informative samples, can lead to biased gradient estimation compared to full-
dataset training. Furthermore, the analysis of this bias and its impact on final performance remains ambiguous. To address these challenges, we propose OrderDP, a plug-and-play framework that aims to obtain stable, unbiased, and near-lossless training acceleration with theoretical guarantees. Specifically, OrderDP first randomly selects a subset and then chooses the top-$q$ samples, where unbiasedness is established with respect to a surrogate loss. This ensures that OrderDP conducts unbiased training in terms of the surrogate objective. We further establish convergence and generalization analyses, elucidating how OrderDP affects optimal performance and enables well-controlled acceleration while ensuring guaranteed final performance. Empirically, we evaluate OrderDP against comprehensive baselines on CIFAR-10, CIFAR-100, and ImageNet-1K, demonstrating competitive accuracy, stable convergence, and exact control – all with a simpler design and faster runtime, while reducing training cost by over 40%. Delivering both strong performance and computational efficiency, our method serves as a robust and easily adaptable tool for data-efficient learning. The code is publicly available at https://github.com/shengze-xu/OrderDP.
Read the full abstract ▾
All authors Chenhan Jin, Shengze Xu, Qingsong Wang, Fan Jia, Dingshuo Chen, Tieyong Zeng
Machine Learning
Titans-as-a-Layer: Test-Time Memory for Conversational Speech Emotion Recognition
Daniel Chen, Qicong Hu, Yang Xiao et al. (+2 more)
Speech emotion recognition (SER) is commonly formulated as utterance-level classification, although conversational
emotion depends on a speaker's usual vocal range and the emotional context established by previous utterances.
Speech-
language models provide strong pretrained acoustic and semantic representations, and can adapts them to SER labels via finetune, but this mechanism still missing per-dialogue state. We study whether test-time neural
memory can supply this missing context while leaving the large audio
language models (LALMs) backbone intact. Building on Titans, we introduce a plug-and-play
Memory-as-a-Layer (MAL) adapter that writes dialogue history into a small neural
memory and reads it back as an audio-
token-aligned residual update, avoiding changes to the host model's
token positions. Across different audio LLMs and
emotion recognition
datasets evaluations, our design improves SER performs across different evaluation metrics, supporting test-time
memory as a residual contextual mechanism for conversational SER.
Read the full abstract ▾
All authors Daniel Chen, Qicong Hu, Yang Xiao, Ting Dang, Hong Jia
Machine Learning · Cross-listed from cs.CL
Calibration of Structured Ignorance Certificates for Diagnosing Unknown Unknowns in Reasoning Models
Subramanyam Sahoo
Large
language models frequently fail in a characteristic way: rather than acknowledging ignorance, they produce fluent but incorrect answers to questions that lie beyond their knowledge boundaries. We introduce Structured Ignorance Certificates (SICs), a JSON-formatted output schema that demands a model explicitly name the missing domain intersection, enumerate required concepts, and propose a productive
retrieval query rather than hallucinating an answer. To train models to produce high-quality SICs we construct a 7,347-sample Unknown-Unknown (UU)
dataset by prompting Qwen3-14B to stitch together questions from seven domains (physics, biology, engineering, CS, economics, medical, legal) into novel cross-domain queries that no single-domain expert could answer. We fine-tune a 14B-parameter model with Group Relative Policy Optimization (GRPO) using a composite reward that combines
retrieval utility, concept specificity, and output-format validity. A paraphrase-divergence probe trained on model responses confirms that SIC-tuned outputs systematically exhibit higher unknown-unknown probability scores. Evaluation on 735 held-out UU questions achieves a 99.46% JSON validity rate, a mean Certificate Specificity Score of 0.967, and a 3.6% ROUGE-L improvement over the base model on
retrieval-grounded generation – demonstrating that explicit epistemic structuring is a learnable and measurable capability.
Read the full abstract ▾
Machine Learning
EinSort: Sorting is All We Need for Tensorizing LLM
Toshiaki Koike-Akino, Jing Liu and Ye Wang
Tensor networks provide efficient representations for compressing large neural networks. By carefully designing shapes and
topologies, they can significantly reduce
memory and computational costs. However, identifying implicit low-rank structures in large foundation models remains challenging due to their enormous scale and un-structured weight distributions. We propose an adaptive tensorization method that discovers inherent low-rank structure in a target tensor by index ordering. Experiments on weight and KV-cache compression demonstrate improved reconstruction quality compared to baselines.
Read the full abstract ▾
Machine Learning
Physics-Guided Dual Decoding and Spectral Supervision for Global 3D Hydrometeor Prediction
Dandan Chen and Yaqiang Wang
While global data-driven models excel at predicting continuous atmospheric variables, three-dimensional hydrometeor
forecasting remains challenging due to the zero-inflated, long-tailed distributions of these variables. Standard deep learning optimization often yields overly smooth forecasts, attenuating extreme events and spatial textures. We propose PredHydro-Net, a physics-guided dual-decoding framework that mitigates this smoothing. To resolve multi-variable optimization conflicts, it employs a decoupled architecture where macroscopic thermodynamic and dynamic fields unidirectionally modulate hydrometeor generation. By integrating wavelet-based frequency decoupling, spectral amplitude matching, and
adversarial training, the model achieves a favorable trade-off between quantitative accuracy and spatial fidelity. In a 72-h global evaluation, PredHydro-Net outperforms both spatiotemporal deep learning baselines (Earthformer and PredRNNv2) and the operational Global Forecast System (GFS) in extreme-event detection and spectral representation. Furthermore, it demonstrates strong climatological consistency with Global Precipitation Measurement (GPM) satellite
retrievals. The model reasonably reproduces the three-dimensional cloud structures in extreme weather events, such as Hurricane Ian. Feature attribution confirms its dependence on physical precursors such as relative humidity and wind convergence, offering a robust, physics-informed approach to long-tailed atmospheric prediction.
Read the full abstract ▾
Machine Learning
A Theoretical Analysis of Memory and Overfitting Phenomena in Stochastic Interpolation Models
Yunchen Li, Shaohui Lin and Zhou Yu
This paper provides a theoretical account of memorization in
stochastic interpolation models. By leveraging closed-form expressions for the optimal velocity field and the associated score function, we show that, in the continuous-time oracle setting, both deterministic and
stochastic generation processes recover training samples. Under Euler discretization, generated samples remain centered around training samples, with deviations controlled by the step size. We further analyze generation in the presence of estimation errors and show that accumulated estimation errors control the endpoint deviation from the training set. These results imply that the generated sample admits a representation as a training sample perturbed by three controlled terms: a discretization-induced bound, an estimation-error-induced bound, and
stochastic Gaussian noise. Based on this characterization, we provide theoretical definitions of overfitting and underfitting in generative models. Synthetic simulations support our theoretical findings.
Read the full abstract ▾
Machine Learning
Routine laboratory trajectories encode the onset of organ-level complications in cancer
Jannik Lübberstedt, Krischan Braitsch, Jacqueline Lammert et al. (+21 more)
Routine laboratory panels drawn during cancer treatment constitute longitudinal physiological recordings of organ function, yet their temporal structure is discarded by single-timepoint prognostic tools. A
transformer trained on 2,777,595 laboratory measurements from 3,905 patients with multiple myeloma or ovarian cancer predicted the two-year onset of 162 treatment-associated complications, including therapy-related myelodysplastic syndromes, spanning eight clinical categories, achieving 1.5- to 6.1-fold enrichment above prevalence at the group level. It matched or outperformed non-sequential baselines across grouped endpoints (AUROC gains up to +0.11), demonstrating that longitudinal laboratory trajectories capture evolving complication-specific physiology inaccessible from isolated measurements. Predictions generalised across both cancers, divergence concentrating in disease-specific complications, and biomarker masking recovered signatures consistent with established pathophysiology. External validation on MIMIC-IV and MMRF CoMMpass confirmed transferability across independent healthcare systems (AUROC up to 0.85). Routine oncological laboratory data encode organ deterioration weeks to months before clinical onset, enabling complication-specific surveillance without additional testing infrastructure.
Read the full abstract ▾
All authors Jannik Lübberstedt, Krischan Braitsch, Jacqueline Lammert, Christof Winter, Florian Gabriel, Tristan Lemke, Christopher Zirn, Markus Graf, Friedrich Puttkammer, Hartmut Häntze, Johannes Moll, Anirudh Narayanan, Andrei Zhukov, Fabian Drexel, Zeineb Ben Chaaben, Sebastian Ziegelmayer, Su Hwan Kim, Marion Högner, Jan Kirschke, Florian Bassermann, Marcus Makowski, Christian Wachinger, Lisa Adams, Keno Bressem
Machine Learning
Autonomous Aerial Manipulation via Contextual Contrastive Meta Reinforcement Learning
Lixuan Jin, Bingxuan Lan, Xinyi Bao et al. (+9 more)
Unmanned aerial vehicles (
UAVs) are increasingly being deployed in logistics, service robotics, and other real-world applications, creating a growing demand for autonomous payload acquisition and delivery. Existing approaches typically assume pre-attached payloads or rely on specialized grippers, leaving versatile end-to-end aerial delivery largely unresolved, where different payloads induce highly variable flight dynamics, requiring a single policy to adapt online without manual calibration or explicit system identification. To this end, we study Autonomous Aerial Manipulation via Contextual Contrastive Meta Reinforcement Learning (\textbf{Aco2}), a fully autonomous aerial delivery setting in which a quadrotor equipped with a lightweight hook continuously picks up, transports, and delivers diverse handle-equipped objects between randomized locations, all without human intervention. First, we design a contextual observation encoder that infers a compact latent context from recent interaction history, enabling the policy to adapt online to payload-dependent dynamics. To further improve the quality of this context, we introduce a contrastive objective that structures the context
embedding around task-relevant variations, improving generalization across diverse payloads without requiring explicit system identification. Trained entirely in simulation with extensive domain randomization, Aco2 can be directly deployed on a physical quadrotor without real-world fine-tuning.
Read the full abstract ▾
All authors Lixuan Jin, Bingxuan Lan, Xinyi Bao, Xiangyuan Xie, Chunjie Zhang, Zheng Chen, Tianshuo Liu, Ruijie Tian, Jinyu Ru, Gang Wang, Lei Yuan, Yang Yu
Machine Learning · Cross-listed from cs.AI
Scaffold Effects on GAIA: A Controlled Comparison
Jason Starace
Published
agent capability scores conflate what a model can do with what its scaffold lets it do, and the magnitude of this elicitation gap is not well characterized under controlled conditions. This study executes a pre-registered controlled comparison of three scaffolds (ReAct, a Planner-Actor-Rater multi-
agent design, and planner-then-executor) across five models from three providers (Claude Opus 4.7, Sonnet 4.6, Haiku 4.5; Gemini 3.1 Pro Preview; GPT-5.5) on
GAIA validation Levels 1 and 2, holding tasks and conditions fixed, with three attempts per question. Scaffold choice alone moves measured accuracy by as much as 28 percentage points within a single model (Opus, Level 2, robust slice), confirming the pre-registered hypothesis that scaffold variation produces gaps of at least 10 points. The pre-registered prediction that more capable models would be less scaffold-sensitive is rejected in direction: scaffold effects vary significantly by model in every
dataset slice, but the most capable Anthropic model gains the most from structured scaffolds at the harder level, and tier-scaling holds only at Level 1 under the robust slice. The multi-
agent advantage over ReAct at Level 2 appears within the Anthropic family but not for the cross-provider models, making model family rather than capability tier the conditioning variable, and the predicted planner-executor advantage on file-reading tasks is falsified. Structured scaffolds make fewer tool calls yet recover more often from mid-trajectory errors at the harder level, and a single
cell (Gemini with planner-then-executor) is the cheapest at both levels and the most accurate at Level 2. These results indicate that single-scaffold capability numbers are scaffold-conditional estimates and that the elicitation gap is not guaranteed to shrink as models improve.
Read the full abstract ▾
Machine Learning
A Joint Finite-Sample Certificate for Adaptive Selective Conformal Risk Control
Xiaoli Yu and Jiamiao Liu
Selective predictors answer on confident inputs and abstain elsewhere; deploying one safely needs a single finite-sample certificate that simultaneously upper-bounds the selected risk, lower-bounds the acceptance probability $\pacc$ above a floor $\pmin$, and lower-bounds the deployment utility. This certificate must be valid under adaptive threshold selection from a finite grid of $m$ pairs on $\ncert$ samples. We give such a certificate for bounded, possibly non-monotone losses by treating the selected risk directly as a ratio rather than through a Hoeffding-style range bound. The construction couples three confidence bounds: a variance-adaptive empirical-Bernstein bound on the ratio risk, a Clopper–Pearson bound on acceptance, and a two-sided closeness bound on utility. Together they lower-bound the certified policy's utility absolutely and to within $2\gammau$ of the best over the certified set, both non-vacuous whenever feasible; a regime-scoped third leg matches an external oracle, informative only where the risk margin $\gammar < \alpha$ and vacuous at the headline operating points. Relative to the range-only Hoeffding-ratio construction this sharpens the acceptance-floor dependence from $1/\pmin$ to $1/\sqrt{\pmin}$, and a closed-form corollary identifies a per-pair regime in which our risk bound dominates a Hoeffding conformal risk control (Hoeffding–CRC) selective bound. Empirically, on ImageNet (three ResNets) and COCO val 2017 panoptic, the certificate opens a $+22$ pp certified-acceptance frontier over Hoeffding–CRC and is ${\approx}10{\times}$ tighter than a non-vacuous matched-valid baseline; these gains are regime-scoped, not universal, and absent on ADE20K. The certifier runs in $O(\ncert m)$ time.
Read the full abstract ▾
Machine Learning · Cross-listed from cs.RO
Towards End to End Motion Planning and Execution for Autonomous Underwater Vehicles Using Reinforcement Learning
Elisei Shafer and Oren Gal
Autonomous
Underwater Vehicles (AUVs) traditionally rely on complex, heavily engineered pipelines for perception, path planning, and motion control. This paper explores the feasibility of an end-to-end Deep Reinforcement Learning (DRL) approach that maps raw sensor data directly to thruster commands, reducing manual engineering. We propose a hierarchical reinforcement learning (HRL) architecture splitting the problem into two Markov Decision Processes. A High-Level (HL) policy operating at 2Hz processes raw $84 \times 84$ pixel
monocular camera frames, stacked $100 \times 100$ pixel forward-looking imaging sonar, and proprioceptive data to generate spatial subgoals. Simultaneously, a Low-Level (LL) policy operating at 10Hz converts these subgoals into thruster commands. The HL policy is trained using Reinforcement Learning from Prior Demonstrations (RLPD) within a modified Sample-Efficient Robotic Reinforcement Learning (SERL) framework, while the LL policy utilizes Soft Actor-Critic (SAC) combined with Hindsight Experience Replay (HER). Evaluated in the high-fidelity HoloOcean simulator, our method demonstrates successful obstacle avoidance, achieving trajectory lengths closely approximating (within 4% to 6% of) an $\text{RRT}^*$ planning baseline. Furthermore, the learned policy exhibits strong robustness to simulated sensor noise and decreased visibility. While the system navigates familiar geometries effectively, experiments reveal generalization limitations when encountering unvisited areas with novel obstacle shapes. Ultimately, this work demonstrates the promise of sample-efficient, end-to-end DRL for
underwater navigation using minimal computational hardware.
Read the full abstract ▾
Machine Learning · Cross-listed from cs.CL
SAEExplainer: Interpreting SAE Features with Activation-Guided Preference Optimization
Jingyi He, Haiyan Zhao, Ruxue Shi et al. (+4 more)
Although Sparse Autoencoders (SAEs) have mitigated the opacity of large
language models (LLMs) by decomposing dense representations into sparse features, explaining these features still remains a central challenge. Current explanation methods, however, typically operate within an open-loop paradigm, failing to leverage mechanistic feedback for further refinement. In this paper, we propose SAEExplainer, a training framework that utilizes activation scores as an objective reward signal to train the model for self-correction and iterative bootstrapping. By iteratively verifying and correcting foundational explanations through a two-round optimization process, SAEExplainer achieves continuous improvement in its explanatory capabilities. This mechanism significantly reduces explanation
hallucinations and reinforces causal triggering patterns. Extensive experiments demonstrate our approach improves upon established baselines across most metrics, especially in causal triggering and discriminative activation. The code is available at https://github.com/he-jingyi/SAEExplainer.
Read the full abstract ▾
All authors Jingyi He, Haiyan Zhao, Ruxue Shi, Yanguang Liu, Xin Wang, Fei Sun, Mengnan Du
Machine Learning · Cross-listed from q-bio.GN
Querying Counterfactuals on Tissue Graphs with Supervised Disentanglement
Abdul Moeed, Stefan Schrod, Martin Rohbeck et al. (+4 more)
Tissue graph counterfactuals ask what a
cell would express if its spatial neighborhood were altered, a question central to predicting
cell behavior in tissues. Existing methods, however, present only specific instances of such queries. In this work, we first formalize tissue graph counterfactuals as a class of spatial interventions that either rewire connections between
cells (edge perturbation) or modify the expression of their neighbors (node perturbation). We then introduce Cellina, a framework that uses supervised disentanglement to decompose a
cell's intrinsic state from its spatial context, using the latter as a conditioning input for in-silico perturbation predictions. Across
benchmarks spanning over 2.5 million spatially-resolved
cells in observational and interventional screens, Cellina outperforms both spatially-informed and non-spatial competitors in in-silico perturbations, while training as fast as non-spatial methods. Moreover, we show that Cellina reveals cancer subdomains with distinct signaling programs in an unsupervised manner and enables targeted neighbor perturbation simulations.
Read the full abstract ▾
All authors Abdul Moeed, Stefan Schrod, Martin Rohbeck, Marc Jan Bonder, Pavlo Lutsik, Oliver Stegle, Daniel Dimitrov
Machine Learning
STELLAR: Spatio-Temporal Environmental Learning with Latent Alignment and Refinement for Long-Tailed Species Distribution Modeling
Shufeng Kong, Tao Yu, Yuanyuan Wei et al. (+6 more)
Joint Species Distribution Modeling (JSDM) is a key enabler for biodiversity monitoring and conservation planning. However, accurate JSDM faces two coupled challenges: environmental drivers and species distributions are inherently spatio-temporal, while species co-occurrence patterns exhibit complex non-linear community structure and severe long-tail imbalance driven by rare species. Existing approaches often address these factors in isolation, learning from static
covariates or neglecting the historical trajectories of dynamic community structure. To overcome these limitations, we propose
STELLAR (Spatio-Temporal Environmental Learning with Latent
Alignment and Refinement), a novel framework that learns a shared latent space where dynamic habitat context and community structure are optimized jointly. Our approach integrates three complementary components: (1) a Graph-Temporal Encoder that employs graph
attention and recurrent units to aggregate spatial neighborhood effects and capture the co-evolving historical dynamics of environmental context and community structure; (2) a Context-Anchored Latent
Alignment mechanism that structures the latent space using a label-activated mixture prior and supervised contrastive learning, actively clustering species based on shared environmental preferences; and (3) an Imbalance-Aware Decoupled Decoding module that utilizes Asymmetric Loss to focus learning on hard, rare species samples, preventing mode collapse in the long tail. Experiments on the large-scale eBird
dataset, curated with domain experts, demonstrate that our framework significantly outperforms state-of-the-art baselines, particularly in predicting rare species and revealing interpretable species interactions.
Read the full abstract ▾
All authors Shufeng Kong, Tao Yu, Yuanyuan Wei, Caihua Liu, Junwen Bai, Yingheng Wang, Marc Grimson, Daniel Fink, Carla P. Gomes
Machine Learning
PIPE-Cypher: Automatic Enterprise Benchmark Generation for Text-to-Cypher Systems
Suraj Ranganath and Anish Raghavendra
Enterprise property graphs vary widely in schema structure, internal terminology, domain assumptions, governance constraints, and user interaction patterns. A deployment-relevant Text2Cypher
benchmark therefore reflects the questions users and
agents actually ask of that graph. Creating such a
benchmark is difficult because schemas and values are unique, and graph structure changes over time. Each NL-query pair must also be executable, use real graph entities, preserve diversity, and remain balanced across query types and difficulty levels. We present PIPE-Cypher, a local
benchmark-generation pipeline that turns a live property graph and optional seed queries from customer questions, analyst logs, or
agent tool calls into balanced NL-to-Cypher
benchmarks. PIPE-Cypher combines schema profiling, reverse-query grounding, constrained generation, deterministic Cypher governance, execution validation, redaction, diversity controls, and a calibrated local LLM judge. Using local Qwen3.5-9B generation and judging, PIPE-Cypher exports 3,000 accepted FinBench/SNB examples, completes three audited ablation suites, calibrates judge behavior with human labels, and evaluates 11 local downstream models. The resulting
benchmark is deliberately discriminative: zero-shot transfer is weak, while a few-shot control shows that schema-specific example banks can help compatible model families. Together, PIPE-Cypher makes Text2Cypher benchmarking a repeatable process that evolves with the graph, its users, and its target workloads.
Read the full abstract ▾
Machine Learning
Adaptive Loss Balancing for Noise-Robust GRPO in Generative Recommendation
Kewei Xu, Junbo Qi, Yanyan Zou et al. (+3 more)
Reinforcement learning (RL) presents a promising avenue for enhancing generative recommendation beyond supervised imitation, leveraging reward signals to guide policy improvement. However, its efficacy is critically contingent on the trustworthiness of the reward model for the samples it evaluates. In practice, production rankers, the widely adopted reward models, are trained on exposure-biased logs, leading to sample-dependent inaccuracies that violate this assumption. Our stratified analysis uncovers a consistent pattern: reward guidance is most beneficial when the policy exhibits uncertainty and the ranker can effectively discriminate the ground-truth item from rollout negatives. On other samples, the reward signal is either negligible or detrimental, highlighting the risk of uniform RL application. To address such an issue, we introduce AdaGRPO, a novel framework that treats reward-guided optimization as selective admission rather than uniform pressure. Training is anchored in supervised negative log-likelihood, while the GRPO objective is gated by a binary, per-sample clip determined by two rollout diagnostics: policy-side difficulty and reward discriminability. Instances failing either diagnostic default to pure supervision, ensuring stability and mitigating the amplification of noisy gradients. We validate AdaGRPO on a large-scale e-commerce
dataset. At the best intermediate checkpoint, it elevates HR@10 from 11.01% to 12.18% while constraining
hallucination below 0.22%, and maintains robustness at the final checkpoint (HR@10 11.63%,
hallucination 0.27%), outperforming fixed NLL–GRPO mixtures across the
retrieval–validity frontier. In production A/B tests, AdaGRPO achieves statistically significant gains in click-through rate and dwell time, confirming its practical utility.
Read the full abstract ▾
All authors Kewei Xu, Junbo Qi, Yanyan Zou, Pengfei Zhang, Xingzhi Yao, Shengjie Li
Machine Learning
Inferring hidden forcing in a biological oscillator using Kolmogorov-Arnold networks
Julian Szereszewski, Facundo Fainstein, Leandro E. Fernandez et al. (+1 more)
Inferring the forces that drive a dynamical system from partial observations is a fundamental challenge across physics, particularly when distinct underlying mechanisms produce similar observable dynamics. Here we show that the effective muscular forcing underlying avian respiratory dynamics can be reconstructed from measurements of air-sac pressure alone. Using an interpretable learning framework based on Kolmogorov-Arnold networks, we infer the governing equations of the system directly from data and uncover a nontrivial structure in the underlying forcing that is not apparent from the pressure signal, which instead suggests a relaxation-like oscillation. The reconstructed dynamics predict a two-phase activation pattern within each respiratory cycle, which we independently validate through electromyographic recordings of expiratory muscles. These results demonstrate that data-driven reconstruction of dynamical laws can reveal hidden physical structure and provide access to unobserved driving variables, establishing a general route to infer latent forces in partially observed dynamical systems.
Read the full abstract ▾
All authors Julian Szereszewski, Facundo Fainstein, Leandro E. Fernandez, Gabriel B. Mindlin
Machine Learning
Physically Consistent Null Space Alignment for Detection of Low-Magnitude False Data Injection Attacks
Xin Li, Chenhan Xiao, Jonathan Cohen et al. (+3 more)
False data injection attacks (FDIAs) introducing small measurement perturbations can still cause large deviations in power system state estimation when the injected signals align with the pseudo-null space of the system model. Existing model- and data-driven detectors may fail to identify such low-magnitude but high-impact attacks because residual tests ignore changes hidden in the pseudo-null space, while subspace learning methods capture correlation patterns without enforcing physical consistency. This paper proposes Physically Consistent Null Space
Alignment (PCNSA), a framework that detects stealthy FDIAs by preserving, through preprocessing, the
geometric correspondence between the physical null space and the measurement-derived pseudo-null space. The key point is a Pseudo-null Space Conserved data Preprocessing (PSCP) step that re-expresses measurements in the physical coordinate frame before subspace extraction. We prove that PSCP preserves the separation between row space and its orthogonal complement, a property that conventional per-feature standardization violates. This keeps the singular value decomposition (SVD)-derived pseudo-null subspace aligned with the physical residual space without explicit knowledge of H. Experiments on IEEE 14-, 30-, 57-, and 118-bus systems confirm this principle in practice: stealthy attacks that evade XTM, LSTM, AE and Isolation Forest baselines appear as clear deviations in the aligned subspace, yielding higher F1-score and detection accuracy while remaining robust under partial observability and realistic PMU noise.
Read the full abstract ▾
All authors Xin Li, Chenhan Xiao, Jonathan Cohen, Aviad Elyashar, Yang Weng, Rami Puzis
Machine Learning
The Confidence Trap: Calibration Attacks for Graph Neural Networks
Cuong Dang, Jiahao Zhang, Hieu Ta Quang et al. (+3 more)
While confidence calibration is essential for trustworthy decision-making in safety-critical applications, the robustness of calibrated GNNs to
adversarial structural perturbations remains largely unexplored. However, studying calibration attacks on graphs presents unique technical challenges: (1) the discrete nature of graph structures complicates gradient-based optimization, (2) existing underconfidence objectives fail to drive predictions toward uniform distributions, and (3) GNNs are highly sensitive to edge perturbations, often causing unintended label changes that violate attack constraints. To address these challenges, we propose a Unified Graph Calibration Attack (UGCA) framework designed for worst-case (white-box) analysis of GNN calibration robustness. UGCA introduces a KL-divergence loss to encourage uniform predictive distributions, a reranking mechanism to reduce label flipping, a hybrid loss to recover labels when violations occur, and beam search to explore a broader
adversarial search space. We further provide theoretical insights linking model generalization,
dataset complexity, and calibration vulnerability, showing that models with higher accuracy or trained on
datasets with more classes are more susceptible under this threat model. Extensive experiments demonstrate that UGCA substantially increases Expected Calibration Error while preserving classification accuracy. Our code is publicly available at https://github.com/CaptainCuong/Graph-Calibration-Attack.git.
Read the full abstract ▾
All authors Cuong Dang, Jiahao Zhang, Hieu Ta Quang, Dung Le, Lu Cheng, Suhang Wang
Machine Learning · Cross-listed from stat.ML
LOTTERY: Learning from Reference-Only Samples in Two-Sample Testing under Size Asymmetry
Xunye Tian, Zhijian Zhou, Liuhua Peng et al. (+1 more)
Data-adaptive two-sample testing assesses if two samples come from the same distribution, using a discrepancy learned from the data (e.g., via kernel-based feature representations). Such methods typically rely on data splitting to decouple learning from testing and control type I error. However, this paradigm is ill-suited to few-shot settings with severe sample-size imbalance: abundant reference samples are available, while only a handful of query samples arrive. In this paper, we show how this imbalance can be leveraged constructively. Using abundant reference data, we learn reference-dependent representations that summarize salient structure of the reference distribution and provide informative signals for detecting departures. We incorporate a collection of representation families that capture both global and local structure, and adaptively weight them using only reference samples via an uncertainty-guided principle. Theoretically, we establish permutation-based type I error control and show consistency of the aggregated test: as the sample sizes grow, the test power converges to one whenever the representation set contains at least one consistent representation. Empirically, our aggregation achieves strong performance across a range of
benchmarks while retaining type I error control.
Read the full abstract ▾
All authors Xunye Tian, Zhijian Zhou, Liuhua Peng, Feng Liu
Machine Learning
Beyond Linear Activation Steering: Invertible Latent Transformations for Controlling LLM Behavior
Tuc Nguyen and Thai Le
Activation steering provides a lightweight inference-time mechanism for controlling large
language models (LLMs) by modifying their internal activation vectors toward desired behaviors. Most existing methods compute a fixed steering direction in the original activation space, typically from pairs of contrastive examples using mean differences, linear probes, or arbitrary separability criteria. While effective to a certain extent, these methods treat behavioral control as a global, linear, additive offset: the same direction is applied across inputs, and behaviors are linearly separable. This can be restrictive when behavioral features vary nonlinearly across the activation space or lie on curved and anisotropic
manifolds, where the optimal intervention may be input-dependent. To address this limitation, we propose INNSteer, a nonlinear activation steering framework based on invertible latent transformations. Rather than searching for a better steering vector in the original representation space, INNSteer learns a lightweight invertible neural network $\phi$ that maps an LLM's activations into a latent space where behavioral classes are more amenable to linear control. At inference time, activations are mapped through $\phi$, steered in the latent space, and mapped back through the exact inverse transformation $\phi^{-1}$. This makes a simple latent-space translation become a nonlinear, input-dependent intervention in the original activation space. Across experiment settings on multiple LLM families, scales, behavioral traits, and safety
benchmarks, INNSteer consistently improves model control over linear, transport-based, and nonlinear steering baselines while largely preserving generation fluency.
Read the full abstract ▾
Machine Learning
Theoretical Foundations of Continual Learning via Drift-Plus-Penalty
Nazreen Shah, Govinda Arya, Bharath B.N. et al. (+1 more)
In many real-world settings, data streams are nonstationary and arrive sequentially, requiring learning systems to adapt continuously without retraining from scratch. Continual learning (CL) addresses this challenge by incorporating new tasks while mitigating catastrophic forgetting, where learning new information degrades performance on previously acquired knowledge. We introduce a control-theoretic perspective on CL that explicitly regulates the
evolution of forgetting, framing adaptation as a controlled process subject to long-term stability constraints. We focus on replay-based CL, where a finite
memory buffer stores representative samples from prior tasks. We propose COntinual Learning with Drift-Plus-Penalty (COLD), a continual learning framework based on the Drift-Plus-Penalty (DPP) principle from
stochastic optimization. To facilitate analysis, we also consider an oracle variant, COLD-ORACLE, as a reference
benchmark. At each task, both methods minimize the current task loss while maintaining a virtual queue that tracks deviations from long-term stability on previously learned tasks, capturing the stability-plasticity trade-off as a regulated dynamical process. We establish stability and convergence guarantees that characterize this trade-off through a tunable control parameter. Experiments on standard
benchmarks demonstrate that COLD consistently outperforms a broad range of state-of-the-art CL methods while providing competitive and controllable forgetting behavior through explicit regulation of stability and plasticity.
Read the full abstract ▾
All authors Nazreen Shah, Govinda Arya, Bharath B.N., Ranjitha Prasad
Machine Learning
Not Just After One: Sleep-Inspired Replay Prevents Catastrophic Forgetting After Sequential Tasks
Anthony Bazhenov, Jean Erik Delanois and Giri P. Krishnan
One of the critical limitations of artificial neural networks is their lack of ability to continually learn: training on new tasks often leads to interference and forgetting of the previous ones. While several
algorithms have been proposed to protect old
memories from interference, they are typically applied during or immediately after each new episode of training. In contrast, humans and animals can learn continuously, acquiring multiple new
memories during active learning before consolidating all of them into long-term storage. Here we show that multiple new tasks can be trained sequentially before an unsupervised sleep-like replay phase is applied to partially restore performance across all previously learned tasks. Our study further suggests that task-specific information remains resilient to new training but decays gradually as network is trained on new tasks. These findings point to novel principles for developing a broad range of continual learning AI solutions.
Read the full abstract ▾
Machine Learning
Sparrow: Sparse Rollout for Stable and Efficient Long-context RL of Large Language Models
Yang Zhou, Ranajoy Sadhukhan, Zhaofeng Sun et al. (+7 more)
Despite being powerful, reinforcement learning with verifiable rewards (RLVR) induces extremely long COT, making it computationally expensive. Since RLVR per-step cost is dominated by long-context rollout generation, sparse
attention offers a promising way to accelerate dense rollout. However, sparse rollouts require a delicate stability-efficiency tradeoff: overly aggressive sparsity causes collapse, while overly lenient sparsity gives insufficient speedup. In this work, we study this tradeoff through sparse-to-dense actor-policy mismatch. We first observe that sparse rollout collapse is not driven by uniform degradation across
tokens: most sparse
tokens align perfectly with dense even under aggressive sparsity. Motivated by this, we hypothesize that sparse rollout training remains stable if the lower tail of per-
token actor-policy mismatch stays above a critical threshold throughout the trajectory. We introduce a dynamic sparsity schedule that keeps this tail statistic constant during generation and validate our hypothesis. Across Qwen3 thinking-family models, keeping the tail mismatch statistic near a consistent threshold generally enables stable training. We then use a cost model to find the sparsity schedule for maximum speedup under this mismatch threshold, achieving 2.2x, 2.4x, and 2.0x rollout speedups when training Qwen3-1.7B, Qwen3-4B, and Qwen3-8B. Empirically, we show the thresholds generalize to a larger model (Qwen3-14B) and another RL domain (coding). Finally, our analysis naturally motivates DistillSparse: lightweight LoRA-based
distillation on sparse rollout lets more aggressive sparsity reach the same sparse-to-dense mismatch threshold, yielding higher speedup.
Read the full abstract ▾
All authors Yang Zhou, Ranajoy Sadhukhan, Zhaofeng Sun, Zhuoming Chen, Souvik Kundu, Saket Dingliwal, Sai Muralidhar Jayanthi, Aram Galstyan, Haizhong Zheng, Beidi Chen
Machine Learning · Cross-listed from stat.ML
Improving Bayesian Optimization via Training-Aware Conditional Diffusion Models
Yilin Zheng, Haowei Wang, Szu Hui Ng et al. (+1 more)
Bayesian optimization (BO) is a widely used approach for black-box optimization that uses a Gaussian process (GP) as a surrogate and guides sequential evaluations via an acquisition function, with the ultimate goal of locating the global optimum $\mathbf{x}^{\star}$. To align with this goal, information-based acquisition functions such as Predictive
Entropy Search (PES) model $\mathbf{x}^{\star}$ as a random variable and reduce the
entropy of its distribution, but approximating this distribution via traditional GP posterior sampling is computationally expensive. To address this limitation, we leverage Conditional
Diffusion Models (CDMs) to efficiently approximate the distribution of $\mathbf{x}^{\star}$ and develop BO-inherent training strategies for CDMs. Motivated by the structural properties of the CDM-learned distribution, we further develop an acquisition strategy termed
Diffusion-based Mode Seeking (DMS) to guide the sequential evaluation. We establish a sub-optimality guarantee for the CDM-learned distribution and demonstrate through extensive experiments that DMS outperforms standard BO baselines.
Read the full abstract ▾
All authors Yilin Zheng, Haowei Wang, Szu Hui Ng, Enlu Zhou
Machine Learning
Provably Efficient Personalized Multi-Objective Bandits with Proactive Conversational Queries
Linfeng Cao, Ming Shi and Ness B. Shroff
Personalized decision-making in multi-objective bandits requires learning user-specific trade-offs among competing objectives. Since arm utility depends on both unknown rewards and unknown preferences, existing methods infer preferences only from utility feedback, entangling preference learning with reward exploration. In practice, however, users often reveal their priorities through proactive conversational queries (e.g., "cheap and clean hotel"), yet this structured signal is not leveraged. We formalize a proactive query-based framework in which user queries provide structured preference signals. Modeling these signals via a Plackett-Luce subset choice model, we show that query-only learning is insufficient due to a fundamental shift-invariance barrier. To resolve this, we introduce MO-PQUCB, a hybrid
algorithm that integrates query-based preference anchoring with bandit feedback through shift-invariant regularization and dual-exploration UCB. We prove that proactive queries accelerate preference estimation and yield improved regret scaling over prior preference-aware MO-MAB methods. Under corrupted queries, we further characterize statistical limits and design a robust
estimator achieving near-optimal performance when the corruption is sparse. Experiments validate both theoretical and practical gains.
Read the full abstract ▾
Machine Learning
When Are Neural Interaction Discoveries Real? Identifiability, Recoverability, and a Pre-Fit Diagnostic
Valentina Kuskova, Dmitry Zaytsev and Michael Coppedge
When a neural time-series model reports that one variable modulates another's effect on a target, is the discovered interaction a property of the data or an artifact of model flexibility? We argue that this is fundamentally a question of identifiability, governed by the geometry of the observed input support rather than by the specific neural architecture. We study the problem in a multiplicative-gating extension of neural additive vector autoregression (GNAVAR), in which source contributions are modulated by other lagged variables. We show that representational capacity is not identifiability: dependent inputs induce leakage between edge-specific interaction terms, and low-dimensional support permits distinct interaction decompositions that agree on the observed data while differing elsewhere. We then prove a population identifiability
theorem for normalized minimal GNAVAR decompositions under explicit support conditions, including settings with shared modulators. The theory yields a simple practitioner-facing diagnostic: the effective rank of the joint lag-block covariance predicts, before fitting, whether interaction recovery is feasible for a given candidate set. When the candidate set is unknown, a two-seed stability check provides a practical operational test. The same support condition organizes empirical outcomes into the three states predicted by the theory. Our results show that interaction recoverability depends on support geometry, that effective rank provides a practical pre-fit diagnostic, and that instability across independent fits is a characteristic signature of non-identifiable interaction discovery. The identifiability phenomenon, the support condition, and the instability signature are model-agnostic; GNAVAR is the vehicle that makes them provable.
Read the full abstract ▾
Machine Learning
The Spectral Dynamics and Noise Geometry of Muon
Pierfrancesco Beneventano, Mahmoud Abdelmoneum and Tomaso Poggio
Muon replaces a
matrix gradient $G=U\Sigma V^\top$ by its polar factor $UV^\top$. This keeps the singular directions selected by the gradient, but makes the update spectrum flat. We study the optimization bias created by this operation. Under explicit
alignment assumptions, we prove that the polar update is the one-step
entropy-maximizing choice among bounded updates that use the gradient singular directions and do not adapt to the current weight spectrum. In an underdetermined regression model, we derive exact singular-value dynamics for continuous-time
Muon and identify a measurement-dependent condition under which the normalized spectrum moves toward equal nonzero singular values. This geometry also rules out a common low-rank interpretation: at fixed Frobenius norm,
Muon's distinguished state has a flat spectrum, whereas
nuclear-norm minimization favors spectral concentration. Controlled
matrix-sensing experiments separate the effect from simple gradient rescaling, show that norm-matched gradient descent does not reproduce
Muon, and recover the predicted flattening trend across broad ablations. In small NanoGPT pretraining,
Muon preserves stable rank, has a broad learning-rate plateau, and improves validation loss relative to AdamW; in a matched small-ViT control, the ranking reverses. The resulting picture is regime-dependent:
Muon is not universally superior, but its flat-spectrum bias can help when many spectral directions need to remain active.
Read the full abstract ▾
Machine Learning
STAR-KV: Low-Rank KV Cache Compression via Soft Thresholding for Adaptive Rank Control
Priyansh Bhatnagar, Ashkan Moradifirouzabadi, Se-Hyun Yang et al. (+3 more)
Low-rank projection has emerged as a promising approach for compressing the KV cache by exploiting hidden-dimension redundancy. However, prior methods rely on fixed or heuristic rank selection and struggle to achieve aggressive compression with minimal accuracy degradation. We propose
STAR-KV, an adaptive low-rank KV cache compression framework with fine-grained rank control.
STAR-KV encompasses 1) a differentiable thresholding mechanism that enables optimal rank selection at both
attention-head and block levels, 2) a hybrid decomposition strategy that applies different low-rank factorizations according to the sensitivity of key and value projections, and 3) a low-rank-aware mixed precision quantization that leverages data statistics for near lossless low-bit quantization. Evaluated across multiple LLMs and
benchmarks,
STAR-KV achieves up to 75% KV cache compression and up to 20x overall KV cache reduction when combined with quantization. Enabled by custom Triton-based GPU kernels,
STAR-KV delivers up to 6.9x speedup for the
attention module and 3.1x end-to-end generation throughput. Our code is publicly available at: https://github.com/PriyanshBhatnagar/
STAR-KV.
Read the full abstract ▾
All authors Priyansh Bhatnagar, Ashkan Moradifirouzabadi, Se-Hyun Yang, SeungJae Lee, Jungwook Choi, Mingu Kang