Decentralized federated learning (DFL) eliminates the central aggregation server but remains vulnerable to data heterogeneity, Byzantine attacks, and channel noise. We show that noise accumulates as models propagate across nodes, becoming increasingly significant as the learning rate decays. Moreover, accumulated noise enables Byzantine attackers to exploit loss-based model selection by mimicking honest model averages. To address these challenges, we propose screened, hop-aware, ranked, proximal averaging, a robust DFL framework (FedSharp). In this approach, each node maintains models from its 2b predecessors, ensuring at least b honest candidates when at most b nodes are Byzantine. FedSharp screens suspicious models using coordinate-wise median deviations, averages the b lowest-loss candidates to reduce fresh channel-noise variance by a factor of b, and applies noise-adaptive gradient-norm regularization and proximal training to mitigate accumulated noise and non-IID drift. Experiments on a 100-node ring with 33 Byzantine nodes show that FedSharp achieves 93.5-94.9% accuracy on MNIST, compared with 29.4-38.1% for a state-of-the-art single-model-selection benchmark, and 32.8-48.8% on CIFAR-10, compared with 10.0-30.5%, across Gaussian and hidden attacks under IID and non-IID data. On MNIST, FedSharp also achieves accuracy within approximately one percentage point of the benchmark operating without channel noise.
In this paper, three operations on Gibbs probability measures are studied: renormalization, normalized log-linear combination, and nesting. The resulting measures are shown to be Gibbs probability measures with explicit expressions for their objective functions, reference measures, and regularization factors. These measures are further characterized as unique solutions to optimization problems involving linear combinations of the original objective functions, subject to either relative entropy regularization or relative entropy constraints. When the reference measures are probability measures, geometric interpretations of these three operations are established involving Pythagorean identities and exponential geodesics.A central result is the equivalence between nesting and normalized log-linear combinations of Gibbs probability measures. These results are then applied to supervised machine learning with Gibbs algorithms, which are represented by Gibbs probability measures. In this context, either through nesting or log-linear combinations, Gibbs algorithms trained on local datasets are used to form new Gibbs algorithms trained upon the aggregation of all local training datasets. Interestingly, the parameters of such operations, determine the weights of each training data point on the aggregated dataset. %Operations on local Gibbs probability measures are shown to control the contributions of training data. This observation is leveraged for achieving: (a) exact machine unlearning in centralized, decentralized, and federated learning systems; and (b) centralized-equivalent performance, in the sense of producing algorithms trained on the aggregation of all local datasets in decentralized and federated learning .
In multivariate conformal prediction, many nonconformity scores yield prediction sets with complex geometry (e.g., ellipsoids or density level sets) that are difficult to interpret and visualize. Hyperrectangles offer a practical alternative by providing one interval per output. We extend this notion to Cartesian products of arbitrary one-dimensional sets, which we call generalized hyperrectangles. We introduce a rectangularization framework that turns any continuous multivariate nonconformity score, or conditional density, into one yielding generalized hyperrectangular prediction sets with finite-sample coverage. We establish key geometric properties of our rectangularization framework and provide useful intuitions on generalized hyperrectangles. We further establish a representation theorem showing that a broad class of hyperrectangular conformal methods based on max-aggregated coordinate-wise scores can be interpreted as rectangularizations of implicit conditional densities. Through synthetic and real-world experiments, we quantify the efficiency cost of rectangularization, while highlighting its benefits, especially for practical applications of uncertainty quantification.
LLMs can repair noisy OCR transcripts of historical documents, but they also change text that was already correct. We propose EditVeto, a veto gate placed after a full-context LLM rewrite. The rewrite is character-aligned to the OCR input, its differences are split into blocks, and a 0.6B-parameter gate keeps or reverts each block. An oracle analysis motivates the design: given the true error locations, per-span generation is worse than no correction, whereas vetoing a full rewrite's edits helps. Trained on rewrites by two Qwen3 models and evaluated, frozen, on the eight HIPE-OCRepair-2026 test sets, the gate improves all five rewriters tested, two of them unseen in training: weighted cMER falls by 0.0026-0.0044 and the preference score rises by 0.15-0.25 (paired bootstrap, p ~ 0.0002 each). On published competition outputs it lowers cMER for three of four teams but raises it for the winning run (0.0050 to 0.0065). In a second, disclosed test look, a retrained gate still improves LoRA-tuned rewriters (cMER -0.0011 to -0.0021), and an ensemble of their gated edits reaches a cMER of 0.0069 and a preference score of 0.928, mostly through fine-tuning and pooling rather than gating. Built only on public models and data, it is behind the winner (0.0050) and tied with the second-ranked run (0.0071), both from the only team using domain pre-training, and ahead of all other published runs. A third, secondary look shows the unchanged gate improving zero-shot Qwen3.8-27B (cMER 0.0114 to 0.0108) but only the preference score of its LoRA-tuned version.
Across a wide range of applications, kernel measures are used to tune hyperparameters, select among candidate kernels, and assess the suitability of kernels for specific learning tasks. For classification, kernel measures are used to predict which kernels may produce accurate kernel classifiers. Many existing kernel measures, such as alignment to the ideal kernel, are designed to reward kernels which captures all samples from the same class as similar. While effective in many settings, this requirement is unnecessarily restrictive for classification, where classes may exhibit substantial internal structure. In this work, the "Confidence Score", a new kernel measure, is proposed. Unlike alignment-to-the-ideal kernel measures, the Confidence Score measures the alignment between an encoding of the ground truth and that ground truth projected by the kernel, allowing kernels that preserve intra-class structure to achieve high scores. Through experiments, it is demonstrated that the Confidence Score exhibits a higher correlation with classification accuracy across a variety of kernels on both real-world and synthetic datasets, compared to select other kernel measures.
In this paper, we provide tight lower bounds for minimizing high-order H\"older-smooth and uniformly convex functions when their orders match. Specifically, we consider functions whose $p^{th}$-order derivatives are H\"older continuous with degree $\nu$ and parameter $H$, and that are uniformly convex with degree $q$ and parameter $\sigma$. When these orders match, namely, $q = p + \nu$, we show that any deterministic algorithm with access to derivatives up to order $p$ requires $\Omega\left(\left(\frac{H}{\sigma}\right)^{\frac{2}{3(p+\nu)-2}}\log\frac{\Delta_0}{\epsilon}\right)$ oracle calls to reduce an initial optimality gap of $\Delta_0$ to $\epsilon$. This matches the corresponding upper bound and establishes a higher-order uniformly convex counterpart to Nesterov's classical first-order lower bound for strongly convex functions. Rather than directly generalizing Nesterov's classical quadratic-chain construction, we develop a multiscale phase-wise construction that preserves previous oracle responses while maintaining global H\"older smoothness and uniform convexity. Our result resolves the matched case left open by existing lower bounds for the two asymmetric regimes.
Sparse autoencoders (SAEs) decompose visual representations into individual features, offering a promising basis for explaining model behavior. Interpreting these features typically relies on images or pixel regions that most strongly activate them. However, this evidence becomes ambiguous when multiple features with different semantics frequently activate in the same locations, such as texture and color in the same patterns. We propose Reconstructive SAE Intervention (RecSAEI), a framework that shifts feature explanation from passive observation to active visual reconstruction. By intervening on representations along SAE decoder directions and reconstructing the visual output, we provide direct evidence of a feature's semantic meaning. We demonstrate that our method successfully disambiguates feature meanings across multiple datasets and vision encoders. Furthermore, our framework grounds [CLS] SAE interventions in a sparse set of patch SAE decoder directions. Beyond stable target response, this attribution also exposes the compositional relationship between elementary visual patterns and high-level semantics. Our results advocate for intervention reconstruction as an essential complementary signal for explaining neural networks.
For humans, referential disambiguation is a collaborative process involving clarification questions asked when they will meaningfully change the outcome. However, large language models (LLMs) have been shown to be biased toward premature answering, preferring definite single-turn responses over more clarification exchanges. In this paper, we study the problem of "strategic clarification" -- asking clarification questions in a reference game when their answers will narrow the set of candidate referents. We construct a controlled test environment from the SIMMC 2.1 dataset that records how many questions each ambiguous reference needs to be successfully resolved, and examine how baseline LLM behavior and common post-training methods ask clarifying questions vs. attempt to answer the referent. We also propose BACE (Belief-Aware Critic with Epistemic tracking), a lightweight action chooser method on top of a frozen LM backbone, and show that BACE more consistently chooses the optimal number of clarifying questions than methods such as GRPO and ArCHer.
Textual information is ubiquitous in real-world time series datasets, appearing as news, reports, and annotations that hold key predictive signals not deducible from numerical history alone. Large Language Models (LLMs) interpret text but lack numerical precision, while Time Series Foundation Models (TSFMs) are quantitatively accurate but cannot reason about the effects of text. Moreover, the lack of large-scale paired multimodal datasets limits progress in bridging this gap. In this work, we address this challenge by providing a scalable approach for generating realistic multimodal datasets by annotating real-world time series with LLMs. Building on this, we propose Context as Covariates (CoCo), a forecasting framework in which an LLM distills its textual reasoning into numerically grounded forecast and confidence covariates that guide a TSFM backbone, thereby exploiting the strengths of both LLMs and TSFMs. We align the two models in stages, first fine-tuning the LLM via Group Relative Policy Optimization (GRPO) with a reward tied to covariate-induced improvement in forecasting performance, and then fine-tuning the TSFM on covariates generated by the aligned LLM. Our experiments show that models trained on our synthetic multimodal corpus generalize to real benchmarks spanning finance, economics, weather, traffic, and security. Notably, CoCo with Qwen3-4B-Instruct-2507 as the LLM and Chronos2 as the TSFM achieves up to 12% MAE improvement over state-of-the-art TSFMs and LLM forecasters, such as GPT-4o. Code, checkpoints, and the evaluation datasets are publicly available.
Humanoid motion tracking has scaled to general policies that reproduce diverse human motion, but the resulting behavior is in free-space: the tracker knows where the body should be, not how it should interact with the environment. Other work makes the object part of the objective, yet abstracts that interaction to contact or geometry, which leaves the demanding cases under-explored. We argue that force is a natural command channel between motion imitation and object interaction, one that specifies not just where to make contact but how hard. We present Ctrl-F, a hierarchical motion-force framework that augments motion tracking with an explicit force command, a 3-D force at each designated contact body such as a hand or a foot. The hierarchical framework comprises: (1) A unified motion-force controller tracks the reference motion while applying the commanded force to the world. (2) A high-level force planner that issues a force command online based on object state and proprioception observations. In simulation and on hardware, Ctrl-F completes loco-manipulation that demands large, sustained interaction force and adjusts its command online as the load changes, both out of reach of a force-blind tracker. On hardware, Ctrl-F demonstrates hand-based dragging, pushing, and lifting, as well as foot-based kicking, successfully manipulating boxes weighing up to 20kg in pushing and kicking tasks where the force-blind baseline fails, while adapting online to additional loads introduced mid-task.
Aligning large language models (LLMs) with several objectives, such as helpfulness, harmlessness, and honesty, typically relies on preference judgment for each objective, growing the annotation cost linearly with the number of objectives. Active preference learning can reduce this cost by selecting the comparisons whose labels are expected to change the policy most. Existing remedies either lack an objective-aware query selection mechanism for the multi-objective setting or regard conflicting preferences as harmful, therefore excluding them from the training data. We introduce OGC-DPO (Objective-conditioned Gradient Coverage DPO), an active learning framework that treats each (response pair, objective) combination as a separate query and measures its informativeness through the policy's own implicit gradient, without relying on an external reward model. OGC-DPO then selects the queries that most expand each objective's gradient coverage. Empirical experiments across varying numbers of objectives demonstrate the effectiveness of OGC-DPO over existing active preference learning methods under the same annotation cost. Further analysis demonstrates that informative conflicts, rather than being uniformly harmful, provide a useful policy training signal.
Instantaneous quantum polynomial-time (IQP) Born machines are attractive quantum generative models, as certain IQP circuit families are conjectured to be hard to sample, yet they can be trained classically. Previously proposed conditional IQP Born machines make the parameters of the diagonal layer condition dependent. We introduce conditional IQP Born machines, in which the condition is encoded in the input state and a single IQP unitary is shared across conditions. Efficient sampling and amplitude-ratio evaluation for the reference state entering the diagonal layer suffice to obtain unbiased classical estimates of a conditional maximum mean discrepancy objective and its gradients. We show that input conditioning can represent conditional families that no diagonally conditioned IQP model with the same number of qubits can reproduce. For the sixteen-qubit Curie-Weiss spin model, a single conditional model captures the magnetization transition at held-out temperatures. Its mean total variation distance is six times lower than that of diagonal conditioning under the same protocol and comparable to that of classical generative models with up to seven times as many parameters. We also train conditional models classically with up to 2048 qubits, which reproduce the mean squared magnetization of the target away from the crossover. These results extend classically trained IQP Born machines to conditional generation.
While discrete tokenization in Unified Multimodal Models (UMMs) establishes strong semantic reasoning, it introduces a strict quantization bottleneck that degrades fine-grained visual details, often resulting in structurally accurate but artifact-heavy images. We introduce FlowShift, a parameter-efficient post-training framework that bypasses this bottleneck by converting pre-trained discrete UMMs into continuous rectified-flow generators. Instead of discarding the backbone's semantic priors, FlowShift replaces discrete visual interfaces with continuous latent projections and applies LoRA to the attention layers. To efficiently boost visual quality without semantic drift, we propose Dual-Objective Velocity Vector Distillation (DOVD), which aligns the model with a high-fidelity continuous teacher and ground-truth continuous trajectories. This approach drastically improves perceptual quality while preserving prompt adherence and reasoning. When applied to leading discrete UMMs, FlowShift significantly improves COCO-30K FID and aesthetic scores, while retaining over 99.5% of multimodal understanding capabilities. Ablations confirm DOVD drives generation quality. Ultimately, FlowShift achieves high-fidelity continuous generation without sacrificing pre-trained semantic grounding.
Learning on boundary representations (B-reps) requires robustness to the pose of a part, since CAD models are not stored in a canonical orientation. Existing B-rep networks either operate on world-frame coordinates and normals, and thus degrade under rotation, or encode each face in a local reference frame. Local frames yield invariance, but reduce inter-face messages to scalars and discard the relative orientation of adjacent faces. We propose SO3brep, an SO(3)-equivariant graph neural network on the B-rep face-adjacency graph. Node features are computed in local frames and mapped into a shared frame by Wigner D-matrices. Clebsch-Gordan tensor products couple these features with the directions between adjacent faces, so that relative orientation is preserved throughout message passing and the output remains SE(3)-invariant. Across three face-segmentation and two classification benchmarks with randomly rotated test parts, SO3brep achieves the highest accuracy among the compared B-rep networks and equivariant point cloud networks, with the fewest parameters. Its advantage is largest with few labeled parts, a setting typical of industrial CAD. Code is available at https://github.com/UGent-CVAMO/so3brep.
Surgical world models have the potential to provide intra-operative guidance, accelerate risk-free training and facilitate transfer of expert surgical maneuvers. However, the use of world models for transferring surgical expertise has been largely unexplored. Moreover, recent approaches in this domain have mainly relied on ground truth kinematics actions or expert annotations for specific tasks, limiting their applicability beyond the narrow range of tasks with ground-truth annotations. In contrast, existing works in the natural domain have shown promising results in learning latent actions solely from unlabeled videos and transferring the learned actions across contexts. We observe that this paradigm is short-sighted for the surgical domain, failing to capture the hierarchical and goal-directed structure of surgical procedures. Thus, we propose a novel multi-step action learning framework to learn context-invariant latent actions that are easily transferable across various surgical contexts. We perform extensive experiments on the publicly available datasets, SurgToolLoc-2022 and PitVis-2023, verifying our approach with both quantitative metrics and qualitative evaluation by surgical experts. We have published our code and the implementation details at https://github.com/bhattarailab/Surgical-Vision-World-Model.
Background: Quantum machine learning (QML) is promoted for clinical prediction, but comparisons with classical models are rarely paired or tested for equivalence. Methods: We compared a quantum-kernel support vector machine (QSVM; 4 qubits, tuned bandwidth) and a variational quantum classifier (VQC) with five tuned classical models across six public clinical datasets. All models used identical four-component inputs and 25 cross-validation splits. Paired AUROC differences were assessed using a corrected resampled t-test, Holm-adjusted equivalence tests (margins: +/-0.02 and +/-0.05), and random-effects pooling. Ablation, noise, finite-shot, and implementation-fault analyses were conducted. We propose QML-Health-READY, an eight-domain reporting framework. Results: Neither quantum model significantly outperformed any classical comparator. Pooled across datasets, QSVM differed from the kernel-matched radial basis function (RBF) SVM by -0.006 AUROC (95% CI: -0.017 to +0.004), equivalent to +/-0.02. VQC differed from the best classical model by -0.007 (95% CI: -0.019 to +0.004). QSVM was equivalent to RBF SVM within +/-0.05 in four of six datasets but was significantly worse than the best classical model on NHANES 2017-18 (-0.076). A classically trivial product-state kernel performed as well as the entangling kernel. With depolarizing noise (p = 0.01) and 1,024 shots per kernel entry, QSVM AUROC decreased by 0.054 (95% CI: 0.036 to 0.071). Untuned bandwidth produced near-chance AUROC (0.48-0.59). Conclusions: Correctly configured quantum classifiers approached but did not exceed tuned classical models. Finite-shot estimation, rather than gate noise, was their main practical cost, while configuration errors could reverse conclusions. QML-Health-READY defines reporting requirements for credible claims.
High-fidelity simulation of dynamical systems is central to scientific discovery and engineering design, yet remains computationally demanding. Here we introduce the Implicit Neural Operator Solver (INOS), which learns the implicit solve and encodes a time integrator. INOS pretrains a neural operator from static equilibrium solutions and reuses it within Newmark-$\beta$ integration. At each time increment, the integrator forms an effective equilibrium problem, the pretrained neural operator infers the next-step solution field, and the Newmark relations advance the dynamic state. Numerical examples in linear and nonlinear elastodynamics demonstrate accurate full-field predictions without dynamic trajectory supervision. Comparative results show that INOS, evaluated zero-shot, achieves higher predictive accuracy than neural solver baselines trained on dynamic simulations. In addition, INOS delivers over 4,600$\times$ speedup in nonlinear simulations over sequential classical finite-element solvers. INOS natively accommodates changes in dynamic conditions and enables reuse of the pretrained operator across spatial and temporal discretizations. Further evaluation demonstrates a stable million-step rollout. These findings establish INOS as a reusable neural implicit solver for accurate, efficient, and robust dynamic modeling.
Gravitational arcs produced by strong gravitational lensing provide a unique approach for investigating dark matter distribution, galaxy cluster mass reconstruction and the structure of the Universe. Automated detection remains challenging because gravitational arcs are rare, often faint and morphologically similar to elongated galaxies.In this paper, we propose \textbf{HLA-Net (Hybrid Lens Arc Network)}, an interpretable multi-stage framework for automatic gravitational arc detection that explicitly exploits both the geometric morphology and physical characteristics of gravitational lensing without requiring manually annotated training datasets.The proposed framework was evaluated on astronomical FITS images containing complex galaxy cluster structures. Starting from 169 initial arc candidates extracted from ordered skeleton representations, the multi-stage verification pipeline reduced the search space to 83 geometrically consistent candidates, while robust circle fitting successfully modelled 113 candidate arc structures. The results demonstrate that HLA-Net effectively suppresses false-positive detections while preserving physically reasonable lens candidates through an interpretable and modular verification strategy.
Spectral graph generation models graphs through the eigenvalues and eigenvectors of their Laplacians, providing a compact and continuous alternative to direct graph construction. However, these eigensystems must retain the structural properties of graph Laplacians. Existing approaches preserve eigenvector orthonormality but do not enforce the Laplacian nullspace constraint, potentially producing eigensystems inconsistent with Laplacian structure. To address this, we introduce Laplacian-Constrained Stiefel Flow Matching (LaSFM), which preserves the nullspace condition alongside eigenvector orthonormality and positive eigenvalues during generation. We characterize the underlying constrained Stiefel manifold and formulate flow matching on it. We further extend LaSFM to property-guided generation, steering samples toward desired graph properties while maintaining these constraints. Experiments demonstrate improved performance on both spectral and graph-level metrics, along with better alignment with target properties.
The RTX 2080 Ti (TU102) ships full second-generation Tensor Cores in hardware, yet the vendor throttles FP16 matrix multiplication at the driver level, leaving usable tensor throughput well below the hardware peak. turing-unlock lifts this restriction with a kernel module that, before the nvidia driver binds the device, writes the FP16 speed selectors (SS0=0x88888888, SS1=0x8) and releases the ACR mutex; published measurements report cuBLAS FP16 matmul throughput rising from 55.7 to 91.4 TFLOPS (about 1.6x). On the same 22G 2080 Ti and under the same from-scratch pre-training protocol as the companion memory ledger (five tiers of Chinese encoder-decoder models, 2,451 steps per tier, 16,384 tokens per step, seed 42), we obtain a before/after throughput account: the small tier (75.7M parameters, 512-wide, 8+8 layers, activation checkpointing on) improves in steady-state throughput from 6,480 to 7,246 tok/s (+11.8%) with wall time cut from 1.77h to 1.61h (9.0% saved); the tiny tier goes from 15,150 to 15,866 tok/s (+4.7%) with wall time savings of 3.8%; the three smaller tiers (pico-micro) gain at most +0.3%; per-tier peak memory stays bit-identical, and final quality metrics are bit-identical or consistent within 10^-3. We conclude that the unlock gain is a function of matrix size, rising monotonically with tier -- imperceptible for pico-micro, visible from tiny upward, and cashed out as a double-digit percentage at small -- and that it is a pure throughput gain that leaves training dynamics untouched.