This paper tackles handwritten Bangla word recognition by casting it as character-level object detection rather than whole-word classification. The authors ensemble SSD-MobileNetV2 and Faster R-CNN-InceptionResNetV2 to spot individual characters and diacritics in unsegmented word images, then rearrange the predictions into text. They introduce a modified Non-Maximum Suppression and a heuristic called Preserving Necessary Overlap (PNO) to cope with the cursive matra line and overlapping diacritics that make Bangla OCR notoriously difficult.

LLMs can hit high accuracy on hard math problems while lacking genuine structural insight. This paper proposes the Mathematical Primitive---a concise, non-procedural statement of the key idea that makes a solution possible---and introduces Prim, a benchmark that disentangles structural understanding from procedural execution across Discovery, Generation, Digestion, and Execution. The authors show that independent Discovery is the dominant bottleneck and introduce Absorb, a primitive-privileged self-distillation framework that selectively transfers structural guidance into the student model without requiring primitives at inference time.

CARM targets off-policy drift in LLM reinforcement learning by preventing signed token log-ratios from canceling in sequence-level response masks. The core idea is to average absolute log-ratios rather than signed ones, which stops bidirectional policy drift from being hidden by the geometric mean. It also proves a joint bound on the fraction and severity of out-of-band token ratios for any accepted response. The method is evaluated on Qwen3.5 models and shows consistent gains over geometric-mean and token-level baselines.

Mechanistic interpretability tries to identify sparse circuit subgraphs that explain model behavior, typically evaluating candidates via intervention-defined faithfulness. This paper demonstrates that faithfulness objectives can prefer an equally sized circuit that reproduces intact-model behavior worse than an alternative, because ablating excluded components distorts the computational context of retained ones. The authors argue this creates an objective-level recovery gap: better discovery algorithms cannot recover true mechanisms if the evaluation metric itself systematically rewards the wrong candidate.

While 3D Gaussian avatars render photorealistic humans efficiently, their animation pipelines are often throttled by costly per-frame neural inference. This paper introduces GALA, which shows that the animation outputs of diverse pretrained avatar networks can be closely approximated by an identity-shared linear blendshape basis extracted via distillation, coupled with a shallow coefficient predictor. By constructing the basis with block-local PCA under a rendering-aware metric and a strict memory budget, the method reduces CPU animation cost by up to three orders of magnitude and enables 60 fps on mobile devices with only modest quality loss.

Most trajectory generators require check-ins from the deployment city, which is unrealistic when only a public POI map is available. This work studies target-trajectory-free generation: synthesizing human mobility in a new city using only source trajectories and the target city's POI coordinates and categories. Its key insight is to replace absolute destination prediction with transferable relative transitions—semantic and geographic displacements between place contexts—and ground them onto the target POI graph via a behavior graph and an exploration–return walk. This factorization yields a theoretically principled and empirically evaluated method for zero-shot mobility synthesis.

The paper asks whether stochastic rounding (SR) can improve low-precision transformer inference over round-to-nearest (RN), and shows that the benefit is site-dependent: SR outperforms RN in long MLP reductions because its $O(\sqrt{n}u)$ probabilistic error envelope beats the $O(nu)$ deterministic drift of RN, while RN is superior at the language-model head because SR's non-uniform variance is heavily penalized by the Fisher-weighted cross-entropy. The authors support this with a fast variable-precision emulation tool (PRISM), a vector Bienaymé–Chebyshev bound, a second-order loss decomposition, and experiments on DistilGPT-2 that motivate mixed-precision rounding recipes.

Training-data selection for LLMs is usually heuristic or sample-specific, which prevents generalization to unseen data. This paper tackles the harder problem of learning a shared meta-network that scores any example, and shows that plugging such a network into the standard bilevel objective (ScaleBiO) causes unstable optimization and poor transfer. The authors trace the failure to two properties of the ScaleBiO loss—competitive weight suppression and easy-feature dominance—and propose TESS, a method that replaces the bilevel objective with a simple pointwise MSE regression on pseudo-labels derived from loss gaps. TESS achieves strong transfer across datasets, corpus scales, and model sizes while cutting peak memory by over 20%.

Recent reasoning models show impressive capabilities thanks to RLVR post-training, but their chain-of-thought traces increasingly drift into unusual, non-standard language. This paper formally proves that RLVR optimization allows unbounded language drift while supervised fine-tuning imposes a hard finite bound, and shows empirically that such drift arises specifically when RLVR must discover novel behaviors on tasks beyond the base model's existing capabilities. The results suggest a fundamental tension: improving chain-of-thought monitorability by constraining drift necessarily limits the expected reward a model can achieve.

This paper studies gradient descent with a large constant stepsize on linearly separable logistic regression, a canonical model for the Edge of Stability phenomenon. It proves that the worst-case exact transition time from oscillatory loss to monotone decrease scales as $\Theta((\log \eta)^{\min\{n-2,d-2\}})$ uniformly over dimension $d \ge 2$, disproving a conjecture that stepsize independence holds beyond $d=2$. The result translates into substantially stronger finite-horizon optimization guarantees by permitting exponentially larger constant stepsizes than prior $O(\eta)$ transient bounds.

This paper tackles the tendency of graph diffusion models to lose global topology—such as cycles and higher-order motifs—during local denoising. It introduces MG-Diff, a framework that extracts a discrete Morse skeleton from a graph and conditions generation on this sparse topological backbone. By framing diffusion as traversing a discrete cobordism, the authors aim to anchor the generative trajectory to structurally critical events rather than treating topology as an afterthought. The promise is better preservation of connectivity and improved uncertainty quantification in spatio-temporal forecasting.

Federated learning suffers from client drift when local data is heterogeneous, slowing convergence and degrading the global model. This paper proposes Activation Sharing Federated Learning (ASFL), a method in which clients exchange hidden-layer activations and labels with a randomly paired peer each round, then perform a partial tail round updating only the deeper layers using the pooled activations before standard local training. The premise is that aligning latent representations directly across clients mitigates heterogeneity more effectively than parameter-space regularization, while partial backpropagation cuts per-round computation.

This paper develops a physics-informed Kolmogorov-Arnold network (KAN) solver for free-boundary PDEs, including elliptic obstacle problems and the time-dependent Stefan problem. The core idea is to replace fixed MLP node activations with learnable univariate B-splines on network edges, aiming for localized resolution of contact sets and moving interfaces. Accurate, parameter-efficient approximation of these coupled unknowns matters because classical mesh-based methods require explicit interface tracking or costly refinement, while standard PINNs may blur sharp free boundaries.

Test-time scaling is often proposed as a cheaper substitute for larger models, but its practical returns remain murky because prior evaluations usually focus on one domain and rarely charge selection costs to the budget. This paper introduces SELF-POT, a cross-domain benchmark and evaluation framework that charges every model call—including selection, critique, and voting—in actual dollars under fixed output-token caps. Across 350 sealed tasks in mathematics, programming, and agentic workflows, the authors show that extra inference yields wildly different returns depending on the domain and selection rule, and that the real bottleneck is often converting generated coverage into a valid submission rather than generating a correct candidate in the first place. By reframing test-time potential as a measured conversion profile tied to price and failure handling, the work connects empirical scaling behavior to concrete deployment trade-offs.

Adapting LLMs at test time usually requires gradients, task labels, or auxiliary demonstrations. This paper studies whether the input query alone contains enough signal to predict a useful parameter update. The authors propose a query-conditioned distributional hypernetwork that learns to output a distribution over LoRA adapters, trained with a differentiable Monte Carlo approximation of the marginal likelihood. They show that the mean update already outperforms deterministic hypernetworks, and that sampling diverse adapted models at test time provides a novel compute-scaling axis competitive with standard token sampling. Results are demonstrated on models from 1B to 4B parameters.

Many recent sampling algorithms built on Wasserstein gradient flows (WGF) and forward-only diffusion processes (FODP) come with exponential-convergence guarantees that are often read as proof of fast practical mixing. This paper argues that guarantee is misleading: the canonical KL-induced WGF and overdamped forward diffusion produce the same marginal Fokker--Planck evolution, so they inherit the same metastability and exponentially slow inter-mode transport known in statistical physics. Using spectral-gap and mean-first-passage-time analysis on multimodal targets, the authors show that well-separated modes induce mixing times that scale exponentially with barrier height, and that standard log-linear annealing does not remove this scaling. The conclusion is that purely local, gradient-driven sampling is structurally limited for multimodal distributions, motivating nonlocal transport mechanisms.

Intrinsically disordered protein regions (IDRs) are essential to cellular function but cannot be designed with structure-based methods because they lack stable folds. This work introduces IDiom, a protein language model pretrained on 54 million predicted IDRs, and RL-SAE, a reinforcement learning method that rewards the activation of interpretable sparse-autoencoder features to control sequence patterns. Together they enable composable, function-targeted generation of disordered sequences, though the results remain purely computational.

Equivariant machine learning interatomic potentials are held back by costly tensor product operations that dominate inference at scale. BranchIP addresses this by introducing lightweight, learned routers inside a single model that adaptively terminate tensor product refinement for individual features, trained with a novel distillation loss. On two realistic reactive systems—CO/Pt catalysis and CsH2PO4 proton conductors—it delivers up to 2.4× inference speedup and 2.6× memory reduction while preserving exact energy conservation and offering chemically interpretable routing patterns.

Rare weather regime transitions such as sudden stratospheric warmings are challenging to emulate because training data is overwhelmingly dominated by stable regimes. This paper develops a probabilistic ResNet-inspired Conditional Variational Autoencoder (CVAE) for the stochastic Holton-Mass model, a 75-dimensional idealized system with two metastable vortex states and rare noise-driven transitions. By training only on one-day-ahead predictions, the authors show that the emulator reproduces both short-term dynamics and demanding long-term statistics, and that unsupervised clustering of its 32-dimensional latent space yields four physically interpretable regime clusters. The results suggest a proof-of-concept pathway toward interpretable early warning systems for extreme events, although the study remains confined to a simplified testbed.

TRACE targets dynamic resource assignment in industrial systems where processing times, energy profiles, and queue capacities are unknown a priori. While existing LLM-based Automatic Heuristic Design learns only from scalar fitness scores, TRACE couples an evolutionary loop with an agentic workflow that distills noisy system logs into insights and executable surrogate models. This feedback turns blind trial-and-error into data-driven design, yielding heuristics that cut cost by over an order of magnitude relative to prior AHD methods and reduce energy by 65% against industry standards without sacrificing reliability.