Peer Revenue Now accepting positions

Papers

Fresh from arXiv, typeset properly. Highlighted words trade on the market.

All of arXiv

Saturday 7 November

Machine Learning

BRAT: Bounded Residual-Aligned Training for Stable Multi-Pass LLM Post-Training

Reinforcement learning for large language model post-training often reuses each batch of sampled rollouts for several gradient passes to amortize the cost of generation. Under such multi-pass reuse, existing objectives fail in different ways: unclipped updates collapse as the policy drifts away from the sampling policy, adding a negative baseline avoids collapse but reinforces every sampled response, and PPO-style ratio clipping cuts off gradients abruptly. We first ask why reuse is fragile. Measuring how well an update estimated from a small rollout group aligns with the KL-regularized target on held-out rollouts of the same prompts, we find that this alignment drops by nearly an order of magnitude from in-sample to held-out rollouts across three base models, suggesting that repeated passes largely overfit the sampled group. Motivated by this finding, we propose Bounded Residual-Aligned Training (BRAT), motivated by a proximal update toward a bounded reweighting of the sampling policy. BRAT assigns each token a gradient weight that decreases linearly with its probability ratio and vanishes at a trust-region boundary, and normalizes group-centered rewards by their maximum magnitude, keeping advantages zero-mean and bounded without extra hyperparameters. With four reuse passes, BRAT avoids training collapse, limits policy drift, and achieves higher average accuracy with lower seed variance on most benchmarks.
Read the full abstract

All authors Xunlan Zhou, Yuzhe Zhao, Xuhong Huang, Feng Chen, Xuanlin Chen, Lang Feng, Bo An

Robotics

DreamVLA: From Generated Human Motions to Humanoid Vision-Language-Action Policies

Training vision-language-action (VLA) policies typically depends on substantial robot demonstration data, making human teleoperation a costly and cumbersome bottleneck, particularly for whole-body humanoid interaction. Reinforcement learning in simulation offers an alternative for scaling this data, but designing task-specific rewards remains challenging. To address this, we propose DreamVLA, which uses generated motion trajectories to guide reinforcement learning (RL) in simulation, converting them into reliable task execution trajectories. These trajectories train VLA models in simulation without new task-specific teleoperation data. Our central contribution is a reference-conditioned DAgger distillation procedure: after a failed learner rollout, we restart the RL expert from a sampled learner state with the same reference suffix and aggregate successful corrective trajectories, addressing the distribution shift left by offline imitation. We evaluate picking from a platform, opening a drawer, pressing a button, and walking and picking from a platform, with ablations of DAgger iterations, residual RL, and generated dataset scale. Project website: ember-center-berkeley.github.io/DreamVLA.
Read the full abstract

All authors Kevin Park, Izat Rasyad, Dvij Kalaria, Shankar Sastry

Statistics Applications

Bayesian Wombling for Tracking Invasive Species

This paper develops Bayesian wombling, or boundary analysis for tracking invasive species. We focus on three invasive species of concern in North America: Lycorma delicatula, Adelges tsugae, and Microstegium vimineum. We aim to demonstrate the applicability of wombling for supporting invasive pest management activities. In spatial science wombling is used widely to track regions of rapid change within the spatial reference domain. We use Gaussian process models for point-referenced spatial data followed by predictive inference on rates of change over curves using line integrals. We extend the wombling model based Bayesian inference framework to feature simple analytic forms that avoid high dimensional quadrature while offering uncertainty quantification over curvilinear spatial boundaries.
Read the full abstract

Machine Learning

Online Imitation Learning for Stabilizing Vlasov–Poisson Plasmas

We study the problem of controlling instability in the Vlasov–Poisson plasma systems, a fundamental challenge for nuclear fusion control. Motivated by the partial observability of plasma states in practice, we investigate imitation learning algorithms that learn from a fully-observable expert controller, but operate under macroscopic measurements. We theoretically establish a separation between online and offline imitation learning — even a small behavior cloning error leads to error compounding that exponentially amplifies instability over time. By way of contrast, the online imitation learning loss can polynomially control the stability of rollout trajectories. We then propose a DAgger-style algorithm that learns the stabilizing controller under partial observability. Simulation results on a 1D Vlasov–Poisson system demonstrate that our algorithm can effectively stabilize plasma instabilities over longer time horizons than behavior cloning.
Read the full abstract

Computer Vision

Interactive Person Retrieval with On-the-Fly Visual Feedback

Traditional text-based person retrieval (TPR) assumes that a witness provides a complete description at once, whereas interactive person retrieval (Inter-PR) refines a partial description through question-answer exchanges. In this setting, the questioner inspects retrieved candidates to decide what evidence to request next, while the answerer sees only the witness-side memory and cannot report candidate-specific visual differences. Existing methods mainly learn offline question orderings, leaving this visual-feedback loop underexplored. To address this limitation, we introduce visual-feedback interactive person retrieval (VFI-PR), a new interactive retrieval paradigm in which retrieved candidates produce visual feedback that becomes evidence for subsequent questioning and re-ranking. To support training and evaluation in this setting, we construct VFI-PEDES, a dataset and evaluation protocol for visual-feedback interaction, and propose VFI-ReID, a framework that learns to use visual feedback during interactive person retrieval. Through the Dynamic Accumulated-evidence Retriever (DAR) and retrieval-aware question optimization, VFI-ReID adapts retrieval and question generation to evolving evidence states, turning visual feedback into complementary evidence for subsequent retrieval. Experiments show that VFI-ReID consistently improves retrieval across VFI-PR interaction stages, demonstrating the utility of visual feedback for interactive person retrieval and outperforming existing Inter-PR baselines.
Read the full abstract

Superconductivity

Multilayer Heater Nano cryotrons: Electrothermal Modeling of Geometry-Dependent Switching

This paper presents a finite-element framework to study how device geometry affects the switching current of multilayer heater nano cryotrons (hTrons). The device comprises a 50 nm Au/5 nm Ti heater, a 100 nm SiO2 spacer, a 23.6 nm NbN channel, and a Si substrate. Two-dimensional electrothermal simulations examine nine heater and channel width combinations using widths of 100, 500, and 1000 nm. Three-dimensional simulations show how heat flows through the layers and spreads across the channel near the heater-channel overlap. Heater electrical losses and heat transport are coupled, whereas the channel switching threshold is obtained separately from a temperature-dependent constitutive relation. Published global temperature measurements support the switching-law calibration, and heater-current measurements assess the resulting device response. The model measurement comparisons show a systematic reduction of switching-current density with heating, but also geometry-dependent discrepancies in the low-drive plateau and transition shape. A common extraction protocol is formulated for heater-current and heater-power requirements at a specified suppression level, while separating fitted device amplitudes from controlled geometry effects. The resulting comparisons connect thermal fields to design metrics while distinguishing calibration from validation and excluding transient-performance claims from the stationary analysis
Read the full abstract

All authors Baibhari Priya Barua, Md Rahatul Islam Udoy, Shamiul Alam, Md Shafayat Hossain, Ahmedullah Aziz

Signal Processing

Missed Interferers in Ray-Traced Radio Maps: A Digital-Twin and Stochastic Geometry Study of Aerial Coverage

Ray-traced digital twins are increasingly used to plan where cellular-connected uncrewed aerial vehicles (UAVs) can fly. In this paper, we build a Sionna RT digital twin around Raleigh, NC, with 18,591 buildings and two overlapping sub-1 GHz operator networks, and study the downlink signal-to-interference-plus-noise ratio (SINR) of UAVs at 40-300 m. We show that the Monte Carlo radio-map solver of Sionna RT misses distant line-of-sight (LOS) interferers on aerial planes. With $10^7$ rays per sector and 5 m cells, it therefore underestimates the coverage at low SINR thresholds and overestimates it at high thresholds, by up to 26 percentage points at $\pm$2 dB and 50-100 m. Using a LOS-aware deterministic engine validated with Sionna's path solver, we then develop a stochastic geometry model with strongest-sector association and own-site leakage that matches the engine without fitting. The analysis shows that the interference grows logarithmically with the modeled network radius, so a twin that omits distant LOS sites overestimates the coverage, and that altitude mainly reduces the coverage at high SINR thresholds. Finally, selecting the better of two operators raises the minimum SINR along the best corridor by 1.0-3.9 dB up to 100 m.
Read the full abstract

Machine Learning

The Surprising Effectiveness of Deleting Weights in LLM Post-Training

How little of a pretrained large language model (LLM) has to change for it to acquire new capabilities? We find that zeroing fewer than 0.05% of its weights, with no other modification, matches full fine-tuning (FFT) at scale across three LLM post-training regimes: reinforcement learning with verifiable rewards, supervised fine-tuning, and on-policy distillation. Our method, Deletion Tuning, learns a binary mask over the gate projection of transformer blocks. Two practical advantages follow: the trained mask is hundreds to thousands of times smaller than the corresponding FFT checkpoint, and Deletion Tuning forgets far less, which addresses a central challenge in continual learning. Under reinforcement learning, its prior-task accuracy stays near the base level while FFT, LoRA, and random sparse fine-tuning lose up to 5.4 points. To understand why masking alone suffices, we analyze the geometry of the learned weight changes under reinforcement learning using stable rank, spectrum drift, and principal-weight overlap. Compared with LoRA and random sparse fine-tuning, Deletion Tuning is closest to FFT across the three measures.
Read the full abstract

All authors Jack Lu, Zhenbang Yang, Mike Lasby, Tejas Pote, Yani Ioannou, Mengye Ren

Numerical Analysis

Neural Operator Surrogates for Landweber Iteration in Solving Inverse Problems

Neural operators can accurately approximate well-posed mappings, but for unbounded or discontinuous operators, common in inverse problems, we verify that their network parameters often blow up in architectures such as DeepONet and FNO. To solve inverse problems stably, we use Landweber iteration, which requires only a forward map and the adjoint of its derivative (linearisation). To reduce repeated PDE forward evaluations, we replace the forward operator and its derivative with a differentiable neural-operator surrogate. We show that encoder-network-decoder operators, including generalised DeepONets and fully coupled Fourier operators, approximate both the forward map and its derivative. A modified discrepancy principle ensures decreasing error before finite stopping. With a compact preconditioner, the iterates remain in a compact set where accurate surrogates exist, and the stopped reconstructions converge as data and surrogate errors vanish under a regularity condition. We also relate the theory to derivative-informed training along finite directions, showing that the preconditioner determines which derivative errors affect the Landweber iteration update. Experiments on identifying a reaction coefficient in an elliptic equation show that smoothing reduces late-iteration deterioration for less accurate surrogates but slows sharp-feature recovery, and that smaller derivative errors along training directions do not by themselves improve reconstruction accuracy.
Read the full abstract

All authors Kuo Gai, Marco A Iglesias, Shuai Lu, Frank Werner

Machine Learning

Stochastic Port-Hamiltonian Gaussian Processes for Learning from Noisy Dynamics on Lie Groups

Learning predictive models for dynamics is difficult because observations are often sparse and noisy, while the underlying dynamics evolve on nonlinear configuration manifolds and are affected by environmental dissipation and stochastic disturbances. Standard neural dynamics models can fit short-term trajectories but often violate physical structure, accumulate error over long horizons, or rely on parameterizations that introduce singularities such as gimbal lock. We introduce LieSPHGP, a stochastic Gaussian-process framework for learning rigid body dynamics with port-Hamiltonian structure and stability-aware robustness. LieSPHGP represents motion on Lie groups and Lie algebras, enabling coordinate-consistent optimization without singular parameterizations. Learned mass, potential energy, dissipation, and control components retain physical roles and a separate diffusion term represents stochastic forcing. Each component is modeled by Matérn kernel Gaussian processes which capture uncertainty induced by noisy observations. To improve physical validity, Lie group exponential map updates preserve rotational constraints, while positive-definite damping ensures nonnegative deterministic energy dissipation. The implicit-explicit integrator additionally provides unconditional stability for the frozen linear damping subsystem. The resulting model learns predictive dynamics that are robust to noise, respect rigid body geometry, and capture both intrinsic motion and environmental dissipation. Experiments on various noisy rigid body benchmarks show improved prediction accuracy, stronger long-horizon stability, and better satisfaction of manifold constraints compared with neural dynamics baselines. Code, data and interactive results are available on the project website: https://thribhuvanrapolu.github.io/liesphgp/
Read the full abstract

Statistical Methodology

Weak-Instrument Inference and Model Diagnostics in Double Machine Learning for Linear Instrumental Variable Models

We study weak-instrument-robust inference in partially linear instrumental-variables models with learned nuisances. Building on an i.i.d. uniform-size baseline, we establish uniform asymptotic validity of clustered cross-fitted Anderson-Rubin inference and cross-fitted score inference under the stated sampling and nuisance-rate conditions. Extensions establish Anderson-Rubin validity for survey-weighted target laws and fold-nested generated instruments. A supporting population benchmark shows that prediction-optimal instrument residualization maximizes concentration when the treatment nuisance is correctly specified. A five-design diagnostic portfolio, a known-truth Monte Carlo, and an exploratory selector examine how nuisance specification and estimation affect realized strength and inference.
Read the full abstract

Combinatorics

Spectral distribution for biased orientations of random graphs

We study the Hermitian adjacency matrix of the second kind for biased orientations of Erdos-Renyi random graphs. In the model $D(n,p,p')$, an edge is present with probability $p$ and, conditional on its presence, is directed from its smaller to its larger endpoint with probability $p'$. Both parameters may depend on $n$, and biased tournaments are the special case $D(n,1,p')$. We prove a global semicircle law whenever $n(p(1-p)+3p^2p'(1-p')) \to \infty$, for both the centered and the uncentered matrix, and also for the expected empirical spectral measure. Our second theorem gives the exact spectrum of the mean matrix and the limits of the extreme eigenvalues on the scale $np$, under a sufficient density assumption. The main difference from the unbiased random orientations (where $p'=1/2$) is emergence of multiple large eigenvalues. The proofs include the closed-walk counting, control of the moment variances, an explicit diagonalization of the deterministic bias, and the passage from centered to uncentered spectral measures.
Read the full abstract

Machine Learning

SWXAI: A Unified Multimodal Explainability Library for Space Weather Forecasting and Heliophysics

Space weather forecasting and heliophysics increasingly rely on machine learning models that process heterogeneous inputs such as multi-viewpoint multi-wavelength solar images and rasters, multivariate time series, spatial annotations, and tabular event-level data. However, explainability workflows remain fragmented across libraries, model interfaces, and output formats, limiting reuse and reproducibility. We present SWXAI, an open-source library that unifies explainable AI (XAI) workflows for image, time-series, and tabular data in space weather forecasting and heliophysics. SWXAI provides a common, extensible interface across commonly used vision, time-series, and tabular learning models, with validation, adaptation, and persistent explanation objects. It also includes domain-grounded examples, reproducible notebooks, tests, and documentation, enabling users to inspect outputs, swap models or explainers, and extend workflows to new heliophysics datasets. SWXAI reduces modality-specific engineering overhead and supports comparable, reusable explanations across heterogeneous scientific machine learning pipelines. A benchmark against direct backend implementations shows that SWXAI substantially reduces the code users write to produce an explanation, with no practical runtime cost.
Read the full abstract

All authors Benedict Antonio Mervyn, Anli Ji, Manolis K. Georgoulis, Berkay Aydin

Machine Learning

Structurally Separated DAG Learning with Multi-Scale Normalized Closure

Continuous DAG learning commonly enforces acyclicity through scalar functions of a weighted adjacency that aggregate evidence from weighted closed walks. We add a complementary layer: matrix-valued closure logits—explicit multi-hop structural relation scores for each ordered pair—supervised alongside data fit, sparsity, and a DAGMA-style barrier on the weighted adjacency. The distinction is not scalarization of the overall objective but how path evidence is represented and exposed before aggregation. SMNC (scale-normalized multi-scale closure) builds these scores on a structural gate by repeated squaring in log space and divides each scale by its path-length budget, counteracting product-style attenuation along long cycles; diagonal entries support node-wise cycle supervision and off-diagonals optional pairwise reachability cues. In the linear model, a bounded structure–magnitude factorization ties strong effective edges to strong gates; a nonlinear induced gate offers another construction with the same intent. For sufficiently strong cycles, normalized logits admit a length-independent lower bound; under a unique-active-cycle assumption, the hard final-scale readout yields a polynomial log-gate gradient factor rather than exponential product-style decay. On cue-free, unstandardized Gaussian linear and nonlinear ER benchmarks, these models improve over DAGMA, run with default hyperparameters at a reduced budget, on denser graphs and are close to it on sparse ones.
Read the full abstract

Representation Theory

Eaton Triples, Polar Representations, and Lie Triple Systems

An Eaton triple consists of a finite dimensional real inner product space, a closed subgroup of its orthogonal group, and a closed convex cone that meets every orbit and satisfies an extremal inner product inequality. In the literature on group majorization and invariant norms it is a standing hypothesis that the triple admits a reduced triple, obtained by restriction to the span of the cone. We prove that this is always the case: the axioms force the span of the cone to be a section of a polar representation, and Gichev's orbit intersection theorem applies. The reduced group is a finite reflection group, crystallographic when the group is connected; Eaton cones are its closed chambers, unique up to the group action; and a compact linear group admits an Eaton cone if and only if its representation is polar with a finite reflection group as Weyl group. For connected groups we construct a Lie triple system on the space, determined by the inner product and the orbits alone, for which the cone is a Weyl chamber; unlike Dadok's orbit equivalence with a symmetric space, it involves no choices.
Read the full abstract

Computer Vision

Beyond Bit Matching: Orthogonal Watermarks for Collusion-Resistant Image Fingerprinting

Image fingerprinting assigns each distributed copy a user-specific watermark for source tracing. Averaging collusion is a central threat: several recipients can average their differently marked copies to suppress each individual mark. Discrete bit-string watermarks are especially vulnerable because averaging drives bit evidence toward ambiguous decisions. We introduce VectorMark, which maps user keys to near-orthogonal high-dimensional unit vectors and identifies colluders by cosine similarity to the extracted watermark direction. Under ideal averaging of $K$ vector keys, the normalized average preserves a cosine signal of order $1/\sqrt{K}$ for each colluder, while unrelated keys remain concentrated near zero. The spherical geometry of random directions gives a shifted-Beta null distribution for cosine similarities, enabling analytic per-key false-positive-rate control after validation on unwatermarked images. VectorMark uses a neural watermark encoder and centered extractor trained with JND masking and a progressive curriculum for photometric, geometric, and collusion robustness. Experiments on MS-COCO, OpenImages, and AI-generated images show that VectorMark maintains robust watermark attribution under photometric and geometric distortions, and achieves the best strict all-colluder detection among evaluated methods, while preserving high visual quality. Our code is publicly available: https://github.com/nurlanov-zh/VectorMark
Read the full abstract

Artificial Intelligence

Polynomial Denoisers on the Image Manifold: Volterra Networks for Self-Contained Pixel Diffusion

Natural images are often conjectured to lie on a low-dimensional manifold within the high-dimensional pixel space. Recent work leverages this to build self-contained pixel-space diffusion models without specialized machinery or pretrained components. A plain Transformer on raw image patches suffices, provided it predicts the clean image rather than the noise and compresses each patch through an aggressive bottleneck. The mixing operator, however, has remained softmax attention, inherited from sequence modeling rather than derived from denoising. For Gaussian noise, the optimal denoiser admits a canonical expansion in polynomial functionals of the observation, the Wiener–Volterra series. We prove that softmax attention, and any normalized mixing rule (including linear attention), cannot realize any polynomial map beyond an affine one. In contrast, Volterra layers realize this class directly and, on a data manifold, achieve approximation rates governed by the manifold dimension rather than the ambient pixel dimension. We therefore propose the architecture these results prescribe; raw image patches passed through a bottleneck embedding, followed by cascaded low-rank Volterra layers, which carry no activation function because each layer is nonlinear in its own right. To our knowledge this is the first pixel-space diffusion architecture whose mixing operators are Volterra layers throughout, with no attention, no convolution, no pretrained components, and no activation function on the residual path. On ImageNet $256 \times 256$ and $512 \times 512$, the resulting model is competitive with self-contained pixel-space diffusion at matched parameter and FLOP budgets.
Read the full abstract

All authors Siddharth Roheda, Rohit Chowdhury, Amit Satish Unde, Aniruddha Bala

Machine Learning

Self-contained Data Cleaning for Unsupervised Anomaly Detection

Training datasets for unsupervised anomaly detection are often assumed to be anomaly-free, but in practice, they frequently contain malicious or anomalous contamination. This contamination severely degrades the dependability and security of downstream detection models. While some contamination-robust methods exist, they rarely generalize across different model architectures and typically rely on a priori knowledge of the contamination ratio\textemdash information that is rarely available in real-world security operations. In this paper, we propose a novel, self-contained preprocessing framework for anomaly detection that identifies and removes training set contamination without relying on known contamination ratios. Our approach performs a dynamic analysis of the cosine similarity between each data point and its reconstruction during the early training stages of an auto-encoder, mapping points into a dynamic space where anomalous regions naturally exhibit lower density. By executing “hard cleaning”\textemdash removing all identified anomalous candidates from the training pool\textemdash we systematically improve data quality. Because this cleaning occurs at the data level, the security and reliability improvements generalize seamlessly across various downstream anomaly detection architectures. Evaluating our framework across 8 distinct models and 10 security-relevant datasets, our approach achieves up to 100% contamination reduction, yielding average improvements of 11.2% in F1 score and 7.3% in classification accuracy.
Read the full abstract

All authors Jordan F. Masakuna, Ahmed Yacine, Djeff K. Nkashama, Arian Soltani, Marc Frappier, Pierre M. Tardif, Froduald Kabanza

Machine Learning

Representing Tensors as Compositions for Efficient Computation on Sparse and Structured Data

We propose a new mathematical formalism in which all kinds of tensors and of operations on those tensors are represented as simple function compositions. Tensors are functions from indices to values. An element-wise operation (e.g. an activation function) is a composition with a function applied to values; re-indexing (e.g. reshape, transposition) is a composition with a function applied to indices; and a reduction is a composition with both an index function, mapping each index to several indices, and a reduction operation (e.g. product, sum, max). Combined with tensor stacking, these compositions can represent most interesting kinds of tensors (e.g. dense, jagged, all kinds of sparsity patterns, matrix factorizations), as well as lazy versions of almost every operation available in modern tensor computation frameworks, including einsums between tensors of any kind. Identities between compositions can then be used to rewrite such expressions into equivalent ones, without evaluating them. For instance, the product of two sparse matrices can be rewritten so that only products between non-zero values are computed, turning a sparse problem into a dense one. Using this algebra as the intermediate representation of a tensor computation engine would shift the focus from defining operations on tensors (e.g. einsum between sparse tensors) to defining an algebra on index functions. This should reduce the codebase complexity from $\mathcal{O}(mn^2)$, with $m$ the number of binary operations between tensors and $n$ the number of tensor types, to $\mathcal{O}(m + k^2)$, with $k$ the number of types of index functions. A draft implementation of a tensor computation framework based on this algebra is publicly available.
Read the full abstract

Differential Geometry

Controlled Riemannian rigidity

Given a compact Lie group acting smoothly and freely on a compact connected manifold, we prove that the space of invariant Riemannian metrics whose symmetry group is precisely the group in question is open dense in that of all invariant metrics provided the orbit space is sufficiently large. This strengthens the author's prior results and recovers Ebin's to the effect that trivial-symmetry Riemannian metrics constitute an open dense set. In the process, the discussion also addresses a number of apparent gaps in said sources, pertaining to the rigidity results in question.
Read the full abstract

Materials Science

A Geometry-Guided Feature-Space Framework for Interpretable Unsupervised 4D-STEM Analysis of Complex Multiphase Materials

Rapid, high-fidelity analysis of four-dimensional scanning transmission electron microscopy (4D-STEM) datasets from heterogeneous materials remains challenging because coexisting phases, orientation variants, local strain gradients, and intrinsic TEM thin-foil effects produce complex, overlapping diffraction signatures. Here, we present a machine-learning (ML) framework that combines physics-informed geometric descriptors with weighted high-dimensional embeddings to identify latent structural patterns with minimal prior knowledge. By constructing a regularized descriptor space before embedding and density-based clustering, the method improves cluster stability, suppresses noise-driven fragmentation, and reduces sensitivity to embedding hyperparameters. Spatial weighting produced the largest stability gain, increasing median cluster persistence from 0.04 to 0.55 while reducing the noise fraction from approximately 22% to 0%. Rotation weighting increased DBCV from 0.80 to 0.92, although the noise fraction rose from 0.2% to 1.0%, highlighting a trade-off between improved density separation and noise rejection. We validate the framework on two benchmark systems: Si/Si$_x$Ge$_{1-x}$ epitaxial layers with continuous lattice variation and deformation, and multiphase Ti-Nb alloys containing distinct phases and orientation variants. Comparisons with a pretrained vanilla autoencoder and model-free pattern vectorization show that the physics-informed geometric descriptor provides higher cluster fidelity, greater interpretability, and substantially lower dimensionality.
Read the full abstract

All authors Arda Genc, Daniel A. Bolden, Albert Nguyen, Daniel S. Gianola, Ravit Silverstein

Computation and Language

KeyASRBench: Measuring Whole-Phrase Keyword Preservation in Real-World ASR

Public automatic speech recognition (ASR) benchmarks increasingly report low sequence error rates, yet production systems still use hotword lists, contextual biasing, retrieval, and post-correction. This motivates a complementary question: does a transcript preserve selected names, products, technical terms, versions, and relational phrases? We present KeyASRBench, a bilingual pilot that operationalizes occurrence-based Keyword Error Rate (KER) over frozen whole-phrase annotations. The 10-hour model-conditioned challenge set contains 1,501 Chinese and English clips with 4,723 keyword occurrences. Across five locally inferred checkpoints and two official API systems, language-specific MER/WER ranges from 5.31% to 12.91%, while KER ranges from 13.61% to 33.81%. A disjoint 30-minute model-blind control yields KER of 10.65%-27.31%. The control confirms that selected-phrase errors extend beyond challenge-set selection.
Read the full abstract

Theoretical Economics

Optimal hybrid tax enforcement

Modern tax design and enforcement rely on classification data reported by taxpayers. Data improve the targeting of tax policy and audits but also open avenues for evasion. How can tax authorities use such data while accounting for their potential strategic distortion? We demonstrate that restoring deterrence generates an endogenous global verification frontier: every report below the frontier is verified through either a profit audit or data verification. Optimal hybrid tax enforcement requires selecting which data classes are subject to profit audits up to the global frontier and which are profit-audited only up to lower class-specific cutoffs, with data verification taking over beyond these cutoffs. This costly data verification requirement limits the benefit from differentiated tax enforcement. Optimal implementation is characterized by profit bunching but involves no data bunching.
Read the full abstract

Group Theory

A novel Non-Homogeneous Local Representation of the Braid Group

We construct a non-homogeneous local representation of the braid group $B_n$ by tracking the labels of individual strands through successive crossings. Each generator acts through a $2\times2$ block whose position is determined by the crossing and whose coefficients depend on the ordered labels of the participating strands. Starting with the ansatz \[ M_{xy}=\begin{pmatrix}a_x&a_y\\ b_{xy}&c_{xy}\end{pmatrix}, \] for $n\geq3$, the cubic braid relations for arbitrary ordered triples of strands determine $c_{xy}=0$ and $b_{xy}=(1-a_x)/a_x$ in the nondegenerate branch. The moving-label rule leads to an explicit semidirect-product representation, for which we prove the Artin relations and use exact evaluations of the resulting symbolic matrices. In the special homogeneous specialization, where $a_x=a$ for every strand label $x$, the resulting matrix representation is diagonally equivalent to unreduced Burau. We prove that a non-homogeneous specialization of our representation detects Bigelow's explicit nontrivial $B_5$ Burau-kernel commutator. Thus the strand-dependent coefficients distinguish a nontrivial braid that is invisible to the classical Burau representation. This separation provides evidence that the non-homogeneous strand-tracking construction is a promising candidate for a faithful Burau-type representation.
Read the full abstract

Representation Theory

A silting theorem for extended module categories

We generalize the silting theorem of Buan and Zhou to extended module categories. For a $(d+1)$-term silting object $\mathbf{P}$ over a proper $d$-truncated non-positive dg algebra $A$, we construct a $(d+1)$-term silting object $\mathbf{Q}$ over the $d$-truncated dg endomorphism algebra $B$ of $\mathbf{P}$. The $s$-torsion pairs induced by $\mathbf{P}$ and $\mathbf{Q}$ yield equivalences of Brenner-Butler type between the torsion class on either side and the torsion-free class on the other. There is also a dg algebra homomorphism from $A$ to the $d$-truncated dg endomorphism algebra of $\mathbf{Q}$, which is a quasi-isomorphism if and only if $\operatorname{Hom}(\mathbf{P},\mathbf{P}[i])=0$ for all $i\leq-d$. Moreover, we describe Auslander-Reiten theory in ${d\text{-}\operatorname{mod}\!B}$ in terms of that in ${d\text{-}\operatorname{mod}\!A}$.
Read the full abstract

Computation and Language

MultiTurnRoute: A Platform for Building and Evaluating Multi-Turn LLM Routers

In a multi-turn conversation, the choice of language model at one turn can affect which model should answer the next. The selected model's reply shapes the context for the next query, and its cached prompt prefix can reduce the cost of using that model again. We introduce MultiTurnRoute, a platform for building and evaluating multi-turn LLM routers. A shared interface gives built-in and user-defined routers the current query, dialogue history, and an estimate of each model's reusable prefix length. Our routing algorithm, Chain-of-Route (CoR), uses a learned gate over recent and recurrent dialogue history to predict response quality and estimates cost from each model's reusable prefix. MultiTurnRoute Arena ranks routers on recorded conversations and compares session quality and cost in interactive runs where a simulated user responds to each reply. On a benchmark drawn from WildChat and LMSYS-Chat-1M with per-turn quality annotations for five LLMs, CoR reduces the normalized cost required to match the strongest model's quality by up to 20.3% relative to the strongest routing baseline, while matching or improving peak response quality.
Read the full abstract

All authors Jiaqi Xue, Mengxin Zheng, Heng Huang, Qian Lou

Computer Vision

Fine-resolution land-use segmentation and synthetic image-label augmentation in the Kathmandu Valley

Parcel-scale land-use data is missing for most rapidly urbanizing Global South cities, as current segmentation architectures rely on multispectral imagery and planned layouts. We introduce a hand-annotated six-class land-use corpus covering 202 km$^2$ of Kathmandu Valley across 408 fine-resolution RGB scenes. The dataset is uniquely challenging: RGB spectral similarity is high, class imbalance approaches 100:1, and key structural features like rivers measure just two to three pixels wide at a standard hierarchical decoder's prediction stride. Evaluating five architectures reveals that standard random data splitting conceals massive spatial leakage. Switching to a strict sequence-block split drops performance by up to 0.095 mIoU. This penalty exceeds the performance gap between the architectures themselves. We modified the SegFormer baseline by integrating a half-resolution residual detail path, boundary-reliability reweighting, a topology-aware loss, and a presence-balanced sampler. Adding only 1.9% more parameters, the enhanced architecture improves mean IoU across every test fold, with gains concentrated in rare, thin-structure classes. However, both models achieve identical results on the external LoveDA benchmark. Finally, augmenting with diffusion-generated pairs at matched compute degraded accuracy across all architectures, especially harming rare classes it aimed to improve. This demonstrates that blindly incorporating synthetic data can severely damage performance, highlighting a critical need to rigorously evaluate and filter generated samples before training.
Read the full abstract

Information Theory

Efficient SC-Flip Decoding for Polar Codes through Decision-Boundary Normalization

Accurate first erroneous position (FEP) identification is key to improving the error-correction performance of successive cancellation flip (SCF) decoding. However, FEP identification remains challenging with a limited number of re-decoding attempts. To address this issue, we propose decision-boundary normalization (DBN) to improve FEP identification and develop a novel DBN-SCF decoding algorithm. DBN normalizes the magnitude of each decision log-likelihood ratio (LLR) by the standard deviation predicted by the Gaussian approximation (GA) for the corresponding bit-channel. This allows a single normalized threshold to be shared across bit-channels within each code configuration, giving it a consistent statistical meaning and avoiding per-channel calibration. DBN-SCF decoding selects candidate flip positions from the information set in the natural successive cancellation (SC) decoding order using a threshold, without sorting. Moreover, DBN-SCF decoding achieves better error-correction performance with a limited number of re-decoding attempts. It also requires lower average decoding complexity than existing SCF decoding methods at a comparable level of error-correction performance.
Read the full abstract

All authors Gulinisha Yiming, Zhongjun Yang, Zhicheng Liu, Zhiming Ma, Xumin Ni, Yuting Liu

Artificial Intelligence

Inference for Reasoning Models: Practical Lessons from Evaluation and Serving

Running a language model involves more than its trained weights. The same model can receive very different benchmark scores, produce different answers, or cost more to serve depending on how it is evaluated and run. In one experiment in this book, changing only the maximum response length moved accuracy on a math benchmark from 0.08% to 93.18%. This book explains how inference settings shape the results we report and the systems we deploy. We begin with evaluation, including generation limits, answer extraction, scoring methods, and quantization. We then examine why outputs can change across batches, attention backends, and caching, even when sampling is disabled. The next chapters cover serving performance, including batching, memory capacity, latency, throughput, and cost. We also study structured outputs, tool calls, reasoning budgets, and long-context tasks, showing what these controls and benchmarks do and do not establish. The book concludes with practical guidance for recording inference runs, checking evaluations, estimating capacity, and comparing systems fairly. Written for researchers and engineers working with language models, it aims to make inference results easier to understand, reproduce, and trust.
Read the full abstract

Artificial Intelligence

What Broke & Why: Practical Lessons from Reinforcement Learning Post-Training

Reinforcement learning with verifiable rewards (RLVR) has become an important way to improve how language models reason, solve problems, and use tools. But getting it to work in practice is not always straightforward. Training can fail because a model receives no useful reward, learns to exploit the reward system, stops exploring, or appears to improve when the evaluation itself is misleading. In this book, we give a practical introduction to reinforcement learning post-training for language models, focusing on what works, what fails, and why. We begin with the foundations of post-training, explain GRPO and the training loop step by step, and then follow five training programs covering math reasoning, search agents, training stability, mixture-of-experts models, and software engineering tasks. Each program shows the problems encountered, the changes made, and the lessons learned. The book also examines what reinforcement learning actually changes inside a model. The remaining chapters provide a practical reference for reward design, evaluation, training tool-using agents, debugging common failures, reproducibility, rollout engines, and hardware. Written for researchers and practitioners working with open models, the book aims to help readers understand how to train models, recognize when something goes wrong, and decide what to try next.
Read the full abstract