Peer Revenue Past citations are no guarantee of future returns

Papers

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

Computer Science

cs Watch Computer Science (sign in to use)

Friday 14 August

Computer Vision

Marionette: Predicting World States, Rendering Geometry, Painting Appearance

Interactive game world models typically autoregress visual observations directly in pixel or latent space, forcing structured properties such as pose, geometry, and occlusion to be implicitly maintained by the same generative sequence. Over long horizons, errors in these latent world properties accumulate, making consistency and controllability fragile. We explicitly model the evolving world state, delegate exact geometric computation to a fixed, zero-parameter renderer, and leave the neural model to synthesize appearance. We instantiate this idea as Marionette, a world model for interactive games with articulated characters. First, a two-stage autoregressive dynamics model predicts an explicit and interpretable 276-dimensional 3D world state comprising multi-entity articulated skeletons, metric root trajectories, and rotations. Second, a zero-parameter graphics bridge converts the predicted state into pose-control videos, computing world-space geometry and occlusion in closed form. Third, a control-conditioned video-diffusion observation model synthesizes photorealistic RGB observations from the resulting structured controls. Our experiments establish two properties of Marionette. First, the predicted world state is directly controllable. Forcing a mismatched action stream changes root-aligned joint error by 31% across 48 held-out segments. Second, long-horizon behaviour is determined in the state, and can be repaired there. Left free, the two generated characters drift to 21.2 m apart (recorded sessions stay near 5 m) and a third of frames show ground penetration. Two rules imposed on the explicit state, a terrain collider and a separation cap, cut penetration by 66% and keep the pair engaged, with no change to the observation model. Routing appearance through the predicted state costs no fidelity we can detect, at an FVD of 831 against 799 for recorded pose.
Read the full abstract

All authors Zian Meng, Zhen Li, Chuanhao Li, Qiang Li, Kaipeng Zhang

Computational Complexity

Polynomial-Factor Deterministic NP-Hardness for SVP in Every $\ell_p$ Norm with $p>2$

We prove that for every constant $2<p<\infty$, the shortest vector problem in the $\ell_p$ norm is NP-hard to approximate within a polynomial factor under a deterministic reduction. More precisely, for lattices of rank $M$, we obtain a factor of $M^\varepsilon$ for every constant $0<\varepsilon<\min\{(p-2)/(4p),1/8\}$. For $p=\infty$, the same holds for every constant $0<\varepsilon<1/8$. Our reduction adapts OpenAI's encoding of a satisfying assignment for the closest vector problem ("Ten Advances in Mathematics and Theoretical Computer Science"). The encoding has one block of coordinates for each evaluation of a polynomial, with one position for each possible value. We allow the entries in a block to be arbitrary field elements, which can cancel when we take weighted sums of powers of these values. We show that on an unsatisfiable formula, a nonzero codeword with small support must have many nonzero entries in each block it uses. Applying a Hadamard matrix to each block, as in a work by Hair and Sahai (STOC 2026), gives the gap in vector length. The main results of this paper were discovered by Codex / ChatGPT 5.6 Sol Ultra using harness components from the UCLA Moonshot Harness Project, in response to a general strategy suggested by the authors. See the end of the introduction for a discussion.
Read the full abstract

Artificial Intelligence

Handover of In-Context Learning State Across Session Boundaries

This study investigates the methodological and theoretical properties of session handover in applications that use large language models. A task may continue in a new session when the context reaches the model's input limit, when the application restarts, or when another agent is asked to finish the task. The application must then decide which information from the earlier session to pass on. We formulate handover as the transfer of a task-relative in-context learning (ICL) state and distinguish exact recovery of earlier material from preservation of the target distribution. Under an exogeneity condition, predictive equivalence characterizes the coarsest deterministic sufficient handover and gives a fixed-length bit requirement. The analysis isolates the effects of the memory constraint, the writer, and the continuation procedure, and quantifies the cost of writing before the realized downstream query is known. We propose a three-part record that stores decisions and constraints exactly, uses task-justified statistics for repeated evidence, and retains original observations whose effect is not preserved by those statistics. Gaussian linear regression gives an exact finite-dimensional handover and finite-bit perturbation bounds, while nonparametric regression gives upper and lower bounds that relate memory to squared prediction error. These results provide a theory and method for deciding what a handover must retain and how its memory requirement depends on the continuation task.
Read the full abstract

Distributed Computing

Validating LLM-Modernized Scientific Software Through Differential Fault Injection

Large language model (LLM) agents are increasingly used to modernize the legacy Fortran underlying production scientific software, but validation of these transformations emphasizes nominal executions and may not test whether a modernization preserves the original code's response to faults, perturbations, and reduced precision. We present a differential fault-injection validation method: a harness instruments the shared self-consistent-field driver of GAMESS at twelve sites and applies identical, deterministic faults to the original and LLM-modernized implementations, isolating the converted integral kernels. Across more than 2,200 runs, transient-fault absorption costs match a contraction-based model (predicted slopes 0.74 and 1.49 iterations per bit; measured 0.82 and 1.50), persistent perturbations halve final-energy error per additional bit, and the campaigns expose phase-dependent parallel deadlocks and false convergence under reduced precision. The original and modernized kernels agree in all 200 paired injections, and a measurement-guided synchronization change composes with the modernization, matching in all 40 pairs.
Read the full abstract

All authors Evan Coleman, Yuzhong Shen, Masha Sosonkina, Peng Xu

Artificial Intelligence

Participatory Moral AI Is Not Neutral: The Invisible Hand of Developers

As AI systems make more morally loaded decisions across society, one response has been moral preference elicitation. In this approach, researchers poll participants on hypothetical dilemmas and use the aggregated votes to train a policy that an AI model then applies at scale. Before any vote is cast, developers make three key choices in the moral AI elicitation pipeline: feature scoping, voter sampling, and question framing. In other words, they decide which features go to a vote, which voters to include, and how to present the question. These choices are often opaque, undocumented, and treated as technical details rather than normative ones. We examine each of these choices within a common empirical study and show that each can shape the preferences produced by moral AI elicitation. Across two phases (N = 809) in three deployment contexts (i.e., AI kidney allocation, AI agents simulating absent workers, and generative AI depictions of the deceased), we examine the three main stages of the moral AI elicitation pipeline. First, morally relevant features shift across contexts. This suggests that feature schemas should not be assumed to transfer across deployment domains. Second, preferences differ by political ideology for roughly one-third of features, with some differences reversing direction. The ideological composition of the voter pool can therefore affect the resulting aggregated preference profile. Third, the wording of the elicitation question can narrow or widen ideological gaps by up to a full scale point. The framing conditions also change how moral foundations are associated with participants' judgments. Taken together, these findings suggest that voting-based alignment cannot deliver fair or transparent AI by aggregation alone; at minimum, each stage of the moral AI elicitation pipeline should be audited and disclosed.
Read the full abstract

Numerical Analysis

Comment on arXiv:2106.08363v3 [math.NA], E. Abreu, A. Espirito Santo, W. Lambert, and J. Perez, Convergence of a Lagrangian–Eulerian scheme by a weak asymptotic analysis for one-dimensional hyperbolic problems

This Comment concerns arXiv:2106.08363v3 [math.NA] by E. Abreu, A. Espirito Santo, W. Lambert and J. Perez, published in Numer. Methods Partial Differential Equations 39 (2023) 2400-2443. That article builds its scheme on space-time control volumes whose lateral boundaries, called "no-flow curves", solve dsigma/dt = H(u)/u and are presented as new. We show, equation by equation, that this object is the space-time integral curve of the locally conservative Eulerian-Lagrangian method of Douglas, Pereira and Yeh [Comput. Geosci. 4 (2000) 1-40], that the no-flow region is the tube those curves bound, and that its scalar forward-tracked form appears in Mancuso, Pereira and de Souza [TEMA 8 (2007) 269-276, 277-286]. At issue is not moving a control volume, which is generic, but which curve moves it: H'(u) and H(u)/u coincide identically only for a linear flux, and only the latter makes lateral mass flux vanish. The commented article itself calls the 2000 paper the first to introduce space-time local conservation, with an integral tube bounded by integral curves, while its abstract calls the same object introduced by the authors. A full-text corpus documents the changing terminology and attribution. Papers published in 2025 and 2026 use the same construction while citing later work but not DPY. Another 2026 paper gives mixed attribution: it calls the no-flow curve an "extension" of the DPY integral curve, although its own zero-flux definition and ratio ODE show identity, and it states that diffusion and dispersion do not modify the defining vector field. It extends the model, scheme and analysis, not the continuous curve. The later discrete contributions are not challenged.
Read the full abstract

Sound · Cross-listed from eess.AS

Singer-Informed Vocal Source Separation for Multi-Singer Music Mixtures

Music source separation systems typically extract a single vocal track and do not distinguish between multiple singers. We study singer-informed vocal source separation for multi-singer mixtures. Our framework introduces a short enrollment recording of a target singer to guide separation through a learned embedding. The singer embedding is incorporated using feature concatenation or feature-wise linear modulation (FiLM), enabling the model to focus on the target singer while suppressing interference. We construct a duet dataset based on DAMP-VSEP with quality filtering and non-overlapping enrollment segments. Experiments on solo and duet settings show that while baseline models perform well for single-singer mixtures, the proposed method improves target-singer extraction in multi-singer cases, increasing target-singer SI-SDR from 0.33 dB to 5.58 dB. Fréchet Audio Distance (FAD) further shows improved perceptual quality and better alignment with target audio distributions. Code and checkpoints are available at https://github.com/jocelynxu01/singer-separation-paper.
Read the full abstract

Human-Computer Interaction

Visualizing Uncertainty in Non-linear Projections with Ensembles

Widely used non-linear dimensionality reduction (NLDR) methods such as UMAP and t-SNE are stochastic–repeated runs on the same data can produce different low-dimensional projections. In this paper, we explore two problems related to projection variability: on some datasets clusters, structure, and outliers may change run-to-run, and on others projections can be extremely stable when overfitting noise. To address the first problem, we propose visualizing the median of multiple NLDR outputs rather than relying on individual projections. To address the second, we perturb input data before creating consensus embeddings. We find that taking the median of multiple projections performs comparably to individual runs on multiple quality metrics, while increasing perturbation emphasizes global over local structure. We show through a set of exploratory visualizations that even relatively simple ensemble presentations can be used to better communicate the reliability of projection patterns.
Read the full abstract

Information Theory

Learning-to-Transition for Large-scale and High-Order MIMO Detection

High-order multiple-input multiple-output (MIMO) detection requires efficient search over a large discrete symbol space while producing reliable soft information for channel decoding. This paper develops a learning-to-transition (L2T) framework that formulates MIMO detection as a stochastic sequence of complete-vector transitions. At each transition, a channel-coupled Transformer updates both the instance embedding and the sampling policy, while a blockwise autoregressive factorization captures inter-stream dependence with moderate sequential complexity. For hard-output detection, a transition network is applied recursively and trained through a residual-to-BER curriculum, which first learns the MIMO search geometry from the exact residual metric and then aligns the policy with transmitted-bit accuracy. For soft-output reception, the well-trained hard policy is cloned at the parameter level into every layer of an untied soft-input soft-output iterative detection and decoding (IDD) receiver. This tied-to-untied transfer preserves the learned zero-prior search dynamics while enabling layer- and round-specific specialization under decoder feedback. Within each IDD round, decoder priors tilt candidate generation according to Bayes' rule, and likelihood-weighted terminal hypotheses produce posterior and extrinsic log-likelihood ratios for LDPC decoding. A multi-stage training strategy further stabilizes the hard-to-soft transfer by progressively exposing the receiver to synthetic and in-loop decoder-generated priors.
Read the full abstract

Artificial Intelligence

Split the Labor: Separating Evidence Interpretation from Decision Aggregation

Systems that ask a language model to reach a conclusion from many sources usually concatenate them into one prompt. This conflates two operations with different requirements. Interpreting a source rewards capacity and context. Combining interpretations rewards fixed arithmetic, comparability across instances, and the option to return nothing. Once separated, the design problem becomes the interface between them. We propose a four-field evidence tuple (hypothesis, reliability bucket, rationale, provenance) and show that fixing it determines both halves. The separation also reveals a failure mode in how such systems combine, which we call count-scale drift. Thresholding a sum of unnormalized weights is exactly posterior thresholding, but at an operating point that slides with the number of sources consulted. The slide grows with reader reliability. When source reliabilities differ, the vote rule and the posterior order instances differently, and no threshold reconciles them. Pooling calibrated log-likelihood ratios addresses both problems. The fix is arithmetic rather than architectural, and applies to a class of rules beyond language models: score-summing triage engines, diagnostic panels scored by counting positives, and additive multi-signal detectors. We then instantiate the principle twice on one longitudinal corpus, once after outcomes resolve and once before. The same partition helps in both, at different granularities: over reading in the first, over learning capacity in the second. There, a small sequence encoder on an easy auxiliary objective plus a tree ensemble carrying the censored survival loss reaches 0.921 AUPRC against 0.805 for a hand-crafted baseline. We separate what transfers from what must be re-estimated per domain, and state five predictions that would falsify the framework, three negative results, and which comparisons remain confounded.
Read the full abstract

Numerical Analysis

Nodal discontinuous Galerkin methods for non-ideal equations of state: pressure equilibrium preservation and entropy correction

Structure-preserving discontinuous Galerkin (DG) methods typically improve the robustness of high order simulations of real fluids. In addition to conservation, key structures include the preservation of pressure equilibrium and satisfaction of at least one entropy inequality. In this work, we investigate conservative discretizations using exactly pressure equilibrium conserving (EPEC) and approximately pressure equilibrium conserving (APEC) flux differencing DG formulations, as well as entropy stable formulations through the use of minimally dissipative corrections for non-ideal equations of state (EOS). We introduce an analysis of EPEC schemes and a new procedure for designing such fluxes based on a generalization of Tadmor's shuffle condition. We also analyze APEC DG schemes and show that the incorporation of dissipative interface penalization terms does not significantly increase pressure equilibrium errors, especially at higher orders of approximation. Finally, we observe that when combined with APEC flux differencing formulations, entropy correction improves robustness for under-resolved solutions and long-time simulations.
Read the full abstract

All authors Jesse CHan, Hendrik Ranocha, Raymond Park, Joshua Lampert, Eric Ching, Ayaboe Edoh

Machine Learning

RecipeNet: A Hierarchical Transformer for Recipe Data

Recipe data arises in domains such as materials synthesis, pharmaceutical formulation, and industrial manufacturing, where procedures are represented as ordered sequences of steps containing heterogeneous structured fields. Existing tabular learning methods typically flatten this structure into fixed-schema representations, limiting their ability to capture hierarchical field interactions and procedural dependencies. We propose RecipeNet, a hierarchical Transformer architecture that encodes field-level interactions within each step and sequential dependencies across steps through stacked Transformer encoders. Experiments on multiple recipe datasets and tasks demonstrate that RecipeNet consistently outperforms existing tabular models, highlighting the value of hierarchical and sequential modeling for recipe representation learning.
Read the full abstract

All authors Pin-Yen Huang, Sachin Chhabra, Prasanth Sai Gouripeddi, Abhinav Kumar, Baoxin Li

Social and Information Networks

RegRole: Regularized Role Detection and Prediction in Temporal Dynamic Networks

This paper introduces a dynamic role discovery technique in temporal dynamic networks, utilizing temporally regularized Non-negative Matrix Factorization (NMF). Our technique differs from existing dynamic role analysis techniques by creating a consistent set of roles across all time periods, as well as a universal transition matrix that describes the probability of transitioning between roles. We also apply a regularization penalty to ensure that role membership does not change dramatically between time periods making our model more robust against real-world noise. We test our data on five real-world and one synthetically simulated dataset using both engineered and automatically generated features. We demonstrate that the proposed regularized role detection method, for appropriate regularization weight parameter reduces prediction errors compared to other techniques. Furthermore, trace analysis of the transition matrices indicates that our method yields a more stable system, that is, individuals are more likely to stay in their roles with fewer arbitrary transitions. Our model learns time-aligned roles, captures behavioral transitions over time, and scales efficiently to large and sparse graphs.
Read the full abstract

Artificial Intelligence · Cross-listed from cond-mat.mtrl-sci

Universal Thermodynamic Interatomic Potentials for Crystalline Materials

Free energies govern solid-state phase stability, yet computational materials discovery still relies largely on ground-state energies because free energy calculations require ensemble averages. We introduce the thermodynamic interatomic potential (TIP), which extends an interatomic potential from its static energy to a thermodynamically consistent Gibbs free energy model, with thermodynamic responses following from temperature and pressure by automatic differentiation. We implement TIP[UMA] using the universal potential UMA, train it on free energies from quasi-harmonic to molecular dynamics fidelity, and calibrate it to higher-resolution calculations or experiment. From a single evaluation, it returns the equation of state of a crystal and locates phase transitions among competing branches, including dynamically stabilized phases. Fine-tuning extends the model to alloy solubility limits and miscibility gaps. TIP makes the free energy as accessible as the potential energy, opening finite-temperature phase stability to high-throughput discovery.
Read the full abstract

All authors Juno Nam, Bowen Deng, Xiaochen Du, Luis Barroso-Luque, Benjamin Kurt Miller, Rafael Gómez-Bombarelli

Cryptography and Security

Lower Bounds on Black-Box Constructions of Pseudorandom Functions

In their seminal work, Goldreich, Goldwasser, and Micali [CRYPTO 1984] constructed a pseudorandom function (PRF) using a black-box access to a pseudorandom generator (PRG). When combined with Levin's domain extension technique, the GGM construction invokes the PRG $\omega(\log n)$ times, where $n$ denotes the input length to the PRG. To this day, no black-box construction achieving fewer calls is known. Recently, Beimel, Malkin, and Mazor [CRYPTO 2024] showed that for a certain family of constructions, which they termed tree constructions, the GGM construction is optimal. However, the basic challenge of whether a PRF can be built with just one invocation of the PRG still remains open. In this work, we consider fully black-box constructions of PRFs from PRGs, where both the construction and the reduction are required to be black-box, and the number of interactions the reduction makes with the adversary is independent of the number of oracle calls the adversary makes to its underlying function within each interaction. Our main result shows that no such construction can have $o(n/\log n)$ and $o(\mathsf{in}/\log\mathsf{in})$ non-adaptive calls to the PRG, where $\mathsf{in}$ is the input length of the PRF. This impossibility holds even for weak PRFs with one-bit output, where the adversary is restricted to making i.i.d. uniformly random queries. In addition, we prove a lower bound for weak PRFs with sufficiently long outputs that holds even when the construction is allowed to make adaptive queries to the PRG.
Read the full abstract

Game Theory

Ex-ante versus Ex-post: Egalitarian Facility Location Mechanism Design

We study the facility location mechanism design problem where $n$ strategic agents report locations in Euclidean space and the mechanism outputs a single facility location. Each agent's cost is its distance from the facility, and our objective is to minimize the egalitarian cost, i.e., the maximum agent cost, in a strategyproof way. The optimal deterministic approximation ratio is $2$, achieved by any dictator mechanism. We study the power of randomized strategyproof-in-expectation mechanisms. Prior work has focused on ex-post evaluation, defined as the expected maximum agent cost. We instead study ex-ante evaluation, defined as the maximum expected agent cost, which is naturally aligned with strategyproofness in expectation. We establish the following results: (1) Low dimensions: Strict ex-ante vs. ex-post separation. In $\mathbb{R}$, we give a simple strategyproof mechanism achieving the optimal ex-ante approximation ratio of $1$. In $\mathbb{R}^2$, we design the "Random Rotated Corner" mechanism, with ex-ante approximation ratio at most $1.598$, breaking the deterministic barrier. For the ex-post objective, we prove a lower bound of $1.605$, yielding a strict separation in $\mathbb{R}^2$. (2) High dimensions: Impossibility. In $\mathbb{R}^d$ for $d \gg 1$, we show that no strategyproof-in-expectation mechanism improves on the deterministic dictator mechanism beyond $o_d(1)$. Thus neither ex-post nor ex-ante evaluation yields improved fairness guarantees in high dimensions. An implication is that the "Random Rotation Coordinate-Wise Median" (RRCWM), currently the best known mechanism for the utilitarian objective, is also best possible for the egalitarian objective in high dimension: we show it achieves an approximation ratio of $2$ for both ex-post and ex-ante objectives in $\mathbb{R}^d$ for every $d \ge 1$.
Read the full abstract

Machine Learning

Rollplex: Cross-Phase GPU Spatial Sharing for Vision Language Model Post-Training

Vision-language models (VLMs) enable embodied agents to reason and act from visual observations and language instructions. Reinforcement learning (RL) post-training enhances these capabilities using task feedback, but current on-policy RL runtimes execute rollout, reference scoring, and actor training in strict serial phases. While effective for text-only RL, this phase-granular execution is wasteful for VLMs, where processing dense video inputs and prompt prefixes occupies a large fraction of each phase. Because prefix processing is independent of the generated response, it can be run alongside rollout decoding, which leaves GPU compute capacity underutilized, without breaking synchronous on-policy semantics. We present Rollplex, a runtime that decomposes the reference and training phase and moves the prefix computation into the rollout decode window. Realizing this schedule requires more than concurrent kernel launches: naive colocation of Qwen2.5-VL-32\,B requires roughly 165\,GiB per GPU, while rollout and training prefer different tensor-parallel (TP) degrees and weight layouts. Rollplex addresses these constraints with two mechanisms. Phase-aware memory management controls HBM residency according to producer–consumer lifetimes. Parallelism-aware weight sharing uses the same physical storage for layout-compatible tensors across distinct TP degrees and reconstructs only incompatible tensors, avoiding a complete second actor copy. On 32 H800 GPUs, Rollplex achieves $1.23\times$–$1.30\times$ speedup over serial colocation and $1.57\times$–$2.24\times$ over disaggregation under the same GPU budget, while preserving the synchronous RL update.
Read the full abstract

All authors Hanfeng Lu, Tianyu Feng, Suyi Li, Yuheng Zhao, Wei Gao, Shaopan Xiong, Ju Huang, Siran Yang, Jiamang Wang, Lin Qu, Wei Wang

Numerical Analysis

Hybrid Dealiasing and Implicit Packing for Real Convolutions

Hybrid dealiasing is an FFT-based method for computing linear convolutions of complex-valued data that reduces the cost of dealiasing by performing zero padding implicitly. We develop two new algorithms that extend hybrid dealiasing to real-valued convolutions. The first algorithm exploits conjugate symmetries in the transformed data and computes each residue contribution directly. The second algorithm employs complex-valued hybrid dealiasing via a new implicit packing technique, which packs real-valued data into complex-valued arrays and performs all unpacking and packing operations implicitly in the transformed domain. Multidimensional versions of both algorithms are obtained by recursive decomposition into lower-dimensional convolutions. Based on these algorithms, efficient routines to compute convolutions of equal-length input arrays are implemented in the open-source FFTW++ library. These routines outperform the standard method of explicit zero padding in one, two, and three dimensions. In many cases, implicit packing without hybrid dealiasing is nearly as fast as the general hybrid dealiasing algorithm, particularly when multithreading.
Read the full abstract

Machine Learning

Generating Benchmark Health Data Using a Tabular Diffusion Transformer

Cross-Tabular Data Generation (CTDG) seeks to learn a generative model from multiple heterogeneous tables and produce new synthetic tabular datasets. However, existing synthetic tabular data generation methods are largely restricted to single-input-table scenarios and struggle to effectively handle multiple heterogeneous tables with diverse feature sets. To address this limitation, we propose a two-stage framework for cross-tabular data generation. In the first stage, each heterogeneous raw table is transformed into a standardized statistical table with the same set of columns across all tables. Each statistical table captures the marginal distributions of the original columns and the pairwise correlations among them. In the second stage, a diffusion transformer model is trained to capture structural patterns across these homogeneous statistical tables and to generate synthetic statistical tables. Synthetic raw tables are subsequently reconstructed from the generated statistical tables via multivariate Gaussian sampling followed by an inverse probability integral transform. This two-stage CTDG framework enables the learning of a unified generative model from multiple heterogeneous tables and supports the generation of an unlimited number of realistic synthetic heterogeneous tables. Experimental results demonstrate high fidelity in the learned statistical representations and a favorable fidelity-diversity trade-off in the generated synthetic data, validating the effectiveness of the proposed approach.
Read the full abstract

All authors Hao Yan, Lisa Pilgram, Dan Liu, Linglong Kong, Fida Dankar, Khaled El Emam

Information Theory

Lossy Compression via Sparse Regression Codes: Generalized Construction and Finite-length Bounds

We study sparse regression codes (SPARCs) for lossy compression under simple greedy encoding rules, including both correlation-based and distance-based methods. We generalize the SPARC construction, and consider the class of additive orthogonal regression codes, of which standard SPARCs are a special case. For this class of codes, we derive nonasymptotic bounds on the squared-error distortion by tracking the evolution of the encoding residual across stages. Our results highlight the role of power allocation in controlling the distortion, allowing us to optimize the allocation based on the parameters of the code. The optimized allocation improves the finite-length compression performance of SPARCs, and our bounds provide distortion guarantees for lower complexity variants of SPARCs, like signed SPARCs and $K$-sparse SPARCs.
Read the full abstract

Machine Learning

Approximate Muon with low-rank adapters

The Muon optimizer shows clear benefits versus alternatives when pretraining neural networks. However, it is used less frequently for parameter-efficient fine-tuning (PEFT). One potential reason is that the most common PEFT method, LoRA, does not naturally combine with Muon since it is not mathematically possible to orthogonalize the weight update given by a low-rank parameterization. In this paper, we address this issue by approximating the solution to a relaxed Muon objective in the low-rank setting via linearization and then least-squares. We provide an efficient implementation that uses matmul operations only, as opposed to more complex linear algebra decomposition routines. Our method, sMuon (small Muon), performs favourably across SFT and a ReLoRA pretraining experiment. While results are model- and eval-dependent, we find overall that using Muon for low-rank fine-tuning provides moderate performance improvements.
Read the full abstract

Systems and Control

Optimal Scheduling of Road Maintenance Jobs Considering Impact on Traffic Flows

Network-level maintenance planning requires repeated evaluations of equilibrium traffic flows under road capacity reductions. While equilibrium traffic assignment models are well established, their repeated solution quickly becomes computationally prohibitive and challenging to embed within maintenance scheduling problems. This paper investigates data-driven surrogate models that approximate equilibrium arc flows directly from origin-destination demand, using optimization-based equilibrium solutions as ground truth. A real-world case study based on traffic data from the Newark, New Jersey area demonstrates the effectiveness of the proposed approach as a scalable building block for future maintenance scheduling frameworks.
Read the full abstract

All authors Charitha Nandepu, Lohitha Kalepu, Gabriele Ciavarella, SangWoo Park

Artificial Intelligence

Twin: Playing an Unknown Game with a Test-Time Digital Twin

We present a Test-time World-model Inference (Twin) system, in which a frontier coding agent writes an executable world model for completing continual learning tasks, such as ARC-AGI-3 games. Traditional approaches hand-engineer such models, one custom design per task. Each game hides its rules and goal, and our system constructs them from simulation and interaction alone. Its inductive prior over grid games is strong enough to recover the true transitions of the game and the goal on nearly all levels. Replay validation happens in a twin world model. The harness enforces that an action is not made until the program reproduces every previous observed game transition. Each mismatch between a world model prediction and the actual action result becomes a counterexample that is used to repair the world model. Twin clears 179 out of 183 levels (97.8%), and does so more efficiently than humans in 158 out of 179 levels (88.3%). The system infers the goal before any reward on 156 of the levels it clears (87.2%), and in the remaining levels automatically discovers the goal by search. The benchmark scores completion and action efficiency, between 0 and 100, against humans playing each game for the first time. Played directly, the base model scores only 7.8%; an off-the-shelf harness increases it to 61.1%, whereas our twin world model increases the same base model to 93.3%, clearing 23 out of 25 games. Building a usable world model is simpler than anticipated, whereas the harder problem is inferring the right goal.
Read the full abstract

All authors Alexy Skoutnev, Kirill Acharya, Gaston Longhitano, Madeleine Udell, Kevin Ellis, Iddo Drori

Data Structures and Algorithms

Isomorphism of tournaments with bounded VC dimension

The tournament isomorphism problem is one of the two fundamental bottlenecks to designing better algorithms for the graph isomorphism problem. Though the problem has been investigated for more than five decades, compared to graphs, there are only very few results on the isomorphism problem of tournaments. For most classes of tournaments neither hardness nor polynomial-time solvability is known. Tournaments of bounded VC dimension are such a class for which no results are available, even though the VC dimension is arguably one of the most robust and central notions of combinatorial tameness. Resolving an open problem of Neuen and Grohe, we show that the isomorphism problem for tournaments of VC dimension $d$ can be decided in time $n^{O(d\log d)}$. Consequently, automorphism groups of tournaments of bounded VC dimension can be computed in polynomial time. To this end, we develop a new method to isomorphism-invariantly decompose tournaments. To facilitate recursion, we introduce the notion of a patched tournament and analyze bounded VC dimension in patched tournaments. We design a recursive algorithm that balances the size of the decomposed pieces against their number and makes use of the structure of near twins. In an orthogonal direction, it is known that a hereditary class of tournaments has unbounded VC dimension if and only if it contains all 2-colorable tournaments. As a second result, we show that also this class does not form an obstruction towards polynomial-time isomorphism testing and indeed show that isomorphism of tournaments of bounded chromatic number is polynomial-time decidable.
Read the full abstract

Numerical Analysis

Geometry-Conforming Finite Element Methods for Interface Problems on Fitted and Unfitted Meshes

We develop an arbitrary-degree geometry-conforming finite element (GC-FE) framework for two-dimensional elliptic boundary value and interface problems on curved domains. Using the Frenet–Serret transformation, curved-boundary and interface-fitted segments are represented exactly, while polynomials in Frenet coordinates generate generally nonpolynomial local shape functions in physical coordinates. For interface-unfitted meshes, GC-FE spaces on curved-boundary elements are coupled with geometry-conforming immersed finite element (GC-IFE) spaces on interface-cut elements, with standard polynomial spaces used elsewhere. We establish optimal approximation, inverse, and trace estimates for the GC-FE spaces. For fitted meshes, we prove well-posedness and optimal error estimates in energy and $L^2$ norms for a symmetric interior penalty discontinuous Galerkin discretization. By retaining the prescribed curves exactly, the method avoids the geometric variational crime associated with curved-geometry approximation and requires no corresponding geometric consistency estimates. Numerical experiments confirm the predicted rates, show global accuracy comparable to nodal isoparametric finite elements and smaller true-interface trace errors in the reported tests, and demonstrate the coupled GC-FE-GC-IFE method on interface-unfitted meshes.
Read the full abstract

All authors Yuanhui Lin, Tao Lin, Xu Zhang, Minfu Feng

Robotics

Ensuring Safe Physical AI in Urban Mobility via Hazard-Informed Synthesized Envelopes

As heterogeneous robotic systems deploy across diverse urban zones, maintaining safety amid complex human-robot interactions remains a critical challenge. We present a unified framework that bridges systematic hazard analysis and runtime enforcement using hazard-informed safety envelopes. Rather than treating safety as a static constraint isolated within individual software modules, we introduce a cross-layer safety transformation process spanning symbolic, spatial, and dynamic world models. We show how this representation naturally interfaces with physical AI runtime harnesses to guarantee safe urban mobility.
Read the full abstract

Machine Learning

LP-NAS: Linear Programming-based Neural Architecture Search

Neural Architecture Search (NAS) aims to automate neural network architecture design, reducing reliance on human expertise. Among the various NAS methods, differentiable NAS has gained prominence due to its efficiency and accuracy compared to conventional NAS approaches. Since differentiable NAS relaxes the architecture search space into a continuous domain, it is possible to apply principles from continuous optimization to NAS. In this paper, we propose Linear Programming-based NAS (LP-NAS), a mathematical programming-based framework for differentiable NAS that is applicable to a wide range of continuous search spaces. LP-NAS formulates a linear program (LP) using the validation-loss gradient and the training-loss Hessian to compute an architecture update direction that improves generalization while preserving the optimality of the model parameters. By following this LP-derived descent direction, LP-NAS efficiently navigates the architecture search space, leading to faster and more effective architecture optimization. We introduce two computationally efficient variants of LP-NAS, namely S-LP-NAS and R-LP-NAS. Applying LP-NAS to the Differentiable Architecture Search (DARTS) search space results in two algorithmic variants, S-LP-DARTS and R-LP-DARTS. Both variants achieve faster convergence and significantly higher validation performance during the early search iterations than the standard DARTS algorithm. Extensive experiments on CIFAR-10 and CIFAR-100 show that LP-DARTS outperforms standard DARTS in both the architecture search and evaluation phases. Additionally, we compare our approach with several DARTS variants (P-DARTS, PC-DARTS, and STO-DARTS) on the CIFAR-10 dataset and demonstrate its effectiveness. Furthermore, we validate the transferability of the discovered architectures through experiments on the ImageNet dataset.
Read the full abstract

Programming Languages

Memory Allocation for Constant-Bounded Programs

This work studies memory allocation for constant-bounded programs, whose execution length is syntactically limited for all inputs. Examples of such programs include verified kernel extensions, cryptographic routines, and fixed-shape machine-learning models. We show that constant boundedness enables a tight, polynomial-time approximation of optimal stack usage by viewing control flow as a tree and applying a tree-scan allocation strategy augmented with memory defragmentation. Our approach guarantees memory usage bounded by the maximum live memory plus, at most, the size of the largest buffer, and is optimal when in-place swapping is permitted. We deploy the proposed allocator in two scenarios. First, in an Elixir-to-eBPF compiler, as a spiller that optimizes stack space. Second, as a static heap allocator for bounded MLIR programs using the Structured Control-Flow dialect. Results demonstrate stack reductions exceeding 90% on real eBPF workloads and show that, even under aggressive code expansion, defragmentation is rarely required and memory usage remains a small fraction of that required by naive allocation strategies.
Read the full abstract

Robotics

Expected Free Energy-based Informative Path Planning for Robotic Mars Exploration

An autonomous robot efficiently exploring an unknown environment, such as looking for water sources on Mars, faces two simultaneous demands: building an accurate information map while quickly finding the regions of greatest value, and paying for every meter of travel and the cost of every measurement it takes. Classical information-seeking and reward-seeking criteria address only one of these objectives at a time. Here, we propose Expected Free Energy (EFE), the principled action-selection objective from active inference, as a unifying criterion for budgeted robotic informative path planning. Maintaining a Gaussian-process belief over the information field, our agent plans continuous trajectories that minimize expected free energy under hard path-length constraints. The results from multiple realizations show that EFE-based planning yields accurate posterior maps and locates the highest-value regions simultaneously, outperforming information-theoretic baselines under the same settings. In robotic exploration, these unified, easy-to-tune principled information-gathering strategies facilitate autonomous deployment while enforcing efficiency and resource constraints.
Read the full abstract

Computation and Language

You Only Pass Once: Answering and Abstaining Together in a Single Forward Pass of a Frozen Language Model

A frozen language model under-uses the evidence in its hidden states and answers confidently when the context contains no answer. We fix both on the residual stream without changing any weight. For accuracy, a small trained network (the probe) adds an input-dependent vector to the residual at several middle layers; it raises Qwen2.5-1.5B on $\alpha$NLI from 51% to 73% accuracy. For abstention, a fixed vector (the direction), the mean hidden state of contexts that contain the answer minus that of contexts that do not, scores the hidden state; a low score abstains. On four QA datasets with human answerability labels (Qwen2.5-7B), moving the direction to a dataset it was not fit on costs under 3 points of decision accuracy (from 0.83 in domain) on average and still beats asking the model whether its context suffices; the same judgment trained as a classifier into the probe's activation edit or into LoRA weights loses 16 and 22 points and ends below asking. We set out to put both functions on a single forward pass, and found that the probe's edit changes the very hidden state the direction scores, degrading it. A small map – trained without labels, only to reconstruct the unedited hidden state from the edited one – removes 63% of that interference on 1.5B. Refit on the hidden-state pairs of LoFiT or LoReFT, the same recipe repairs those methods too: the steering module is replaceable. The resulting system, YOPO, answers and abstains in one forward pass. On $\alpha$NLI it raises three-way accuracy from 0.38 to about 0.75 on 1.5B, and stays within 0.02 of a two-pass reference on 1.5B, 3B and 7B, at half the cost. We also observe improvements on the four QA datasets, where answerability is human-labeled and answers are generated as free-form text. Near-parity with the two-pass reference holds across nine backbones spanning six model families without backbone-specific hyperparameter tuning.
Read the full abstract

All authors Ziyang Luo, Zhongyao Chu, Xinjie He, Youting Wang, Xukui Qin, Runxiong Wu, Yan-Syuan Chen