Artificial Intelligence
Papers
Fresh from arXiv, typeset properly. Highlighted words trade on the market.
All of arXiv
Monday 14 September
Vibe design agents translate natural-language requests into interfaces, yet repeated generations often converge on similar visual directions. We present an inference architecture that makes design exploration an explicit stage before implementation. A pre-pass proposes structured design specifications with typicality weights, an external policy samples a direction, and the generator implements it under fixed decoding settings. Candidate elicitation draws on Verbalized Sampling. Applied to UI themes and visual-asset prompts, this architecture provides direct control over aesthetic exploration without changing code-generation temperature. Across 168 benchmark prompts, theme sampling expands observed selection coverage from one option to nearly three and increases rendered visual variation. For visual-asset prompts, the strongest observed LLM-judge win/loss ratio is 1.32, with only a small change in whole-screen similarity. Localized aesthetic choices can thus influence judged preference even when the overall screen remains similar. We further evaluate joint theme and asset sampling in a randomized online study conducted over nearly two months, covering 3.39 million tasks and 87,416 labeled conversations. These studies show how explicit design specifications support exploration of alternative UI concepts and help us evaluate their effects on rendered interfaces, judged preference, and user interaction.
Machine Learning
Ensemble-Conditioned Molecular Design
Molecular design is typically approached as a problem of finding molecules which can adopt a single bioactive conformation. In reality, molecules occupy a distribution over conformations, and many of the properties which determine whether a candidate is viable depend on that distribution rather than on any single conformer. We reframe molecular design as an optimisation of both the modes and properties of molecules' conformational ensembles, where modes can be represented as shapes, pharmacophore profiles or protein pockets, and properties are aggregate scalars computed over the whole distribution. To realise this we introduce ensemble-conditioned guidance, a framework which conditions 3D molecular generative models on both axes simultaneously. Mode conditions are composed adaptively at inference by combining the vector fields produced under each condition. Conditions may be targeted or avoided, mixed across modalities and combined in arbitrary numbers, allowing a wide range of design tasks to be expressed with a single trained model. We introduce adaptive symmetry learning to allow conditions from different reference frames to be composed, and extend our generative framework to enable flexible-size generation. We evaluate on new benchmarks for multi-mode conditioning and ensemble property optimisation, and apply the framework to two practical drug discovery tasks, dual-target binder design and active-state-selective agonist design, where in both cases conditioning on the additional state improves the desired outcome over single-state conditioning.
Read the full abstract
Spectral Theory
Singular value decomposition of unbounded operators
The singular value decomposition has been established for matrices, Hilbert–Schmidt operators, trace-class operator, compact operators, and bounded operators, but surprisingly not for unbounded operators. Unfortunately, most interesting operators in applied math are unbounded, as any operators involving some form of derivatives — gradient, exterior derivatives, Laplacians, Fourier and other transforms of derivatives, Hamiltonians, etc. — are likely unbounded. In this article, we fill in this last missing piece by establishing the existence of singular value decompositions for unbounded operators in three natural forms: multiplication-operator, direct-integral, and operator-valued-measure. We show it inherits classical properties of finite-dimensional singular value decomposition including approximation results, relationships with fundamental subspaces, and the Moore–Penrose inverse. This discovery opens the door to the singular value decompositions of a myriad of well-known unbounded operators in mathematics, physics, statistics, and finnance — gradients on Euclidean spaces and manifolds, Petrov–Galerkin method, finite-difference operators, Hilbert–Schmidt operators, Hilbert complexes, supersymmetric quantum mechanics, Sturm–Liouville theory, nonparametric density estimation, and the Black–Scholes equation. The resulting decompositions reveal a number of novel insights, among many others: bosonic and fermionic states in supersymmetric quantum mechanics arise as left and right singular vectors of generalized ladder operators; the Riesz transform appears as the left singular operator of the Euclidean gradient; and the Hodge decomposition follows directly from the singular value decompositions of the exterior derivatives.
Read the full abstract
Complex Variables
Partial positivity is not preserved by marginalization
In this paper, we show that partial positivity is not preserved by marginalization in both the real and complex settings. We also give a sufficient condition for the preservation of partial positivity under marginalization.
Plasma Physics
Shape effects and Shafranov shift reversal in analytical Grad–Shafranov equilibria with non-convex boundaries
Analytical Solov'ev equilibria with freely prescribed plasma boundaries can be constructed by enforcing the boundary condition in a least-squares sense on a polynomial basis of homogeneous solutions. Within this approach, a systematic scan of boundary shape is carried out well beyond the convex regime: non-convex star polygons of arbitrary symmetry order and concavity are examined alongside conventional convex tokamak cross-sections, using a quantitative boundary-fidelity criterion to delimit the range of shapes the polynomial basis can represent. For standard convex shapes, poloidal beta is insensitive to shaping due to the Solov'ev current profile, while polygon boundaries raise it monotonically with the number of sides, driven by the flat sides compressing flux surfaces toward the axis. For the Shafranov shift, odd-fold boundaries admit a critical concavity below which the geometric centre of the boundary overtakes the magnetic axis, reversing the sign of $\Delta/a$; no such reversal occurs for even-fold boundaries, and the effect has no analogue among convex shapes.
Read the full abstract
Analysis of PDEs
Sharp non-uniqueness for the hyper-viscous compressible 3D Oldroyd-type models: Beyond the Lions exponent
We prove the non-uniqueness of weak solutions to the three-dimensional hyper-viscous compressible Oldroyd-type model with general pressure laws and Newtonian viscous. Using an intermittent convex integration scheme, we construct non-unique weak solutions in $L_t^\gamma W_x^{s,p}$ with $(s,p,\gamma)$ lies in the supercritical regime relative to the Ladyzhenskaya-Prodi-Serrin criteria. In addition, the Newtonian viscosity we considered contains the fractional viscosity $(-\Delta)^\alpha$, where the viscous exponent $\alpha$ can be larger than the Lions exponent $5/4$. In the classical viscous case $\alpha=1$, our construction also yields sharp non-uniqueness in $L_t^pL_x^\infty$ for $1<p<2$. Furthermore, we obtain two singular-limit results. First, we prove the strong vanishing viscosity limit for any weak solutions in $H_{t,x}^{\widetilde{\beta}}$ to the compressible fundamental elastodynamic system. Second, $H_{t,x}^{\widetilde{\beta}}$ weak solutions to the incompressible Oldroyd-type model can be obtain as a low Mach number limit (incompressible limit) of a sequence of weak solutions to the compressible Oldroyd-type model. The main ingredient of our result is a new cancellation mechanism that overcomes the difficulty caused by the relative rigidity of the pressure. To the best of our knowledge, this is the first non-uniqueness results of weak solutions to the hyper-viscous compressible hydrodynamics equations.
Read the full abstract
Machine Learning
Branched Optimal Transport Amortization
Methods of Branched Optimal Transport (BOT) mimic the economy and efficiency of natural tree-like structures, such as those found in rivers and biological systems. These methods are widely applicable for designing efficient networks in society, from river basins and blood vessels to mail and gas distribution systems. However, they remain understudied in the context of designing deep generative models, particularly at a large scale. Standard continuous-time generative models, such as the flow matching approach, fail to capture the inherent hierarchical and branching patterns present in real-world data. Current models provide no mechanism for flows to merge or share pathways to minimize total transport cost. Inspired by the "economy of scale" principle in BOT, we introduce a novel, scalable branched flow-matching algorithm designed to solve the branched optimal transport problem in high dimensions. Our method adapts the Benamou-Brenier continuous-time optimal transport formulation to learn branched generative flows. These flows allow probability mass to aggregate along common pathways before branching out to diverse targets. Parametrized by neural networks, our method effectively learns complex branched generative processes. We demonstrate its effectiveness on challenging high-dimensional tasks in biology and image generation.
Read the full abstract
Combinatorics
The Komlós conjecture for complex discrepancy
The Komlós conjecture is a classic problem in discrepancy theory; it asks whether an absolute constant $K$ exists such that given any $n$ vectors $a_1,\ldots,a_n$ inside the $m$-dimensional Euclidean ball, regardless of how large $m,n$ are, there is always a selection of signs $\varepsilon_1,\ldots,\varepsilon_n$ guaranteeing $$\|\varepsilon_1a_1+\ldots+\varepsilon_na_n\|_\infty \leq K.$$ We show that if the $\varepsilon_i$'s are allowed to take not just the values of $\pm 1$ but any unit modulus complex number, which we refer to as complex discrepancy, then the above inequality holds for a finite, explicit constant $K_{\mathbb{C}}$. Here, the $\ell^\infty$ norm of the resulting vector in $\mathbb{C}^m$ is the largest modulus of its entries, and thus the complex discrepancy of real vectors is equivalent to their rank-$2$ vector discrepancy. This quantity provides an upper bound (modulo uniform constant prefactor) on Gaussian discrepancy – a discrepancy measure introduced by Chewi, Gerber, Rigollet and Turner. Thus, we also resolve the Komlós conjecture for Gaussian discrepancy. Our paper builds upon the recent work of Bansal and Jiang on the Beck-Fiala and Komlós conjectures, which we approach from the formalism of Burkholder and the Bellman function method from probability and harmonic analysis. Our work was in part motivated by the realization that the complex discrepancy of the columns of any unitary matrix is equal to 1, a fact that follows from a straightforward calculation based on Idel and Wolf's generalization of the Sinkhorn normal form for unitary matrices.
Read the full abstract
Computation and Language
DA-DLM: Explicitly Modeling Token Dependencies in Diffusion Language Models
Diffusion Language Models (DLMs) generate text by iteratively denoising a masked sequence, independently predicting multiple tokens at each step. This conditional independence discards inter-token dependencies and degrades coherence-an issue that parallels the multi-modality problem in Non-Autoregressive Translation (NAT). Drawing on the Directed Acyclic Transformer (DAT), which tackles this problem in NAT via a Directed Acyclic Graph (DAG), we propose DA-DLM, a model that adapts DAG-based dependency modeling to DLMs' iterative setting through a position-oriented DAG design. The position-oriented DAG binds node groups to fixed output positions so that tokens fixed in earlier steps anchor neighboring predictions via learned transitions, and evolves with denoising to focus on remaining uncertainty as anchors accumulate. On language modeling, open-ended generation, and summarization, DA-DLM consistently outperforms Block Diffusion, especially under fewer denoising steps, and matches autoregressive models while preserving the parallel generation advantage. Our code is publicly available at https://github.com/jipy0222/DA-DLM.
Read the full abstract
Analysis of PDEs
Existence of pullback \(\mathcal{D}\)-attractors for Kirchhoff wave equations with strong damping and delay
This paper studies the existence of pullback $\mathcal{D}$-attractors for non-autonomous Kirchhoff wave equations with strong damping and delay effects in the phase space $\mathcal{E} = C_{H^1_0(\Omega)} \times C_{L^2(\Omega)}$. The model contains a strong damping term $-\Delta \partial_t u$, a nonlocal Kirchhoff term $\Psi(\|\nabla u\|^2)$, a state-dependent delay term $\phi(t, u_t)$, and a time-dependent external force $h(x, t)$. There appear to be no results on the pullback attractors of Kirchhoff wave equations involving both strong damping and delay effects in the literature. To this end, we establish delicate uniform estimates and employ the contraction function method to prove the pullback asymptotic compactness of the associated process, which yields the existence of pullback $\mathcal{D}$-attractors.
Read the full abstract
Algebraic Geometry
Ulrich Bundles on certain Grassmann Bundles Over Curves
Let $E$ be a slope-semistable vector bundle on smooth projective curve $C$. Assume, $ \pi : G= Gr_{C}(r, E) \to C$ denote the relative Grassmannian of rank $r$ quotients of $E$. We prove that for every very ample polarization $L= O_{G}(1) \otimes \pi^{*} A$ the polarized variety $(G,L)$ possess an Ulrich bundle. The construction globalizes homogeneous ulrich bundle on Grassmannian fibres. The remaining cohomology vanishing problem reduced to an acyclicity criterion on the base curve. We finally use Falting's cohomological charaterization of semistability.
Read the full abstract
Sound
Rethinking Procedural Audio Pre-training: Source Scaling and Objective Adaptation
Procedural audio has emerged as a viable source for transferable audio representation learning, but its design principles remain unclear.We revisit two questions: how a procedural source should be scaled, and whether training choices developed on natural audio should transfer unchanged to procedural data.Using a controlled source, we separate scale into formula-class coverage C and within-class rendering diversity I.Experiments with FDSL and AudioMAE show that these two forms of scale provide different benefits and depend on the learning formulation and downstream task. A matched AudioMAE study further shows that procedural audio favors low mask ratios (10%–25%), whereas AudioSet-28K favors 50%–75%. Shared-codebook analysis reveals lower patch diversity and stronger temporal predictability in procedural audio. These results motivate source-aware procedural pre-training, where source scaling and learning configuration are considered jointly.Code is available at https://github.com/Cross-Innovation-Lab/Formula-Bank.
Read the full abstract
Computation and Language
Salesforce Koa: An Enterprise Language Model for Agentic Tool Use
We present Salesforce Koa, an enterprise language model built by post-training the open-weight Nemotron-3-Super-120B foundation model with reinforcement learning using Group Relative Policy Optimization (GRPO), and deployed in FP8 for production. Koa is trained only on public and synthetically generated data, and specialized for the agentic tool use that enterprise workflows demand: routing a request to the correct action, invoking the right tool with valid arguments, and completing multi-turn business tasks. The distinctive component of our pipeline is specification-driven task construction: declarative Agent Script specifications are expanded into persona-conditioned multi-turn environments whose rewards are grounded in successful tool use. Applied to enterprise CRM specifications, the same pipeline produces the in-domain training distribution on which Koa is specialized. On CRMAgentBench and the human-labeled production tool-calling set, Koa outperforms both its untuned open-weight base and GPT-4.1 and is competitive with the strongest frontier models. It reaches 87% Task Success Rate on CRMAgentBench (vs. GPT-4.1 at 82% and the base at 79%), is at or near the top of every metric on the human-labeled portion of an internal production benchmark, and preserves the base model's general capability on public benchmarks (Tau2Bench, BFCL). A controlled comparison with architecture and RL recipe held fixed shows that the additional in-domain RL stage improves argument accuracy and full tool-call success on the human-labeled enterprise benchmark.
Read the full abstract
Machine Learning
Sensory Precision Inference for Multimodal Arbitration under Uncertainty
Autonomous agents operating on multisensory data cannot assume that all sensory modalities remain consistently informative. In real environments, sensory streams are frequently corrupted by noise, missing data, or inter-modal incongruence, requiring adaptive arbitration between competing sensory hypotheses. While active inference provides a principled framework for uncertainty-guided inference, the role of dynamically inferred sensory precision in generative multimodal arbitration under sensory conflict remains comparatively underexplored. We propose a multimodal perceptual inference model in which latent beliefs and modality-specific sensory precisions are jointly updated through iterative free-energy minimization. In our proposed model, sensory precision dynamics not only reflect sensory uncertainty but actively shape the evolution of latent beliefs during multimodal conflict. In addition, we introduce a learned prior over sensory precisions that induces structured, class-dependent precision patterns and influences cross-modal inference dynamics. We evaluate the model using a synthetic multimodal MNIST dataset combining visual, auditory, and tactile representations of digit classes under controlled sensory noise and inter-modal incongruence. Results show that dynamic precision inference improves reconstruction robustness under corrupted sensory evidence, supports coherent latent inference from reduced sensory evidence, and enables stable arbitration between conflicting modalities. Furthermore, learned precision priors generate interpretable precision structures that shape inference dynamics and cross-modal latent structure. These findings support sensory precision inference as a mechanistic control process for adaptive multimodal belief formation under uncertainty, highlighting precision dynamics as a computational mechanism for robust and interpretable multisensory integration.
Read the full abstract
Machine Learning
What Does an LLM Learn from Reinforcement Learning? A Mechanistic Interpretability Perspective with Fixed-SAE Track
Reinforcement learning (RL) is widely utilized in large language model training to improve targeted capabilities, yet how RL reshapes a model remains poorly understood. Prior attempts to explain how RL works largely offer behavioral perspectives, leaving open what RL gives a model at the representation level: can RL create genuinely novel features, and which existing features does it enhance or suppress? Recent developments in mechanistic interpretability suggest sparse autoencoders (SAEs) as a promising lens to decompose internal activations into human-interpretable features; however, they cannot be directly applied to tracking change across training. In this work, we introduce Fixed-SAE Track, a framework that trains one shared SAE per considered layer on activations pooled across the base model and all RL checkpoints, holding every feature direction fixed so that representation shifts are rigorously defined through the activations of interpretable SAE latents, including the detection of emerging novel features. Validated across multiple datasets and RL algorithms, we find that RL-induced drift is small, gradual, concept specific, and concentrated in late layers, mainly enhancing the sampling rates of a small set of ladder tokens, formatting scaffolding such as step breaks and answer delimiters, rather than reshaping problem content. Steering these features into the base model recovers around 80% of RL's performance gain, suggesting that RL primarily elicits capabilities the model already possesses, much as steering does. We further design a synthetic benchmark with features known by construction to test whether RL can instill genuinely novel features. We believe Fixed-SAE Track provides a principled approach to tracking representation shifts and offers representational evidence for understanding how reinforcement learning changes the inner representation of LLMs.
Read the full abstract
Computational Complexity
Randomized query complexity can beat certificate complexity
A long-standing open question in query complexity asks whether there is a total Boolean function f with R(f) << C(f), where R(f) and C(f) denote its bounded-error randomized query complexity and certificate complexity, respectively. We construct a function with R(f) = O~(sqrt{C(f)}), which is optimal up to log factors. The same function also has $Q(f) = O~(C(f)^{1/4}), where Q(f) is the bounded-error quantum query complexity of f, which is also nearly optimal.
Read the full abstract
Computational Complexity
The Exact Growth Rate of Space-Optimal Reversible Pebbling on Chains
We determine the exact time exponent of space-optimal reversible pebbling on chains as $1.331742379256310\ldots$. The growth rate of space-optimal reach exists as a limit and admits a variational formula. The same exponent governs complete computations at minimal space, uniformly in the chain length.
Read the full abstract
Strongly Correlated Electrons
Non-Abelian chiral spin liquid in a spin-$1$ antiferromagnet on the square lattice
Non-Abelian chiral spin liquids (CSLs) host non-Abelian anyon excitations and are promising platforms for topological quantum computation. However, realizing non-Abelian CSLs in short-range interacting spin systems remains elusive. Here, we study a spin-$1$ square-lattice model with the first- and second-neighbor bilinear ($J_1,J_2$) and biquadratic interactions ($K_1,K_2$), as well as the three-spin scalar chiral coupling $J_{\chi}$. Inspired by the evidence of the Moore-Read state in exact diagonalization and infinite Projected Entangled Pair States calculations, we fix $J_1=1.0$, $J_2/J_1=0.623$, $K_1/J_1=-0.176$, and obtain a quantum phase diagram for $0 \leq K_2/J_1 \leq 0.4$ and $0.3 \leq J_{\chi}/J_1 \leq 0.6$ by using the density matrix renormalization group (DMRG) calculations. We identify a stripe antiferromagnetic phase, a Néel antiferromagnetic phase, a magnetically ordered chiral spin state phase, and a non-Abelian CSL phase emerging near the boundaries of the ordered phases. The unbiased DMRG results provide strong evidence for the Moore-Read state, including the three topological sectors, the quantized spin Chern number $C=1$, and the level counting of low-lying entanglement spectra which is consistent with the chiral SU(2)$_2$ conformal field theory. Our results may shed new light on searching for non-Abelian CSLs in other higher-spin ($S > 1/2$) systems with short-range couplings.
Read the full abstract
Strongly Correlated Electrons
Observation of complex orbital order in MnV2O4
Orbital ordering in vanadium spinel oxides with a geometrically frustrated pyrochlore structure has been a subject of controversy, owing to competing theoretical models and the absence of direct experimental evidence. Here we combine high-precision single-crystal synchrotron x-ray diffraction with core differential fourier synthesis to visualize the valence electron density (VED) of the orbital-ordered ground state in real space. By carefully investigating multiple-scattering artifacts, we identify the low-temperature structure as belonging to the I41/amd space group. The reconstructed VED around the V sites reveals an orbital-ordered state distinct from both previously proposed real- and complex-orbital models. Our results resolve the long-standing controversy in MnV2O4 and establish a route to identifying orbital states in frustrated spin-orbital systems.
Read the full abstract
Quantum Physics
Fermionic quantum error correction is never free
Fermionic platforms offer compelling architectures for quantum computing, ranging from topologically protected Majorana-based qubits to fermionic cold atoms. To achieve scalability, however, they require quantum error correction. In this work, we prove that any exact and sufficiently accurate approximate fermionic quantum error correction necessarily requires non-Gaussian operations, beyond the free-fermion regime of quadratic dynamics. This is in sharp contrast to the qubit setting, where the efficiently classically simulable stabilizer operations form the standard framework for quantum error correction. Specifically, we show that the logical space of any non-trivial fermionic error-correcting code contains no pure fermionic Gaussian state, utilizing a fundamental incompatibility between fermionic error correction and Wick's theorem. We further show that the required number of bounded-weight non-Gaussian gates for unitary codeword preparation grows at least linearly with both the code distance and the number of encoded modes, revealing an intrinsic resource overhead that increases simultaneously with error-protection strength and logical capacity. Furthermore, we analyze the performance of fermionic Gaussian operations in entanglement distillation, revealing a distinction from their bosonic counterparts. Our results reveal fundamental difficulties for fermionic error correction from the perspectives of both physical implementation and classical simulation, suggesting connections to fermionic phases of matter and state preparation complexity.
Read the full abstract
Databases
Fast Label-Filtering Approximate Nearest Neighbor Search via Progressive Label Set Stratification
Approximate nearest neighbor search (ANNS) retrieves the most similar vectors to a query vector in high-dimensional space. Label-filtering ANNS (LFANNS) extends ANNS with a label filter that the labels of base vectors must satisfy a set relation (e.g., equality, containment, or overlap) with the query labels. Existing LFANNS indices suffer from inconsistent performance across different filter types and degraded scalability under varying label scale and distribution. In this paper, we define label-stratified similarity graph (LSSG), where edges connect neighboring vectors whose label sets fall within stratified similarity thresholds. To implement LSSG efficiently, we design an incremental insertion algorithm to prune redundant edges in both vector and label spaces, and leverage a MinHash structure to ensure scalability for large-scale labels. We analyze stepwise probabilities under explicit label models and explain why stricter label tiers reduce ineffective in-filtering expansions. Benchmark experiments show that LSSG achieves ideal optimality for equality queries, and 1.06x-92.9x and 1.08x-84.1x faster than the best competing index for containment and overlap, respectively, in query speed with identical accuracy and 0.35x index size.
Read the full abstract
Operator Algebras
Applications of the Projection Characterization and Symmetry Bootstrap for Operators that are Nearby Commuting Operators
This paper applies the Projection Characterization and the Symmetry Bootstrap for two operators $A, B$ (with $A$ unitary or self-adjoint) nearby commuting operators $A', B'$ (with $A'$ unitary or self-adjoint) from arXiv:2412.20795 to problems related to structured almost commuting unitaries, self-adjoint operators that almost anti-commute, and unitaries that almost commute up to a rational phase. We prove several new stability theorems for these types of relations for matrices. We also provide estimates for how close the constructed operators are for several known stability theorems that did not previously have asymptotic estimates.
Read the full abstract
High Energy Physics (Phenomenology)
Open-channel radiation zeros and nonlinear damping of a critical-bubble internal mode
We study radiation and nonlinear damping of a spherically symmetric localized mode of an $O(3)$ thermal critical bubble. The leading second-harmonic radiation amplitude is determined by the overlap between a nonlinear source and a continuum scattering wave. This overlap can vanish through destructive interference as the supercooling is varied, even when the second harmonic lies above the continuum threshold. A signed-overlap scan resolves five zeros in a finite parameter interval, including three on the thinner-wall side of the two representative roots studied in detail. For these representative roots, half-line Jost–Green calculations, threshold-preserving wall deformations, and constrained radial evolution verify the cancellation and its robustness. Near a selected root, fourth-order perturbation theory predicts an $A^2$ displacement of the finite-amplitude radiation minimum and an $A$-linear width of the third-harmonic-dominated region; radial evolution quantitatively tests both relations. At the selected zero, third-harmonic loss predicts $A\sim\tau^{-1/4}$, instead of the generic $A\sim\tau^{-1/2}$ law. Measured harmonic powers and a slow-envelope reconstruction support this hierarchy. Independent center-stable shooting recovers the unprojected third-harmonic coefficient, which differs from the fixed-projector value. The results connect a tunable radiation form factor to nonlinear relaxation of an internal mode on an unstable nucleation saddle.
Read the full abstract
Signal Processing
Bridging the Gap in ECG-Based Emotion Recognition: A Unified Evaluation of Deep Learning Models
Deep learning has led to numerous proposed architectures for Automated Emotion Recognition (AER) from electrocardiogram (ECG) data, but inconsistencies in preprocessing, training, and evaluation make direct comparisons difficult. Most studies train and validate models on individual datasets collected under homogeneous conditions, limiting variability and raising concerns about generalizability. Cross-dataset validation is sometimes used but primarily assesses model adaptability rather than true generalization. This study presents a comparative analysis of prominent deep learning architectures in AER, emphasizing model generalization over dataset adaptability. To enable this benchmark, we introduce two open-source frameworks: Affective Research on Representations and Classifications (ARRC), a standardized benchmarking toolkit, and Affective Research Dataset Toolkit (ARDT), a framework for inter-dataset training and validation. Using ARDT, we consolidate three publicly available AER datasets, CUADS, ASCERTAIN, and DREAMER, into a single dataset, increasing variability in sensor types, recording conditions, and participant demographics. We then use ARRC to evaluate three widely studied deep learning models and two CNN baselines through hyperparameter optimization and 10-fold cross-validation. Our findings provide insights into the trade-offs between classification accuracy and model complexity, establishing a reproducible benchmark for AER research. All source code for ARRC, ARDT, and model evaluation is publicly available to ensure transparency and facilitate further research.
Read the full abstract
Artificial Intelligence
BusMA: A Bus Communication Substrate for Multi-Agent Systems
Multi-Agent (MA) systems are effective at solving complex tasks that demand planning, tool use, and the synthesis of evidence from multiple sources. Existing systems typically adopt Hierarchical Manager-Worker (HMW) or Router-based Message Passing (RMP) structures as their communication protocol. However, these designs restrict agent autonomy: Worker agents cannot directly consult specific "peers", and misrouted messages can propagate errors. Inspired by bus architectures in computer systems, we propose BusMA, a communication framework that allows any agent to address other agents through a shared channel, i.e., the Bus. It consists of agent registration, message routing, and shared memory management components. Worker agents, each equipped with tools, have their own local memory and can reason, act (tool usage), and communicate by posting shared messages with specific intents. We introduce four intents: discussion, challenge, guidance, and request for explanation, which support fine-grained communication among agents. A Chair agent monitors the shared memory to coordinate interactions and facilitate convergence among Workers. To evaluate the effectiveness of BusMA, we conduct extensive experiments with two frontier LLMs across 13 tasks spanning visual reasoning, mathematical reasoning, and knowledge retrieval. The results demonstrate that BusMA consistently outperforms state-of-the-art HMW and RMP methods.
Read the full abstract
High Energy Physics (Experiment)
Improved amplitude analysis of $\eta^\prime\to\pi^+\pi^-\pi^0$ and $\eta^\prime\to\pi^0\pi^0\pi^0$
Using a sample of $(10087\pm44)\times 10^6$ $J/\psi$ events collected with the BESIII detector at BEPCII, we perform an amplitude analysis of the decays $\eta^\prime\to\pi^+\pi^-\pi^0$ and $\eta^\prime\to\pi^0\pi^0\pi^0$, where we observe significant $\pi^\pm\pi^0$ $P$-wave and $\pi$-$\pi$ $S$-wave interactions. Two different parameterizations, a $\pi$-$\pi$ scattering phase shift and the Gounaris-Sakurai Breit-Wigner formalism, are used to describe the $P$-wave propagator. Due to the large interference, the branching fractions for both the $P$- and the $S$-waves are found to be strongly model dependent.
Read the full abstract
Number Theory
Laurent coefficients of Zagier-type zeta function à la Ishibashi and arithmetic aspects of extended Ramanujan period function
One of the remarkable contributions of Don Zagier was the Kronecker limit formula for a real quadratic field, where he connects the double series $\mathcal{Z}(s,w,w^\prime)$ to the Dedekind zeta function associated to a real quadratic field. Later, Ishibashi determined all the Laurent coefficients of $\mathcal{Z}(s,w,w^\prime)$ at $s=1$. Recently, Choie and kumar have studied the analytic behaviour of the analogous double series $\tilde{\mathcal{Z}}(s,w,w^\prime)$. In this article, we derive all the Laurent coefficients of $\tilde{\mathcal{Z}}(s,w,w^\prime)$, akin to Ishibashi. These Laurent coefficients involve an interesting function $\mathfrak{F}_k^0(x)$, which was earlier studied by Dixit et. al. (Ramanujan for $k=1$), where they obtained a beautiful symmetric relation for $\mathfrak{F}_k^0(x)$. We establish both the two term and the three term functional equation of $\mathfrak{F}_k^0(x)$, derive the action of the period-like Hecke operator on $\mathfrak{F}_k^0(x)$ and connect an important integral with $\mathfrak{F}_k^0(x)$.
Read the full abstract
Machine Learning
Not All Prompts Are Equal: Exploration-Guided Prompt Scaffolding for Multimodal Reinforcement Post-Training
Training prompts in online reinforcement learning (RL) differ substantially in how informative they are for the current policy: some are already saturated while others are too difficult to yield reliable learning signals, yet both receive equal rollout budget under standard training. We propose an exploration-guided prompt scaffolding framework that adapts the training prompt distribution dynamically throughout RL post-training of multimodal large language models (MLLMs). Central to our approach is the $\textit{Exploration Potential Score} (EPS)$, a lightweight rollout-based proxy for prompt utility derived from KL-regularized policy improvement theory, computable directly from on-policy rollout statistics without additional overhead. Rather than discarding low-utility prompts, we use a teacher model to generate scaffolded rewrites that preserve the original task intent while making subsequent training more informative, reframing teacher supervision as training-data refinement rather than output imitation. Integrated with GRPO on Geo3K and MMK12, our method consistently outperforms the baseline on both in-domain and out-of-distribution benchmarks, achieving up to 9.7% relative improvement in-domain and gains of 11.5% on MathVision and 11.1% on MMMU-Pro.
Read the full abstract
Materials Science
Scalable construction of force-constant basis sets for large-scale anharmonic lattice-dynamical calculations
A projector-based formulation of force constants in crystalline materials provides a systematic framework for constructing force-constant bases and determining force constants from force–displacement datasets while rigorously satisfying crystal symmetry, permutation symmetry, and translational invariance. In this work, we develop an efficient eigenvalue solver for projection matrices and integrate it into the projector-based framework. The proposed method substantially reduces the computational cost and enables practical calculations for large-scale systems, low-symmetry crystals, and higher-order force constants that are difficult to treat using conventional approaches. Applications to self-consistent phonon calculations for assessing the grain-boundary excess free energy, lattice thermal conductivity calculations in complex compounds, and fourth-order force-constant estimations demonstrate the efficiency and robustness of the proposed framework for large and complex materials systems.
Read the full abstract
Machine Learning
Ensemble Complexity in Photovoltaic Forecasting
An ensemble can improve photovoltaic forecasts while adding components that contribute little or increase computation. We assess these effects through matched comparisons and ablations of a fixed heterogeneous predictor bank. Hourly experiments use GEFCom2014 and three additional public datasets, with chronological partitions and three seeds. Under retrospective ERA5 assistance, static fusion reduces scaled mean absolute error against matched boosting by 1.11%, 4.41%, and 1.63% on PVDAQ, OPSD, and Ausgrid; only OPSD remains supported after multiple-comparison correction. Weather gating offers no consistent incremental benefit. Exploratory member removals show group-level dependence alongside individual redundancy. A separate, previously inspected fifteen-minute case replaces one neural member with a tree predictor: normalized error falls by 1.72%, but measured inference is slower. These findings support component-wise evaluation with explicit limits on weather availability and test-set reuse.