Machine Learning
Dyad: Extending Large Language Models with Native Typed Decision-Making
Yundaichuan Zhan, Weishi Wang, Wenbiao Liu et al. (+5 more)
We study how to build more capable general-purpose
agents by extending large
language models (LLMs) with native typed decision-making. We introduce Dyad, an architecture that augments a pretrained LLM with an environment-conditioned action encoder that embeds each candidate action description in parallel, then scores these
embeddings against the LLM's internal state to yield a distribution over typed actions. By factorizing decision-making into representations of the evolving interaction state and environment-specific action semantics, Dyad introduces an inductive bias for learning reusable representations while keeping action scoring efficient even as the action space grows. We investigate two complementary reinforcement learning settings driven by environment interaction. With the LLM frozen, training the action encoder alone achieves consistent gains across four unseen environments, enabling modular adaptation without modifying any LLM parameters. Jointly optimizing both components outperforms conventional RL post-training across diverse interactive tasks and model scales, including a 3.80% average absolute gain on ALFWorld with a 9B model, while improving general knowledge,
reasoning, and coding.
Read the full abstract ▾
All authors Yundaichuan Zhan, Weishi Wang, Wenbiao Liu, Daniel Dahlmeier, Chengwei Qin, Juncheng Li, Fredrik D. Johansson, Zhongqi Yue
Machine Learning
Koa-decision: Fast and Consistent Structured Decision Making with Generative LLMs
Shenghong Dai, Shiva Kumar Pentyala, Yingchi Liu et al. (+6 more)
Industry applications often demand low-latency classification, yet current large
language model (LLM) approaches remain poorly suited for latency-critical applications. Existing prompting and constrained decoding produce verbose, multi-
token outputs that require expensive
token-by-
token generation, while encoder-based models achieve faster inference but sacrifice task flexibility. We propose Koa-decision, a framework for low-latency atomic actions – fast, single-step decisions such as classification, semantic endpointing, Boolean checks, and scoring – formulated as constrained generation with single-
token outputs. By introducing atomic label
tokens and applying supervised fine-tuning, our method reduces classification to a deterministic one-step decoding problem. Across standard
benchmarks, Koa-decision delivers competitive accuracy with consistently low and stable latency. On a production intent-routing
benchmark, Koa-decision reaches 85.5% accuracy – competitive with the strongest frontier models (Claude-4.8-Opus, Gemini-Pro-3.1) and ahead of GPT-5 and Gemini-2.5-Pro – while answering in about half a second, several-fold faster than every frontier model (up to ~7.5x at the median) under identical serving conditions. Against the dedicated single-
token system Jev/TypeSafe, Koa-decision is competitive on accuracy and faster at the median, while also handling
multimodal inputs and multi-label outputs that single-label text systems do not.
Read the full abstract ▾
All authors Shenghong Dai, Shiva Kumar Pentyala, Yingchi Liu, Shubham Mehrotra, Suman Banerjee, James Zhu, Bin Bi, Sitaram Asur, Phil Mui
Hardware Architecture
Automated Pre-Silicon Verification of High-Speed DDR5 and LPDDR5/6 Memory Controllers: Closed-Loop Timing, Mode Register, and PHY Synchronization in UVM
Manan Patel and Anirban Majumder
External
memory interfaces (such as LPDDR5/4 and DDR5) are essential components of modern mobile, cloud, and enterprise computing systems. While
memory manufacturers focus on physical DRAM die development, the vast majority of
semiconductor firms design custom application ASICs that require a dedicated
Memory Controller to interface with these standardized external
memories. Consequently, pre-silicon design verification of the
Memory Controller RTL – acting as the Device Under Test (DUT) against a third-party DRAM Verification IP (VIP) – is a ubiquitous and critical challenge across the global
semiconductor industry. This article presents an automated, pre-silicon configuration and closed-loop initialization framework for
Memory Controller verification. The proposed solution parses JEDEC timing and configuration parameters directly from the DRAM VIP's database files (such as Denali SOMA files) to configure the
Memory Controller DUT's registers, while a custom Mode Register Register Abstraction Layer (MR-RAL) tracks volatile DRAM VIP states in real-time. Furthermore, a dynamic, re-compilation-free PHY initialization flow randomizes interface parameters directly in the testbench, executes on-the-fly configuration generation via system calls, and parses the resulting register write sequences at runtime. By integrating these automated methodologies into pre-silicon verification flows, engineers can eliminate setup overhead, prevent false protocol violations, and enable comprehensive randomized testing of complex PHY and
Memory Controller configurations.
Read the full abstract ▾
Machine Learning
LoopICL: Looping a single transformer block to solve tabular tasks
Amir Rezaei Balef and Katharina Eggensperger
Tabular foundation models using in-context learning have recently surpassed gradient-boosted trees on predictive tabular tasks. However, recent mechanistic insights suggest that parameters in these models are largely redundant. We introduce LoopICL, a looped
transformer whose core design decouples parameter count from computational depth. LoopICL consists of a single block, processing data through two coupled streams: a
cell stream capturing per-
cell feature representations and a row stream capturing in-context example representations, jointly refined through within-column and cross-column
attention. During pre-training, we vary loop counts, allowing the block to be unrolled for a varying number of iterations at test-time and use a learned exit-gate to automatically exit. In its standard setting, LoopICL performs competitively with TabICLv2 on TabArena and TALENT at the same computational cost (FLOPs), while using nearly 90% fewer parameters. Furthermore, its recurrent design enables users to also trade off inference cost and performance, providing a resource-aware TFM.
Read the full abstract ▾
Robotics
Scouting the Dynamics Gap: Test-Time Policy Adaptation via Action-Outcome Feedback
Yishu Li, Liyuan Geng, Xinyi Mao et al. (+2 more)
While pretrained robotic policies exhibit impressive capabilities in controlled environments, unobserved physical properties and dynamics require these policies to rapidly adapt during deployment. Existing test-time adaptation methods typically rely on sparse scalar rewards, failing to exploit the rich
geometric and dynamic feedback from the environment during physical interaction. To address this challenge, we propose SCOUT, a dynamics-aware meta-learning framework that enables manipulation policies to rapidly adapt by continuously revising their internal beliefs about environment dynamics. Our approach couples an action-prediction policy with a forward dynamics model via a shared belief latent space. During meta-training, an inner loop updates this shared belief latent by minimizing the dynamics prediction error against the observed action outcome, while the outer loop optimizes the network for action selection. At deployment, this structure allows the
agent to infer and adapt to unknown physical dynamics on the fly. By updating its latent belief based on action-outcome mismatches, the policy automatically adapts without risking catastrophic forgetting. We demonstrate that SCOUT significantly accelerates online adaptation across simulated manipulation
benchmarks and achieves robust sim-to-real transfer in the real world. Project webiste can be found here: https://liy1shu.github.io/SCOUT/
Read the full abstract ▾
All authors Yishu Li, Liyuan Geng, Xinyi Mao, Amber Li, David Held
Machine Learning · Cross-listed from eess.SP
Fundamental Limits of Transferability and Equivariance in Algebraic Signal Models I: Finite Dimensions
Alejandro Parada-Mayorga
We study the fundamental limits of transferability in algebraic signal processing through homomorphisms between algebraic signal models. Homomorphisms are linear maps between the signal spaces of two models that commute with filtering, so filtering a signal and transferring it across domains can be done in either order. The existence of such maps is governed entirely by coincidences among the filtered
eigenvalues of the two models' shift operators, but existence alone is insufficient: the space of homomorphisms always contains trivial elements that destroy all information. We introduce the spectral transfer efficiency $\eta(\theta)\in[0,1]$ to quantify information-preserving quality, prove that every homomorphism decomposes into unconstrained blocks over coincidence classes, derive the dimension of the homomorphism space, and characterize exactly when lossless transfer is achievable. Beyond normal shift operators, we quantify a departure-from-normality penalty and show how filter derivatives can repair spectral defectiveness. The theory yields concrete consequences in three settings: for sampling,
eigenvalue interlacing converts transferability under subsampling into an explicit filter design constraint; for compressed sensing, $\eta(\theta)$ controls the restricted isometry constant and the coherence of the resulting measurements, and recovery decouples across coincidence classes; and for machine learning, spectral aliasing emerges as the controlled
symmetry breaking that makes transfer between mismatched domains possible at all.
Read the full abstract ▾
Artificial Intelligence
An Exact Generate - Transform Decomposition of Small-LLM Team Scaling Across Orchestration Architectures
Blaz Bertalanic and Carolina Fortuna
Replacing one LLM
agent with a collaborating team can raise accuracy, but whether scaling the team helps, and which architecture to scale, is unclear. Sweeping eight
agent orchestration architectures across five instruction-tuned 7-9B models, five short-answer
benchmarks, and an executable-code
benchmark up to 30 calls, we find that the returns to team scaling are sharply task-dependent: from three to thirty calls accuracy rises by up to 17 points on the two arithmetic word-problem
benchmarks (GSM8K, GSMHard) but by at most four on ARC, GPQA, and MMLU, for every architecture, a split the usual task-averaged number conceals. Proposer-Critic captures the arithmetic gains, scaling steepest and, in aggregate, surpassing every other architecture at the largest budget (item-clustered intervals exclude zero), though it ranks among the weakest elsewhere, and no architecture wins across tasks. We explain these trajectories with an exact generate-transform decomposition. Partitioning any workflow into proposal coverage and a downstream transform, any accuracy change splits exactly into an extensive coverage dividend and an intensive transformation change. The decomposition diagnoses each task: arithmetic offers coverage headroom that a critic-guided transform converts, whereas the multiple-choice
benchmarks either saturate in coverage or fail to convert it, and on open-ended code generative recovery nearly vanishes so accuracy tracks coverage. At equal call budgets
token cost still varies 2.1x. Extra calls therefore create candidate opportunity that only some architectures, on some tasks, convert. Team scaling is a task- and architecture-specific bet, not a uniform lever.
Read the full abstract ▾
Computer Vision
One Geometry, Different Outcomes: Readout-Dependent Effects of the Modality Gap in Vision-Language Models
Aditya Sharma and Divya Saxena
Contrastive
vision-
language models learn shared
embedding spaces by aligning matched image-text pairs, yet their representations remain separated by a modality gap. Prior work reports divergent effects of modifying this gap: reducing it can improve zero-shot classification and cross-modal
alignment, whereas removing gap-related structure can degrade image-text
retrieval. In this paper, we provide a unified
geometric explanation for these task-dependent effects. Across CLIP and SigLIP encoders, we find that a single dominant direction captures 94.4-99.9% of the squared norm of the image-text mean separation, revealing that the mean-separation component is approximately rank-one. A decomposition of the similarity score then identifies three task-specific roles. In zero-shot classification, query-side fixed gap-offset subtraction is exactly equivalent to an additive class bias. In standard cross-modal
retrieval, projecting out the gap direction and renormalising residuals discards candidate-specific norm information, inducing a multiplicative ranking distortion; a geometry-derived exponent tracks the grid-search optimum (Spearman rho = 0.93) and restores performance in some settings, although the gains transfer unevenly. In mixed-modal
retrieval, the gap direction sorts candidates by modality; its removal can improve cross-modal ranking, unlike random or non-gap controls. Residual semantic structure after removal defines the limits of the rank-one account. Together, these results explain why gap modification can improve, degrade, or restore performance across downstream settings. By clarifying when and why gap modification changes model behavior, this account provides a principled basis for selecting gap interventions in similarity-based
vision-
language systems across evaluated downstream tasks.
Read the full abstract ▾
Machine Learning
Data Unlearning via Inverse Distillation
Aleksei Leonov, Nikita Kornilov, Zhenhe Zhang et al. (+3 more)
Multi-step matching models, including flow and
diffusion models, produce high-quality outputs but incur substantial inference costs and may reproduce unwanted components of their training
datasets. We introduce Inverse
Distillation Unlearning (IDU), a unified framework that simultaneously distills a teacher multi-step matching model into an efficient one-step student generator and suppresses outputs corresponding to a designated training subset. We first formulate
distillation as a min-max objective over a data distribution and then represent this distribution as a mixture of the forget-set and the generated distributions. This allows us to compare this mixture with the teacher's training distribution and recover only the retained data at the optimum. Our method requires only a pretrained full-data teacher and data from the forget set, without access to retained training examples, extra feature extractors or classifiers. Extensive experiments on MNIST and CIFAR-10
datasets under flow-matching and score-based
diffusion settings demonstrate that IDU substantially reduces the generation frequency of forgotten classes while preserving generation quality on the retained classes. To the best of our knowledge, IDU is the first unified framework for simultaneous unlearning and
distillation in unconditional flow-matching and score-based models.
Read the full abstract ▾
All authors Aleksei Leonov, Nikita Kornilov, Zhenhe Zhang, Evgeny Burnaev, Iaroslav Koshelev, Alexander Korotin
Computation and Language · Cross-listed from q-bio.NC
Better Behavioral Prediction, More Faithful Model Ablations? Evidence from Sequential Choice
Hanbo Xie
Using predictive models to explain cognition requires more than accurate behavioral predictions. Input ablations offer an appealing route: remove information from a model and interpret the resulting performance change as evidence of its importance for behavior. Yet this inference assumes that the model's dependence on information reflects the dependence of the process generating the behavior. We test it in two synthetic sequential bandit tasks with known generating policies, where past choices can remain informative when feedback is unavailable to a predictor. We compare GRUs and
Transformers trained from scratch, a fine-tuned LLaMA model, and cognitive models across systematically varied reward contributions. Our analyses distinguish prediction after training without reward observations from the response of a fixed predictor to donor-reward replacement. Three findings emerge. First, in the restless task, neural models trained without rewards predict held-out choices better than four simple training-fitted behavioral baselines. Second, under matched donor replacement, accurate predictors can respond much less than the known generator. Third, at some reward weights, neural networks predict better than a pooled reinforcement-learning model but have less faithful changes in choice probabilities; the model ordering differs between the two tasks. These independent-test results separate information sufficient for prediction from response fidelity under a specified ablation in sequential choice. They motivate validating model-ablation responses independently of predictive performance before using them to infer how the observed behavior was generated.
Read the full abstract ▾
Systems and Control
Emergency Control of Transmission Voltages by Coordinating DERs From Multiple Substations
Francisco Escobar and Gustavo Valverde
Voltage stability can be endangered in transmission networks that operate under high strain and exhibit a large separation between generators and loads. During emergencies associated to voltage instability, one promising supporting countermeasure is to control large populations of Distributed Energy Resources (DERs) connected to medium- and low-voltage networks, so that they locally reduce the net load by modifying their power generation or consumption. However, it is challenging to achieve this control while anticipating interactions with legacy voltage regulators, coordinating efforts across substations, and exchanging small amounts of information with DERs. In this paper, we propose a control scheme that coordinates DER aggregations with the load tap changers of substation
transformers, asks them to contribute power depending on the substation they are connected to, and relies on lightweight communication. Our simulations show that the proposed control can successfully coordinate DERs and load tap changers to enhance long-term voltage stability.
Read the full abstract ▾
Robotics
MagNav: A Dual-Core Magnetic Track Guidance Framework for Lighting-Invariant Navigation in Two-Wheeled Robots
Arpita Kumari, Anupam Chatterjee and Richa Srivastava
Two-Wheeled Inverted Pendulum (TWIP)
robots are useful for studying how to control systems that are naturally unstable and have fewer actuators than degrees of freedom. Adding autonomous line-following to these
robots is challenging because steering and balancing are closely linked. Most existing systems use infrared sensors, which can be affected by changes in lighting, such as sunlight or shadows, making them reliable only indoors. This paper presents a self-balancing
robot that can follow a line using a
magnetic track guidance system. By using a five-channel analog
Hall-effect sensor array, the
robot is not affected by optical interference. The control system uses a cascaded PID structure: the inner loop keeps the
robot balanced using data from an inertial measurement unit with a complementary filter, while the outer loop adjusts steering based on the
magnetic sensor readings. Stepper motors provide precise torque control without needing extra rotary encoders. For comparison, an optical sensor module was also included. Tests show that the
magnetic guidance system keeps accurate tracking even in very bright lighting, over 10,000 Lux, while the optical system loses accuracy and sometimes fails. This design provides a reliable, lighting-independent solution for autonomous
navigation in places like factories, warehouses, and outdoor paths.
Read the full abstract ▾
Systems and Control
Control and Estimation Co-Design via Envelope-Theorem Gradients
Mohammad S. Ramadan, Philip Dinenis and Mihai Anitescu
Instead of the sequential plant–control–
estimator design pipeline, which is in general not optimal, we propose the Control and Estimation Co-Design (ContEst) framework, which poses the entire system design as a single two-stage problem solved by first-order methods. A key enabler is that the gradient of each inner control/estimation optimal cost with respect to the design parameters is available from the dual variables the inner solver already returns, by the envelope
theorem: exactly under mild regularity (a unique inner minimizer), and as a subgradient under more relaxed conditions. That is, no differentiation through the optimizer or its optimality conditions is needed. We first present the framework in a linear quadratic Gaussian (LQG) regime, where the control and estimation costs become two semidefinite programs (SDPs) whose design gradients are read from the duals of the Lyapunov constraints. We extend the results to robust ($H_\infty$) formulation to cover worst-case control and filtering problems. We also present an extension to constrained and nonlinear systems, using an extended Kalman filter (eKF) and a local linearization of the information-state dynamics to yield an approximate but convex model predictive control (MPC) problem whose design gradients are read directly from the costates. We apply our methods to a variety of realistic design examples from different fields and report significant design improvements.
Read the full abstract ▾
Machine Learning
PHASE: A Physiology-Guided Hierarchical Foundation Model for Intracranial EEG
Yipeng Zhang, Chenda Duan, Yuanyi Ding et al. (+10 more)
Clinicians and neuroscientists have long analyzed intracranial electroencephalography (iEEG) through directly measurable physiological characteristics, which carry much of the information that downstream tasks depend on. Recent iEEG foundation models learn by reconstructing or predicting their inputs, which leaves the retention of these characteristics implicit. They are also evaluated mainly on cognitive decoding and a narrow clinical task, i.e., seizure detection. On a broad, clinically relevant
benchmark such as Omni-iEEG, they remain below task-specific models when used frozen. We introduce PHASE, a physiology-guided foundation model that makes these characteristics explicit learning targets, pairing them with masked latent prediction in a temporal stage (PHASE-T) within each channel and a spatiotemporal stage (PHASE-ST) across synchronized channels. PHASE is pretrained on heterogeneous recordings from 222 participants at nine clinical sites. On all five Omni-iEEG clinical tasks, frozen PHASE-T outperforms every evaluated foundation model by up to 31%, and fine-tuned PHASE-T surpasses the task-specific models, setting a new state of the art. PHASE-T benefits from physiological supervision, outperforming variants trained with latent prediction alone or auxiliary waveform reconstruction on every task in matched ablations. PHASE-T generalizes to unseen institutions, outperforming the compared models with few or no local labels. PHASE-ST further improves seizure-onset-zone identification over PHASE-T and, when frozen, decodes sound volume and pitch on BrainTreebank better than published models. Beyond task performance, PHASE learns to encapsulate the physiological characteristics clinicians recognize, from seizure onset and its propagation to anatomical region identity, even though its pretraining contains no ictal recordings or anatomical labels.
Read the full abstract ▾
All authors Yipeng Zhang, Chenda Duan, Yuanyi Ding, Tianyi Wang, Atsuro Daida, Masaki Izumi, Yuta Tanoue, Naoto Kuroda, Shaun A. Hussain, Nishant Sinha, Eishi Asano, Hiroki Nariai, Vwani Roychowdhury
Computation and Language
PADMÉ: Preference Alignment Data Synthesis for Meta-Evaluation of LM Agent Evaluators
Cheng Chang, Yining Mao and Peng Qi
Language models are frequently employed to evaluate other
language models. An LM evaluator scoring agentic behaviors across multiple criteria is valuable, provided that its decisions align with human judgment. We call the problem of evaluating this
alignment Meta-Evaluation. Tackling it directly is difficult: collecting human data is expensive, absolute scoring is hard to align, and using an LM meta-evaluator recurses the question of trustworthiness. We adopt a reformulation of meta-evaluation as a preference judgment problem: rather than comparing human and LM evaluator scores of a trajectory, we ask whether their implied preferences align. Building on this, we introduce PADMÉ, a data synthesis method that generates reliable criterion-based meta-evaluation data for agentic settings. PADMÉ uses only small
language models, requires no human involvement during evaluations, and operates under a low computational budget. We build a prototype of PADMÉ and synthesize a
dataset of 1,000 samples across four agentic domains and three evaluation criteria. Human validation on a 150-sample subset demonstrates that PADMÉ improves agreement with human judgment from 73% to 85% over a naive baseline. Meta-evaluating 25 common models with our
dataset demonstrates the correlations between evaluation performance and scoring granularity, leniency, and model size, among other factors.
Read the full abstract ▾
Artificial Intelligence · Cross-listed from cond-mat.mtrl-sci
Measuring trainable degrees of freedom in materials graph neural networks: a random-subspace intrinsic dimension analysis
Shehroz Ahmad Shoaib and Kangming Li
Final predictive accuracy is the standard basis for comparing graph neural networks (GNNs) in materials-property prediction, but it does not show how strongly performance depends on access to trainable parameter-space directions. Here, we introduce trainable-degree dependence as a complementary characterization of materials GNN learning. Using random-subspace intrinsic-dimension analysis, we train CGCNN, ALIGNN, and DimeNet++ in randomly oriented parameter subspaces across six prediction tasks and measure how performance recovers as independent trainable degrees of freedom are restored. The resulting recovery curves separate endpoint accuracy from the trainable-dimensional demand required to recover it. They reveal distinctions that final errors alone miss: metallic classification and log-bulk-modulus regression recover near-reference performance from small fractional subspaces, formation-energy and band-gap prediction show stronger architecture dependence, and
phonon prediction is most sensitive to dimensional restriction.
Dataset-size sweeps show that band-gap models require larger fractional subspaces as training data grows, whereas formation-energy and bulk-modulus responses are more stable. A width sweep shows that fractional thresholds can remain stable while absolute threshold dimensions increase with model size. Random-subspace analysis therefore provides a targeted stress test for how materials GNNs use their optimization space.
Read the full abstract ▾
Systems and Control · Cross-listed from physics.ins-det
Design and Analysis of a 2D Vernier Structure
Hüsna Yildiz and Rick Helmer
This paper presents the design and simulation of a 2D Vernier Time-to-Digital Converter (TDC) targeting less than 20 picosecond time resolution and 5-bit output resolution. The TDC comprises voltage-controlled delay buffers, NAND-based SR latches with reset functionality, and a calibration loop using a phase-frequency detector (PFD) and charge pump. The calibration loop dynamically adjusts delay mismatches to maintain consistent timing across process-voltage-temperature (PVT) corners, significantly improving linearity. Simulations show that the TDC achieves a resolution of 16.9 ps for tt corner and maintains DNL and INL below 0.5 LSB under all tested PVT conditions
Read the full abstract ▾
Artificial Intelligence
GeoOutageBench: Benchmarking Ambiguity-aware, Ontology-grounded Geospatiotemporal KGQA for Multimodal Power Outage and Resilience Analysis
Ethan D. Frakes, Amy Kvien, Rishabh Kundu et al. (+8 more)
We introduce GeoOutageBench, a
benchmark for assessing LLM-based geospatiotemporal KGQA for
multimodal outage and resilience analysis. Unlike existing KGQA
benchmarks for Web knowledge, GeoOutageBench considers a spatiotemporal KG that integrates visual, textual, and structured data from outage records, remote sensing, weather observations, storm and power events, geographic entities, and domain ontologies. It provides a competency query taxonomy at different difficulty levels from spatiotemporal containment and proximity, spatiotemporal co-occurrence analysis,
multimodal evidence, to hypothetical evaluation. Over
multimodal KG and query classes, GeoOutageBench provides user-configurable evaluation of three important, highly coherent yet less studied tasks: (1) LLMs' understanding for ambiguous geospatiotemporal questions in terms of NL to SPARQL interpretation, (2) query-driven assessment of ontology utility, and (3) answer accuracy of
multimodal KGQA
retrieval. GeoOutageBench provides a design principle and foundation for assessing LLM-KG systems that support real-world infrastructure resilience analysis. Our
benchmark, source code, data, results, and other documentation are available at https://github.com/UCF-SAGE/GeoOutageBench.
Read the full abstract ▾
All authors Ethan D. Frakes, Amy Kvien, Rishabh Kundu, Redad Mehdi, Van D. Tran, Vibha S. Mandayam, Kristopher O. Davis, Erika I. Barcelos, Roger H. French, Yinghui Wu, Mengjie Li
Machine Learning
Early Learning Shapes Later Directions Of Representation Change In Continual Learning
Yuantao Deng, Jinnuo Liu, Kaizhen Tan et al. (+1 more)
Representations continually change as a network learns new tasks. We ask whether early representational changes naturally form a
geometric structure that continues to shape later learning. We identify a low-dimensional subspace of early representation drift, which we call a scaffold, and test whether it is reused across subsequent tasks. Across four pretrained visual encoders and two
datasets, later representational changes consistently favor this early-defined subspace over matched random alternatives. This reuse is history-dependent: when networks experience different early tasks but identical later training inputs, each network preferentially reuses the scaffold induced by its own learning history. The same preference appears in individual optimizer updates, even though the network's dominant local response directions shift away from the original scaffold. Finally, constraining motion within the scaffold slows new-task acquisition more than matched random constraints, while effects on old-task retention are less consistent. In summary, these results suggest that early experience leaves a persistent
geometric imprint on how neural networks adapt to future tasks.Code is available at https://github.com/YuantaoDeng/latent-scaffold.
Read the full abstract ▾
All authors Yuantao Deng, Jinnuo Liu, Kaizhen Tan, Yuchen Liu
Artificial Intelligence
A Polyphonic Conception of AI Understanding
Matthieu Queloz and Pierre Beckmann
When a doctor, a judge, or an engineer must decide whether to trust an AI model's output, they cannot avoid asking what the model understands. Purely mathematical or statistical descriptions struggle to distinguish trustworthy from untrustworthy outputs without reintroducing the question of AI understanding in all but name. Yet the question is ill-framed as it stands, because the inherited concept operates within a monophonic paradigm: the idea that a cognitive system's understanding of something must be localised to a single mechanism underpinning all the capacities conferred by such understanding. Drawing on a wide range of mechanistic evidence, we show that LLMs are pervasively polyphonic: outputs emerge from coalitions of parallel mechanisms of uneven reliability, which variously complement, duplicate, or drown out one another, with several coalitions sufficing for a task without any one being indispensable. Polyphony not only complicates attributions of understanding, but renders monophonic inference patterns hazardous. In response, we develop a conception of understanding fit for polyphonic AI. It centres on sound circuitry that is reliably and correctly recruited and in control of outputs. Attributions of understanding thereby become tractable claims about internal organisation, and can do the work of guiding trust in AI.
Read the full abstract ▾
Machine Learning
GEM-KMeans: Memory-Efficient and Accurate Clustering on Massive Scale with GPU Optimization
Peng Xu, Nihar Koganti, Volodymyr Kindratenko et al. (+1 more)
Memory-efficient scaling on clustering problems without sacrificing statistical accuracy is of central interest for large-scale data analysis and machine learning problems. Nonnegative low-rank (NLR)
matrix factorization for $K$-means is a scalable clustering method, which connects to semidefinite relaxations with optimal average-case exact recovery guarantees. However, a direct GPU implementation of NLR requires multiple large factor-sized buffers and substantial data movements that are essentially
memory-bound. In this paper, we introduce GEM-KMeans, a spectrally normalized yet mathematically equivalent NLR formulation that fuses the gradient update, nonnegative projection, and sufficient statistics for normalization and iterate movement into a
matrix-multiplication epilogue. Instead of retaining three massive factor-sized arrays, our IO-aware GPU implementation materializes only one single factor with small tile-reduction arrays as additional storage in the High Bandwidth
Memory (HBM). We derive explicit
memory costs and spectrally normalized smoothness bounds for optimizing the clustering objective function. Accurate clustering is demonstrated at massive scales on synthetic and real
datasets, where performance gains of GEM-KMeans over existing GPU-accelerated Lloyd's
algorithms involve data-dependent runtime tradeoffs.
Read the full abstract ▾
All authors Peng Xu, Nihar Koganti, Volodymyr Kindratenko, Xiaohui Chen
Computers and Society
Argus: Academic Integrity in the Era of Generative AI
David Racovan, Ajay Rawat, Christopher K. May et al. (+1 more)
The rapid proliferation of large
language models (LLMs) in the context of education has introduced significant challenges in enforcement of academic integrity, especially in programming courses. We present Argus, an automated detection system for LLM-assisted student work in undergraduate C programming assignments. Argus integrates behavioral and stylistic indicators to create a holistic picture of the student's progress through an assignment and surfaces anomalies that point to potential misuse of LLM assistance. We quantify and analyze data over six years of Spring semester offerings in a large-enrollment CS2 course at Purdue University using Argus, finding that 45% of enrolled students exhibited patterns consistent with LLM-assisted code development in Spring 2026. To contextualize these results, we analyze the relationship between flagged LLM use and student performance on written, in-person proctored examinations, and find a significant negative correlation. We also explore the problem of mitigating false positives, recognizing that erroneous accusations of academic integrity carry significant consequences for students and instructors alike, particularly in the context of large enrollment courses. We argue that any automated detection system must be accompanied by a structured process for human review. We discuss the consequences for future course design, changing academic policy as these tools become more ubiquitous, and the pedagogical implications of LLM-based tools in computer science education.
Read the full abstract ▾
All authors David Racovan, Ajay Rawat, Christopher K. May, Jeffrey A. Turkstra
Computational Complexity
Spectral Methods for the Complexity of Planar Graph Homomorphisms
Ashwin Maran, Jin-Yi Cai and Zhuxiao Tang
We explore the frontier beyond the recently discovered barrier represented by the
quantum automorphism group $qut(M)$ in the classification theory of planar graph homomorphisms $PlGH(M)$. We show that analyzing the spectral relations of $M$ can prove \#P-hardness when traditional vertex separation and domain-reduction methods with planar edge gadgets provably fail due to the $\qut(M)$ barrier. We prove two criteria of \#P-hardness for $PlGH(M)$: a spectral criterion and a determinant criterion. It is known that the core problem for the classification of $PlGH(M)$ for nonnegative
matrices $M$ is for positive definite entry-wise positive
matrices. We use the spectral criterion to show that $PlGH(M)$ is \#P-hard for all circulant
matrices of
prime order $q \ge 3$, while for $q=2$ it is precisely the matchgate case and is P-time computable by the FKT
algorithm (for planar perfect matching). We also prove a complexity dichotomy for $\PlGH$ problems defined by tensor products of 2 by 2
matrices. This gives a complete complexity classification for this class of
matrices, and the FKT
algorithm together with a
holographic transformation is universal—every $PlGH(M)$ is either (1) P-time computable over all graphs, or (2) \#P-hard in general but P-time computable over planar graphs, or (3) \#P-hard over planar graphs; furthermore, $PlGH(M)$ in (2) consists of precisely those computable by FKT with a
holographic transformation.
Read the full abstract ▾
Artificial Intelligence
LongCat-DeepResearch Technical Report
Meituan LongCat Team: He Zhu, Yue Xu, Wanli Wu et al. (+27 more)
We present LongCat-DeepResearch, a deep research system that combines an enhanced LongCat model with a multi-
agent workflow for producing comprehensive, evidence-grounded reports. The workflow separates global planning from detailed investigation and coordinates revision at the section level. Multiple planning
agents first explore external sources and refine an actionable research plan, termed ResearchSpec. Research
agents then investigate and draft their assigned sections in parallel, gathering additional evidence in separate contexts as their analyses develop. Once the sections are assembled, global review guides targeted local revisions, reducing reliance on repeated full-report rewriting. This workflow also supports the construction of research tasks and trajectories for the mid-training and post-training of LongCat's general-purpose models. LongCat-DeepResearch achieves 55.25 on DeepResearchBench, 51.35 on DeepResearchBench II, and 79.83 on ResearchRubrics. On an in-house
benchmark, it scores 76.04, ranking second among four compared systems. Development-set analyses show benefits from combining planning perspectives, while further planning refinement has mixed effects. Additional editing improves average automatic readability preference across two
benchmarks, with different trends on each.
Read the full abstract ▾
All authors Meituan LongCat Team: He Zhu, Yue Xu, Wanli Wu, Haolin Ren, Yuxin Bian, Jiarui Zhao, Rongzhi Zhang, Quanchi Weng, Jinghao Cui, Yu Fan, Yuhan Liu, Yunhu Ye, Jiyuan Ren, Fengcheng Yuan, Zhao Yang, Jiacheng Zhang, Yuchuan Dai, Ruixuan Xiao, Haozhe Sun, Xiangyuan Liu, Cheng Sun, Yao Du, Yiming Hao, Hongbo Guo, Shuo He, Lei Wang, Xunliang Cai, Yan Chen, Fan Yang, Lingchuan Liu
Distributed Computing
Mixture-of-Kittens: MoE Megakernel for NVL72s
Stuart H. Sul, Nash Brown, Henry Wildermuth et al. (+3 more)
AI accelerator systems are rapidly consolidating into scale-up architectures, where tens to thousands of GPUs communicate over high-bandwidth, single-hop fabrics. We find that existing Mixture-of-Experts (MoE) training systems, optimized for conventional scale-out networks, transfer poorly to this setting, often running slower than a naive baseline built with PyTorch and NCCL. With industry roadmaps pointing toward even larger scale-up domains, understanding the performance tradeoffs of this hardware regime is increasingly important. We present Mixture-of-Kittens (MoK), an MoE training system designed for Nvidia NVL72. MoK builds on three insights that unlock performance on scale-up domains: (1) choosing push- or pull-based communication per operator, (2) restructuring the computation-communication overlap, and (3) fully eliminating CPU-GPU synchronization. MoK distills these insights into a single deterministic training megakernel that fuses
token dispatch, shared and routed expert FFNs, and
token combine. Across MoE layer shapes from four widely used open-weight models, MoK delivers up to $2.37\times$ the throughput of the strongest publicly available baseline. In a production run on 512 GPUs spanning multiple GB300 NVL72 racks, MoK improves end-to-end training throughput by $1.41\times$.
Read the full abstract ▾
All authors Stuart H. Sul, Nash Brown, Henry Wildermuth, William Lin, Federico Cassano, Christopher Ré
Software Engineering
The Uneven Decline of Collective Knowledge Production: Evidence from Stack Overflow After Generative AI
Myokyung Han, Taegyoon Kim, Jinhyuk Yun et al. (+1 more)
Generative AI (Gen AI) is reshaping how individuals learn and work, but its consequences for collective knowledge, the shared body of knowledge that online communities produce together, remain poorly understood. Prior work has documented an aggregate decline in participation on knowledge-sharing platforms, but it remains unclear which specific kinds of knowledge are being lost first. We study this question using Stack Overflow, one of the largest online communities for software engineering, treating the release of ChatGPT-3.5 as a natural shock. Analyzing over two million questions posted between 2020 and 2025, we track how two dimensions of collective knowledge, difficulty and data availability, change following Gen AI's release. Using diverse methods and robust checks, we find consistent patterns. Easy questions decline sharply while difficult questions become more common, a pattern corroborated by rising code complexity. Data-rich topics and tags lose share of questions, while data-scarce ones gain ground. The two dimensions also interact: the decline in easy questions is concentrated specifically within data-rich domains, while difficult questions increase regardless of data availability. This pattern extends beyond Python across programming
languages, with more prevalent
languages showing sharper shifts. Together, our findings reveal that Gen AI's impact on collective knowledge is uneven, eroding easy, accessible knowledge first while more complex, less common knowledge persists.
Read the full abstract ▾
All authors Myokyung Han, Taegyoon Kim, Jinhyuk Yun, Lanu Kim
Information Theory
A Hyperbolic Bound for File Retrieval in DNA-Based Data Storage
Constantinos Vasilios Argyris Vlachos and Daniella Bar-Lev
In DNA-based storage systems, data are retrieved by sequencing molecules sampled from a DNA pool. We study how the way coding redundancy is shared between files affects their expected
retrieval times. Our focus lies on the case of two files that are encoded by a systematic linear code over an arbitrary finite field. We consider the
conjecture that the sum obtained by dividing each file dimension by its expected
retrieval time is at most one whenever the dimension of at least one file is more than one. For this, we develop a
geometric view of the
retrieval process. As molecules are sampled, we follow the growing span of the corresponding columns of the generator
matrix and track how much of this span comes from each file. Among the samples that enlarge the overall span, this lets us compare those that make progress toward recovering both files with those that make progress toward neither. We call a column mixed if the corresponding encoded symbol combines information from both files. In this work, we sharpen a projection bound and use it to control the effect of mixed columns. We prove the
conjecture whenever the total information dimension is at least twice the number of mixed columns plus two. For equal-sized files, this allows up to one fewer mixed column than the dimension of either file and extends the previous result for codes with no mixed columns.
Read the full abstract ▾
Computer Vision
AerialDojo-200K: A Large-Scale Benchmark Suite for Open-World Aerial Object-Goal Search
Tongtong Feng, Xin Wang, Haoran Hou et al. (+9 more)
Open-world aerial object-goal search is a foundational yet challenging task, requiring aerial
agents to autonomously explore large-scale, unstructured three-dimensional environments and reach target objects specified by semantic descriptions or reference images, rather than following route-specific instructions. However, research in this task remains at a nascent stage and relies on small, environment-specific
benchmarks with heterogeneous action spaces and data formats. These limitations hinder large-scale training and cross-
benchmark evaluation, constraining the scalability and generalizability of aerial
agents. To address this problem, we propose AerialDojo-200K, a large-scale
benchmark suite for open-world aerial object-goal search, with 3 times as many scenes and 18.7 times as many task instances as the largest existing
benchmark for this task. Specifically, we construct 42 simulation scenes spanning four scene families and 21 scene types, including 18 urban, 12 natural, six infrastructure, and six disaster scenes. To ensure data quality, 12 annotators spent two months manually annotating 109 landmarks, 2099 target objects, and 2099 object anchors across these scenes. We further construct 205,732 task instances, comprising over 100K semantic-goal and over 100K image-goal instances across Base, Standard, and Long-Horizon settings. Each task instance includes a collision-free reference trajectory and corresponding multi-view
video recordings. We also develop a unified evaluation framework with a scene partition comprising 21 in-distribution scenes and 21 out-of-distribution scenes. Finally, our evaluation of five open-source and four closed-source
multimodal large
language models reveals that there is still a long way to go toward achieving general-purpose aerial
agents. All can be found at https://fengtt42.github.io/AerialDojo/.
Read the full abstract ▾
All authors Tongtong Feng, Xin Wang, Haoran Hou, Ren Wang, Weiran Wang, Shaokai Zhu, Ziqi Jia, Hao Wang, Yu-Wei Zhan, Zongyuan Wu, Jinghao Cui, Wenwu Zhu
Programming Languages
Irene: Equivalence Checking of Hybrid Quantum Programs via Structure-Preserving Symbolic Reduction
Jingyu Ke, Jingyang Li and Guoqiang Li
Equivalence checking is essential for validating compiler transformations of hybrid
quantum programs, which combine
quantum operations, measurements, and classical control. Measurement-dependent control limits
unitary reasoning, while dependencies between classical outcomes and
quantum operations can enlarge intermediate symbolic states. We present Irene, an equivalence-checking framework for bounded hybrid
quantum programs based on structure-preserving symbolic reduction. The framework progressively simplifies equivalence obligations through three levels of
reasoning. At the gate level, algebraic identities simplify
unitary regions. At the hybrid path-sum (HPS) level, reduced symbolic execution states are represented as typed graphs, whose isomorphism certifies equivalence. Remaining obligations are handled by density kernels that characterize transformations of input density operators into observable outputs, allowing comparison even when internal measurement histories differ. Residual coefficient differences are encoded as SMT queries. A common set of symbolic reductions supports HPS and density-kernel
reasoning by preserving factored Boolean and arithmetic expressions, eliminating reducible dependencies before expanding residual sums. We evaluate Irene against five equivalence checkers on 1,982 program pairs from seven
benchmark suites. Irene solves 1,584 pairs (79.92%), compared with 57.52% for MQT QCEC, the baseline with the highest aggregate coverage, with a mean end-to-end time of 3.93 seconds per solved pair. Applied as an equivalence-checking oracle, Irene also identifies 15 previously unknown bugs in
quantum compilers, including Qiskit, Cirq, and PennyLane.
Read the full abstract ▾
Machine Learning
FLOORA: A Human-Aligned Domain-Specific Language Model for Architectural Design
Sahand Rezaei-Shoshtari, Patryk Wozniczka, Shu Ishida et al. (+14 more)
Foundation models are powerful generators, but many engineering domains require structured representations that general-purpose systems handle poorly. We introduce FLOORA (Floor Layout Optimization with RL
Alignment), a family of small domain-specific
language (DSL) models for architectural layout generation. With specialized data and
alignment, our 0.6B model outperforms much larger frontier models, achieving VLM judge win rates up to 92.0% on out-of-distribution real-world buildings and 96.0% on synthetic buildings. Human evaluations further corroborate these results, with FLOORA selected as the best model in 89.3% of evaluations. FLOORA combines a
token-efficient DSL, custom tokenization, domain-specific pretraining, supervised fine-tuning (SFT), and reinforcement learning (RL) with learned human-preference and verifiable rewards. This pipeline improves architectural and
geometric validity, supported by extensive empirical evaluation and ablation studies. Although focused on architecture, our results suggest that similar domain-specific recipes may be useful in other engineering domains with structured, verifiable outputs.
Datasets, models, and inference code are available at https://github.com/AutodeskAILab/floora.
Read the full abstract ▾
All authors Sahand Rezaei-Shoshtari, Patryk Wozniczka, Shu Ishida, Gregg Streuber, Farnoosh Javadi, Jeffrey Landes, Angela Ju, Muhammad Azam, Bryan Lim, Johan Luttun, Indrajeet Haldar, Jonathan Shaw, Beatriz Guerra, Ivan Sosnovik, James Stoddart, Robert Giaquinto, Adam Gaier