PyTorchKR Papers

최신 논문

총 568,830편
태그·기간 등 상세 필터

카테고리 탐색

arXiv:2610.23968v2cs.CV

IMPULSE: Push First, Then Predict in V-JEPA 2 Feature Space

Latent world models such as V-JEPA 2 forecast in the feature space of a video encoder, but one look at an object does not reveal how hard it is to rotate, how much it bounces or how much it slides. People find out by trying: a few pokes reveal how an object responds. IMPULSE, named after the impulse response it infers, gives a predictor in frozen V-JEPA 2 feature space the same ability. A Bayesian contact memory summarizes three short calibration pushes, a predicted object pose carries it into the predictor, and a lightweight random-Fourier pose head lets the predictor use the pose. On 512 fresh objects, in a pre-specified held-out confirmation with models sealed before the data existed, the memory lowers the 1-second feature-forecast error by 11.6-12.8% relative to the same predictor without experience, for all 8 training seeds, with a 95% seed-and-object interval excluding zero. The gain also holds for 128 objects with out-of-range physics (11.7-14.0%), and the full extension beats a plain predictor by 15.5-16.6%. An information-ladder argument, exact for ideal predictors, shows that physics reaches the features only through motion, so we measure how much the pathway lets pose help. On fresh objects, the Fourier head roughly triples the gain of a perfect memory over plain pose tokens (2.8-3.2 times). Raw push measurements fed through the same pathway give a similar gain, so the Bayesian summary is not required; it condenses the trials into eight interpretable numbers. The gain is absent at contact onset and grows afterwards.

arXiv:2610.23908v2cs.CV

ContraSafe-VL: A Contrastive Safety Preference Dataset for Vision-Language Models

Vision-Language Models (VLMs) are increasingly deployed as general-purpose multimodal assistants, making multimodal safety alignment an important challenge. Safety depends on the interplay between textual and visual content, with visual context often playing a critical role. However, VLMs may fail to use visual information appropriately, either overlooking safety-relevant visual content or over-interpreting sensitive visual content, leading to unsafe responses or over-refusal. Existing multimodal safety alignment datasets largely follow the response-preference paradigm, which teaches models how to respond but provides limited supervision on how visual context should inform the safety judgment, making it difficult to balance safety and over-refusal. To address this limitation, we introduce \textbf{ContraSafe-VL}, a \textbf{Contra}stive \textbf{Safe}ty preference dataset for \textbf{V}ision-\textbf{L}anguage Models. For each fixed textual query, ContraSafe-VL constructs paired visual contexts that result in multimodal queries with different safety labels, providing direct supervision for how visual context should inform multimodal safety judgments. It further provides response preference pairs based on harmlessness and helpfulness to encourage models to generate responses that are both safe and helpful. Experiments across three VLM architectures and multiple benchmarks demonstrate that training with ContraSafe-VL improves cross-modal safety recognition and achieves a better safety--over-refusal trade-off while preserving general multimodal capabilities. The dataset is available at https://github.com/pikepokenew/ContraSafe-VL

arXiv:2610.23550v2cs.RO

Reason-guided Human Reward Modeling for Preference-based Reinforcement Learning

Reinforcement learning (RL) policies optimized for task rewards may produce behaviors that are efficient but misaligned with human expectations, safety preferences, or intent. This paper presents a reason-guided preference-based reinforcement learning (PbRL) framework that learns a human reward model from trajectory-level feedback. In addition to preference labels, the framework incorporates reason tags explaining why a behavior is preferred, enabling the learned reward to capture both human choices and their underlying rationale. Compared with preference-only reward learning, the reason-guided formulation provides additional supervision for learning human-aligned rewards and improves the interpretability of the resulting policy. The learned human reward is combined with the environment reward during policy optimization, allowing the agent to improve task performance while producing more interpretable, human- aligned, and preference-sensitive behavior

arXiv:2610.23298v2physics.flu-dyn

How the elastic axis folds the transonic flutter boundary: shock energy transfer and conformally calibrated active learning

Isogai's case A, a pitch-plunge aerofoil whose elastic axis lies half a chord ahead of the leading edge, has a folded transonic flutter boundary in inviscid flow, with several flutter speeds at one Mach number. We map this boundary and its elastic-axis family with coupled Euler simulations, a Gaussian-process surrogate of the damping trained by batch active learning, and distribution-free split conformal prediction, which bounds the marginal error rate of stability calls. Three failure modes of damping labels are invisible in the residuals, and the half-degree mean incidence left by an impulsive start at $1^\circ$ freestream incidence raises the flutter speed at the dip by 2.8%, more than any numerical setting. With 80 simulations the learned case A boundary labels 63 of 64 independent runs correctly, and at most 1.8% of calibrated calls on held-out runs are wrong. Over Mach number, elastic axis and flutter speed index, 193 simulations show that, for the case A structure about the axis, the inviscid fold exists only for axes more than about 0.4 chord ahead of the leading edge, on two meshes. Chordwise energy budgets, closed three ways, explain both branches. On the lower branch the falling reduced frequency, at unchanged shock phase, raises the shock input and lowers the upstream damping similarly. On the upper branch the shock is pinned at the trailing edge; moving the axis aft brings the flutter-mode pivot towards the section, so the upstream damping falls faster than the residual shock input and the fold closes.

arXiv:2610.22852v2cs.RO

Sample-Efficient Reinforcement Learning for Policy Improvement via Flow-Reversed Action Anchors

Generalist robot policies offer strong behavioral priors, but achieving deployment-level reliability and throughput remains difficult. Reinforcement learning (RL) can improve these policies, yet its effectiveness is constrained by the cost of online interaction. Existing methods adapt frozen flow-matching policies by treating noise space as a continuous action space. We show that this behaviorally unstructured space makes exploration inefficient, often preventing these methods from extracting meaningful learning signals on challenging tasks. We introduce AnchorQ, which enables coarse-to-fine policy improvement through a structured, refinable latent action representation. At each decision step, AnchorQ uses flow reversal to map reference behaviors to state-conditioned latent anchors, samples local particles around those anchors, and applies value-guided sampling to explore promising refinements. Our analysis shows that anchors reduce identification complexity at the cost of support bias; particle refinement trades this bias against action-class complexity; and critic accuracy determines the reliability of KL-regularized, value-guided selection. Across simulated and real-world manipulation tasks, AnchorQ outperforms state-of-the-art post-training baselines in sample efficiency, delivering a 37% improvement in real-world success rate and 97% higher simulated task throughput under the same interaction budget. Additional visualizations are available on our website https://anchorq-rl.com/.

arXiv:2610.22420v2cs.AI

Sensitivity Analysis of Decision Rules in an Agent-Based Agro-Ecological Model

Agent-based models are well-suited for modelling agro-ecological systems for decision-making, yet turning their simulations into actionable insight requires identifying which decisions actually drive the behaviours. In such a context, sensitivity analysis stumbles upon a structural feature of many complex systems: the parameters are not a flat vector but a decision tree in which many parameters appear only when a higher-level choice activates them. Classical global sensitivity analysis presupposes independent inputs over a fixed Cartesian product and is ill-posed on such a branched space and biased by construction. We instead treat the structure of the decisions as a first-class modelling object. Using advanced surrogate modelling techniques, we encode the crop-management itinerary of the high-resolution agro-ecological modelling platform MAELIA as a hierarchical design space featuring both categorical and numerical variables. We propose a novel dedicated method to analyse this kind of hierarchical space, a random-forest-supervised HSIC-ANOVA, which ranks main effects and interactions natively on the tree while placing rarely and always active decisions on an equal footing. We show how, when applied to simulation outputs such as nitrogen leaching, soil organic carbon variation, and yield predictions, it can deliver an interpretable, decision-oriented reading.

arXiv:2610.22100v2cs.CL

When the next byte's type is fixed: why entropy patching skips computed values under tight budgets

The Byte Latent Transformer (BLT) runs a large model once per patch and starts patches where a small model's next-byte entropy is high. Under tight budgets this rule skips computed values, such as the number after "= " in math. On BLT-1B at inference we trace the miss to the cue before a value: when the preceding bytes fix the next byte's type, the patcher's entropy is capped near log2 10 bits, close to a 10% budget's threshold. Entropy then starts a patch at 0% of values after a dollar sign and 19% after "= "; writing "is " instead of "= " raises the 19% to 66%. Version 1's two-byte lookup table, motivated by LZ novelty, acts as a learned detector of cues that tend to precede new values in its fitting text: on the solutions rewritten as prose it covers every value after a dollar sign and none after a plain space, and its margin over entropy falls from +42.1 to -8.9 points, partly because prose gives entropy more starts per value. In BLT-style models trained from scratch (1M-50M parameters, 128-byte context, 3 seeds per rule) and scored by teacher-forced exact match on answer spans, the table beats entropy at 50M by 36.8 points on computed results and 29.6 on final answers, which mostly copy the last computed result. A hand-written marker rule is not distinguishable from the table at 1M and 12M; its 50M control was not run. At 1M and 12M the table ran on a T4 GPU in fp16 and the baselines on an A40 in bf16; at 50M all ran on the A40. End to end, no 1M-50M model solves more than 2 of 100 GSM8K problems. A label-free probe of BLT-1B's patcher finds prose values, mostly after a dollar sign, but costs final answers. In six domains, no rule we tested beat entropy on exact match of whole words and numbers in more than one domain. The positive findings concern tight budgets on cue-formatted text. This version corrects version 1; the work is closed, with no venue intended.

arXiv:2610.21743v2cs.LG

Pairwise AUC Optimization Needs Corrective Power: A Unified View

Direct pairwise AUC-surrogate training is attractive because it trains toward the ranking metric used at test time, especially under class imbalance. This creates a direct alignment between the training objective and the evaluation metric, but such metric alignment alone does not guarantee that stochastic training remains corrective. We study this gap through Corrective Power: whether pairwise mini-batch updates produce stronger and more reliable improvements for more severe positive-negative ranking errors. Our main observation is that direct pairwise training can enter regions where severe errors remain common but no longer produce useful corrective motion. In such regions, even the mini-batch noise that would normally help SGD move along the steepest severe-error correction direction can become much weaker. In this paper, we seek a theoretical explanation for this by exploring the degenerate property of the U-statistics formed from the pairwise gradients therein. Our results provide a mechanistic explanation for several empirical patterns: (a) squared-hinge loss is often an effective AUC surrogate; (b) cross-entropy warm-up can help models that enter the AUC phase from a poor starting region; and (c) parameter-efficient fine-tuning from a pretrained model can reduce that need by starting closer to a favorable ranking geometry. Experiments on 8 CV tasks and 8 NLP tasks, using ResNet50, DenseNet121, DistilBERT-base, and Qwen2.5-1.5B-Instruct, support these predictions.

arXiv:2610.20619v2physics.comp-ph

pcNN-SN: a physics-enhanced predictor-corrector neural network algorithm coupled with the discrete ordinates method for neutron transport equations

Super-resolution machine learning methods hold promise for addressing the high computational costs associated with high-dimensional features in traditional transport simulations. This paper proposes a physics-enhanced neural network coupling algorithm for discrete ordinates (SN) transport calculations. Based on high- and low-accuracy data, we develop a predictor-corrector neural network (pcNN) to predict correction factors for coarse-grid solutions. Based on transport equations, local physical features containing subgrid-scale information are selected as network inputs. Dimensionless transformation of physical quantities reduces the input space dimensionality and narrows the range of parameter distributions, thereby facilitating network training. Physical constraints are incorporated into the network model, and the corrected solution retains its original conservation properties. pcNN and SN transport sweeps are internally coupled, and numerical solutions are corrected online. Numerical results demonstrate that this algorithm can effectively improve coarse-grid solutions to achieve a level of accuracy comparable to that of fine-grid results. The physics-based network design enables the model to demonstrate a certain degree of reusability under different computational conditions. Compared to mesh refinement, the pcNN-SN algorithm reduces the computation time for the three-dimensional test problems by more than 60%.

arXiv:2610.20108v3cs.LG

Exhaustive Canonical Skeletons Generation: A Formal Dataset for Decomposable Symbolic Regression

Neural symbolic regression (SR) models require diverse datasets of mathematical expressions for robust training. However, existing datasets rely predominantly on stochastic sampling methods, which suffer from high structural redundancy and incomplete functional coverage. To address this, we introduce an exhaustive dataset of univariate canonical skeletons generated via a deterministic constrained context-free grammar. A canonical skeleton acts as a generic structural representation that subsumes more specific functional forms, eliminating redundant expressions and avoiding repeated computational efforts. Although our dataset consists of univariate skeletons, it is designed to support the decomposable SR paradigm to solve multivariate SR problems without assuming linear separability. Furthermore, we pair this dataset with a sampling strategy that generates clean numerical data without triggering numerical instabilities or relying on artificial protected operators. Canonicalizing two existing neural SR datasets reveals that our dataset is substantially more compact, eliminates structural repetitions, and achieves a uniform operator distribution free of simplicity bias. Finally, we establish a benchmark of 150 SR problems across three tiers, on which state-of-the-art SR methods are shown to struggle significantly.

arXiv:2610.19732v2cs.NE

Controlling the latency in temporal coded spiking models for neuromorphic datasets

Spiking neural networks (SNNs) promise energy-efficient, event-driven computation on neuromorphic hardware, where response time, or latency, matters as much as energy for real-time applications. Time-to-first-spike (TTFS) coding with learnable synaptic delays is a natural way to exploit the timing of events. However, in state-of-the-art spiking neuron models, spike times emerge from nonlinear membrane dynamics and thresholds: no closed-form expression links network parameters to spike times, so latency can only be measured empirically, not specified or guaranteed at design time. Moreover, TTFS models are mostly evaluated on static datasets, which hides their latency on truly temporal data. Our hypothesis is that, if spike times are computed directly as closed-form functions of presynaptic spike times and synaptic delays, bypassing membrane potential and thresholding, latency becomes an explicit function of the learnable parameters that can be bounded, and therefore controlled, analytically. We build on this hypothesis with the Delay Neural Network (DeNN) in real time, which turns event streams into successive single-spike images, computes each spike time from standardised inter-spike intervals and synaptic delays, and links successive images through a long-term memory. We prove lower and upper bounds on the latency of every neuron of a layer, and show that bounding the synaptic parameters after training with a single hyperparameter places the latency anywhere between these bounds, without retraining. On a ten-class subset of the Google Speech Commands dataset, this post-training control exposes the accuracy--latency trade-off: a three-hidden-layer network reaches 94.25\% accuracy while deciding, on average, 477~ms after the end of the one-second input, and about 250~ms with confidence-based early decisions. Learning is also shown to concentrate the long-term memory on a few neurons. Latency thus becomes a design parameter of spiking models rather than a quantity measured after the fact, a step towards neuromorphic systems with real-time guarantees.

arXiv:2610.19730v2cs.NE

Design and Analysis of Temporal Representations in Neural Networks

Neural networks, spiking or not, compute with amplitudes: activation values, spike counts, firing rates, never event times. No measure exists of how much a network uses time. We propose one, the degree of temporal representation: the fraction of a network's four functions (encoding, activation, decoding, learning) whose variables or parameters are temporal, and hypothesise that it predicts performance on temporal tasks. At fixed architecture, accuracy and energy efficiency increase with the degree: a delay network of maximal degree reaches 98.94\% on Google Speech Commands with 175k parameters, above transformers. To test whether networks of different degree really differ, we convert the delay network into artificial and spiking networks with the same inputs and outputs. From this common start, the artificial one learns differently and temporal gradients depend on event times. Hence we obtain a design rule: for temporal data, make event times variables of a network's functions. Spiking networks, converted after training, keep the delay network's accuracy, unreachable by spiking networks trained from scratch.

arXiv:2610.19011v2cs.LG

PMO-Dock: Benchmarking Docking, Specificity, and Generalization in Molecular Optimization

The Practical Molecular Optimization (PMO) benchmark standardized evaluation in molecular optimization, but it is built on simple property-based oracles that do not capture structure-based drug design. As the field has moved to docking-based objectives, evaluation practice has become unstandardized, and different methods report results on different docking tasks, often without a shared protocol for fair comparison or sample-efficiency constraints. To address this, we introduce Practical Molecular Optimization for Docking (PMO-Dock), a benchmark and protocol for docking-based optimization consisting of 25 tasks covering hit generation, lead optimization, and a new specificity task requiring strong on-target binding while penalizing off-target interactions. The protocol separates model development from final evaluation: validation tasks are used for hyperparameter tuning, and test tasks are held out for final evaluation and are not used during development. Evaluation on the test tasks is done under a strict oracle budget to reflect realistic drug-discovery settings. We benchmark four diverse high-performing methods spanning different optimization paradigms, Saturn (reinforcement learning), GenMol (discrete diffusion), Genetic-guided GFlowNet, and ChemLlama (LLM-based evolutionary search). Our results show no universally best method across tasks, additional analyses of hyperparameter transfer (cross-task and cross-category) and sensitivity show that transfer is often limited and varies by method, motivating method-dependent selection strategies. This benchmark provides a concrete standard for evaluating generalization to unseen tasks and sample efficiency under a limited test-time oracle budget.

arXiv:2610.18654v3cs.LG

Kwai Summary Attention V2 Technical Report

Summary-token architectures such as Kwai Summary Attention (KSA) compress distant context into learnable Summary token per chunk, and hybrid variants interleave these layers with full attention to retain exact access to the original text. In such hybrid KSA-Full architecture, the interleaving inserted summary tokens are served as a streaming flow role across different layers. Although effective, however, it also introduce some bottlenecks: (1) the different layers summary compression chunk must be same, therefore hard to reach a flexible KV Cache compression, e.g., different layer set different summary compression chunk window. (2) in the full-attention layers, the summary token introduced additional computation cost, but such layer ignore them and still pay quadratic prefilling, therefore unfriendly to the computation side. Motivated by them, we developed the KSA V2, which not only achieves a more flexible KV cache memory compression but also reduces the computation cost in Full attention layer. Specifically, we found that the summary tokens can serve two roles: the KV memory and sparse indexer.

arXiv:2610.18272v2cs.CL

JevGPT: Thinking Fast and Slow in Autonomous Research

Autonomous research agents must acquire and reason over evidence within a limited budget. Yet relevant sources can remain uninspected or unused, leaving a gap between retrieval coverage and the evidence that informs research. The challenge is to align fast local judgments with changing evidence needs and feed their findings back into research decisions. We introduce JevGPT, a dual-system architecture for autonomous research that couples fast evidence processing with deliberative research planning. System Two sets evidence objectives and integrates findings, while System One locates support, assesses relevance, and guides further inspection. Independent source judgments run in parallel; selected original passages inform the next research step. Experiments on fixed-corpus question answering and document ranking show that JevGPT achieves leading task performance, with faster completion and lower cost than comparably effective research agents. Evidence analysis shows broader source inspection and greater coverage of supporting documents in the research loop, highlighting efficient evidence processing as a route to more capable autonomous research.

arXiv:2610.17623v3cs.LG

RELAY: Selective Self-Distillation with Adaptive Hand-off for Rubric-Based Reinforcement Learning

Rubric-based reinforcement learning grades a response against many written criteria but collapses them into one reward, which discards the detail the rubric carries and gives no GRPO reward gradient when a group's responses tie. A frozen teacher that reads the rubric as privileged context provides token-level supervision, but we observe an early plateau in the gains over GRPO. RELAY uses advantages and rewards to select positive-advantage and low-scoring zero-advantage rollouts, and monitors the distillation loss to decide when to retire the teacher. Stage 1 combines selective distillation from a frozen teacher with GRPO; a plateau detector then removes the teacher, and Stage 2 continues with GRPO alone. Calibrated on one model-domain pair and reused unchanged in three others, the plateau detector fires between steps 122 and 171. RELAY scores highest in-domain among the compared methods in all four settings, ahead of GRPO on RaR-Medicine by 4.62 points with Qwen3-1.7B and 6.02 with Qwen3-4B. On Qwen3-1.7B/Medicine, removing the teacher from a shared checkpoint outperforms keeping, weakening or refreshing it. Distilling selected rollouts covering about half the response tokens also outperforms distilling all rollouts at a matched average coefficient.

arXiv:2610.17020v2cs.CR

CWE-bench: Refocusing Agentic Security from Exploitation to Defense

Cybersecurity evaluations of AI agents have focused largely on offense, yet a working exploit does not show that an agent can make software safe. An attacker needs to exploit only one vulnerability; a defender must find and fix all of them. We introduce CWE-bench, a benchmark of 250 repository audit-and-repair tasks spanning 185 CWE types and eight languages, grouped into ten weakness classes. We release 100 tasks publicly and keep 150 held out. Agents are given a repository to audit and a description of its threat model, but are not told the weakness, where it occurs, or how many vulnerable sites there are. Each patch must block the exploit while preserving original functionality, as checked by a deterministic programmatic verifier. Verifiers were curated with panels of agentic judges and reviewed by hand, so that they reject incomplete repairs without penalizing correct ones. The best agent reaches 67% pass@1 and 79% pass@4. We further conduct analyses showing that (i) agents with similar pass@1 differ widely in retry gains and cost, and (ii) no single agent is best across languages or weakness classes. As the offensive capabilities of agents improve, CWE-bench provides a broad and rigorous measure of whether their defensive capabilities keep pace.

arXiv:2610.16822v3cs.LG

DTA-GT: Direction- and Topology-Aware Graph Transformer for Neural Network Representation Learning

Architecture attribute predictors reduce the cost of neural architecture search by estimating accuracy or latency from a small set of evaluated candidates. Neural architectures are operation-labeled directed acyclic graphs (DAGs), where edge directions determine computational roles and execution dependencies. Existing graph Transformer predictors incorporate topological and directional information in different ways, but primarily focus on specific mechanisms rather than coordinated encoder-wide modeling of these semantics. We introduce encoder-wide directional consistency, a design principle that coordinates the modeling of directed computational semantics across node initialization, global spectral encoding, pairwise interaction, and feature updates. To implement this principle, we propose DTA-GT, a Direction- and Topology-Aware Graph Transformer that integrates direction- and topology-aware node initialization, magnetic Laplacian spectral encoding, direction-dual structural attention, and a Direction-Sensitive MoE-FFN. Across accuracy and latency prediction on NAS-Bench-101 and NAS-Bench-201, DTA-GT consistently outperforms representative RNN-, CNN-, GNN-, Transformer-, and hybrid predictors under limited-budget settings. Ablations and magnetic spectral controls support the contribution of the direction-aware stages and orientation-sensitive spectral information, while mechanistic diagnostics characterize module behavior. These results support encoder-wide directional consistency as an effective design principle for architecture DAG representation.

arXiv:2610.16776v2cs.AI

Game-Creation-Agents: Autonomous Game Production

Game-Creation-Agents (GCA) turns a natural-language request into a runnable Phaser 4 browser game. The pipeline first builds a multimodal game design document that couples mechanics and scene contracts to grounded visual references. It then projects that document into a runtime scaffold, machine-readable contracts, and deterministic tests. Eight specialized agents next carry out production: scenes are implemented serially, assets are produced in parallel, and one reduced manifest joins the two branches into a single build. Build-linked real-UI checks and a code-blind critic then verify that build and decide whether the game may ship. Each blocker returns to the agent that owns the affected artifact for a bounded repair, and the game is released once all the gates pass. We evaluate finished games separately, through their rendered interfaces alone, over prompt-matched tasks run under a common protocol. GCA achieves stronger requirement satisfaction, core mechanics, content depth, and functional visuals than a comparable open pipeline. In a side-by-side human study, raters also score GCA's games higher on every dimension and prefer them overall.

arXiv:2610.14423v3cs.CL

Scaling Laws for Multimodal Data Mixtures

Frontier AI systems are increasingly natively multimodal, jointly pretrained on multiple modalities, such as text, vision, and audio. However, the scientific understanding of the multimodal data mixtures used to pretrain these systems --- how much of each modality to mix, and at what potential interference cost to the others --- is largely absent or gatekept. In this work, we derive data mixing scaling laws for three modalities: text, vision, and audio. Multimodal Large Language Models (MLLMs) increasingly adopt Mixture-of-Experts (MoE) architectures, as MoEs scale efficiently and partially limit cross-modal interference. We therefore use MoEs to conduct 268 experiments ranging from 457M to 8.3B parameters, on up to 150 billion tokens. Our work leads to multiple actionable insights into multimodal data mixing for training MLLMs. First, compute optima are modality-specific: audio modality is model-heavy, favoring scaling parameters over tokens, while text and vision require balanced scaling of model size and data. Second, repeating vision or audio data beyond 2x yields negligible benefits. Finally, cross-modal interactions are highly asymmetric: vision strongly interferes with audio, whereas audio's interference with learning text and vision is negligible. Overall, we provide a principled foundation for understanding the multimodal data mixtures needed to train frontier MLLMs.