arXiv:2512.17866v4 Announce Type: replace Abstract: This paper demonstrates a novel method to extract photomultiplier tube (PMT) calibration timing constants in large liquid scintillation detectors from physics data using the machinery of unsupervised deep learning. The approach uses a simplified physical model of optical photon transport in the loss function, with PMT calibration constants treated as free parameters, and the simple assumption that individual events represent point-like emission. The problem is, thus, effectively reduced to that of regression on a very large scale, made tractable by deep learning architectures and automatic differentiation frameworks. Using data from the 9,300 PMTs in the SNO+ detector, the method has been shown to reliably extract 3 calibration constants for each of the over 7,500 online PMTs using radioactive background events. We believe that this basic approach can be straightforwardly generalised for a wide range of applications.
Science Journals
arXiv:2512.18295v2 Announce Type: replace Abstract: Continual graph learning (CGL) aims to enable graph neural networks to incrementally learn from a stream of graph structured data without forgetting previously acquired knowledge. Existing methods particularly those based on experience replay typically store and revisit past graph data to mitigate catastrophic forgetting. However, these approaches pose significant limitations, including privacy concerns, inefficiency. In this work, we propose AL GNN, a novel framework for continual graph learning that eliminates the need for backpropagation and replay buffers. Instead, AL GNN leverages principles from analytic learning theory to formulate learning as a recursive least squares optimization process. It maintains and updates model knowledge analytically through closed form classifier updates and a regularized feature autocorrelation matrix. This design enables efficient one pass training for each task, and inherently preserves data privacy by avoiding historical sample storage. Extensive experiments on multiple dynamic graph classification benchmarks demonstrate that AL GNN achieves competitive or superior performance compared to existing methods. For instance, it improves average performance by 10% on CoraFull and reduces forgetting by over 30% on Reddit, while also reducing training time by nearly 50% due to its backpropagation free design.
arXiv:2606.07140v2 Announce Type: replace-cross Abstract: We experimentally demonstrate an improved optical biasing scheme for superconducting nanowire single-photon detectors (SNSPDs), which employs a cryogenic InGaAs-InP photodiode (PD) as a local bias source. It is found that, under illumination from a stable external light source, this PD generates a stable photocurrent in a cryogenic environment (~2.3 K), with fluctuations in the photocurrent primarily attributed to fluctuations in the incident optical power. Furthermore, by screening and effectively blocking stray photons leaking from the PD, which give rise to background dark counts, we have achieved an SNSPD exhibiting an ultra-low intrinsic dark count rate of 1e-4 cps. Utilizing this improved optical biasing technique, our SNSPD achieved performance comparable to that obtained under conventional electrical biasing: a system detection efficiency of 80.7%, a background dark count rate of 32.6 cps, and a minimum timing jitter of 57.5 ps. These results indicate that cryogenic-PD-based optical biasing serves as a viable, low-noise, and low-jitter alternative to traditional electrical biasing. Moreover, this work offers useful design guidance for the future development of PD-based low-noise bias sources and for the construction of all-photonic SNSPD systems tailored for high-precision quantum photonics applications.
arXiv:2512.22560v2 Announce Type: replace Abstract: Agentic Reinforcement Learning (RL) trains LLMs through multi-turn interactions with environments, producing workloads that mix compute-bound prefill, bandwidth-bound decoding, CPU-heavy environment execution, and bursty reward evaluation. Existing systems either colocate all stages on a single GPU cluster or decouple them only at a coarse granularity, overlooking hardware heterogeneity and incurring substantial synchronization overhead across stages. We present ROLLART, a system for multi-task agentic RL on disaggregated infrastructure. ROLLART maps each pipeline stage to best-fit hardware, routing prefill-heavy tasks to compute-optimized GPUs, decode-heavy tasks to bandwidth-optimized GPUs, and environments to CPU clusters. It decouples rollout at the trajectory level, allowing generation, environment interaction, and reward scoring to proceed independently, so that slow or failed environments never block the others. ROLLART offloads stateless reward computation to serverless infrastructure and overlaps rollout with training via staleness-bounded asynchronous weight synchronization. Our results demonstrate that ROLLART effectively improves training throughput and achieves 1.31--2.05 \(\times\) training time reduction compared to various RL systems. We also evaluated ROLLART by training a hundreds-of-billions-parameter MoE model for Qoder product on an Alibaba cluster with above 3,000 GPUs, demonstrating its stability and scalability.
arXiv:2512.22827v2 Announce Type: replace Abstract: Code often suffers from performance bugs. These bugs necessitate the research and practice of code optimization. Traditional rule-based methods rely on manually designing and maintaining rules for specific performance bugs (e.g., redundant loops, repeated computations), making them labor-intensive and limited in applicability. In recent years, machine learning and deep learning-based methods have emerged as promising alternatives by learning optimization heuristics from annotated code corpora and performance measurements. However, these approaches usually depend on specific program representations and meticulously crafted training datasets, making them costly to develop and difficult to scale. With the booming of Large Language Models (LLMs), their remarkable capabilities in code generation have opened new avenues for automated code optimization. In this work, we proposed FasterPy, a low-cost and efficient framework that adapts LLMs to optimize the execution efficiency of Python code. FasterPy combines Retrieval-Augmented Generation (RAG), supported by a knowledge base constructed from existing performance-improving code pairs and corresponding performance measurements, with Low-Rank Adaptation (LoRA) to enhance code optimization performance. Our experimental results on the Performance Improving Code Edits (PIE) benchmark demonstrate that our method outperforms existing models on multiple metrics. The FasterPy tool and the experimental results are available at https://github.com/WuYue22/fasterpy.
arXiv:2606.03489v2 Announce Type: replace Abstract: While Large Language Models (LLMs) excel in code generation, they remain prone to replicating subtle yet critical vulnerabilities endemic to their training data. Current alignment techniques, such as Supervised Fine-Tuning (SFT) and Reinforcement Learning (RL), typically apply coarse-grained optimization at the sequence level. This approach often fails to address the localized nature of security flaws, where a single incorrect token choice can compromise an entire program. To bridge this gap, we introduce Tree-like Self-Play (TSP), a framework that reframes secure code generation as a fine-grained sequential decision process. Unlike standard methods that blindly maximize likelihood, TSP constructs a decision tree where the model explores branching trajectories--generating both secure "golden paths" and vulnerable variants. By treating code generation as a self-play game, the model learns to strictly discriminate against its own localized errors. This provides a dense, on-policy learning signal that forces self-correction precisely at the critical decision nodes where vulnerabilities typically emerge. Our experiments demonstrate that TSP fundamentally enhances model reliability. In Python security benchmarks, TSP boosts CodeLlama-7B's pass rate (SPR@1) to 75.8%, significantly outperforming SFT (57.0%) and unstructured self-play baselines. Crucially, TSP induces robust out-of-distribution generalization: the model not only reduces vulnerabilities in unseen categories (CWEs) by 24.5% but also successfully transfers security principles learned from C/C++ to diverse languages, including Python, Go, and JavaScript. This suggests that TSP does not merely memorize patches, but internalizes abstract, language-agnostic security logic.
arXiv:2606.05692v3 Announce Type: replace Abstract: Deep learning has enabled significant advances in time-series causal inference, yet progress remains constrained by the lack of realistic benchmarks with observable counterfactual outcomes. Existing datasets either rely on real-world observations without ground-truth counterfactuals or on simplified simulations that fail to capture complex causal dynamics. To address this gap, we develop a large-scale benchmark for counterfactual prediction in epidemic time series under dynamic interventions. Unlike existing benchmarks, it supports static and time-varying treatments, as well as both single-policy and multi-policy intervention settings, enabling evaluation of causal inference methods across a broad range of causal inference scenarios. Leveraging a calibrated agent-based model grounded in real-world demographic, mobility, epidemiological, and policy data, we generate realistic counterfactual trajectories across more than 150 U.S. counties. Using this benchmark, we evaluate widely used and state-of-the-art causal inference methods, revealing substantial performance differences and highlighting the challenges of realistic time-series causal reasoning.
arXiv:2606.12791v2 Announce Type: replace Abstract: No model of the Korean transmission system at native resolution is publicly available, which makes reproducible research on one of the world's most distinctive grids difficult-an islanded interconnection with extreme separation between generation and the Seoul Metropolitan Area load center, low renewable penetration, and heavy reliance on extra-high-voltage (EHV) transmission. Working strictly from public data, and for research purposes only, we present the GIST 2217-bus test system, a geographically grounded synthetic model of the Korean grid. Unlike fully synthetic cases, whose lines match no real corridor, and aggregated public Korean models, it derives its 345 and 154 kV layout from the OpenStreetMap/OpenInfraMap power layer by a multi-source shortest-path reassembly of overhead-line geometry, gap-fills unreachable substations with a geographic minimum-spanning-tree backbone, and calibrates the aggregate circuit length to published national statistics (94/100/109% at 765/345/154 kV). The model spans 2217 buses, 512 generation and renewable sources (144 GW), 3708 AC line circuits plus four high-voltage direct-current (HVDC) converter links, 3324 transformers, and reactive resources (shunts and 11 FACTS devices), serialized to a PSS/E-compatible CSV schema. The model is distributed as a frozen operating point-taps, setpoints, and bus voltages settled once offline and baked into the data-so a single deterministic pandapower Newton-Raphson pass (with reactive limit enforcement and HVDC converter settling) reproduces an 85 GW high demand snapshot at a single connected operating point (mean transmission voltage 0.995 pu, 2.6 % losses), structurally consistent with the independent public KPG193 model. The dataset, maps, and tooling are released as a citable platform for power flow, planning, and decarbonization studies.
arXiv:2606.16774v1 Announce Type: new Abstract: Equipping Large Language Model (LLM) agents with effective skills is crucial for solving complex tasks in real-world systems like OpenClaw. In this work, we aim to develop a framework that automatically constructs such reusable skills to enhance LLMs in tool use, multi-step reasoning, and dynamic environment interaction. To this end, we propose Collective Skill Tree Search (CSTS), a novel tree-search-based skill construction framework that constructs structured, diverse and generalizable tree of skills. The core idea of CSTS is to leverage collective intelligence to jointly search, identify and compose effective skills via two iterative phases: Collective Skill Node Generation (CSN-Gen) and Collective Skill Node Assessment (CSN-Assess). CSN-Gen exploits collective knowledge from multiple models to explore diverse candidate skills for each subtask, enabling comprehensive skill exploration. CSN-Assess employs multiple models as judges to evaluate and select skill nodes with two scoring mechanisms: (1) collective quality scoring that aggregates independent evaluations to produce a robust estimate of skill effectiveness, and (2) collective transferability scoring that explicitly verifies whether a skill generalizes well across different models. With CSTS, we construct a set of comprehensive tree of skills along with skill-augmented training data, enabling models to effectively learn and utilize skills. Besides, we introduce Collective Skill Reinforcement Learning, which actively selects multiple relevant skills from the tree to broaden solution-space exploration, avoid being trapped by a single skill and its resulting homogeneous or suboptimal solutions. As a result, our trained model, OpenClaw-Skill, exhibits outstanding agentic capabilities in long-horizon planning, tool use and generalization over challenging benchmarks.
arXiv:2505.19699v2 Announce Type: replace Abstract: Federated Learning (FL) is a decentralized machine learning paradigm that enables clients to collaboratively train models while preserving data privacy. However, the coexistence of model and data heterogeneity gives rise to inconsistent representations and divergent optimization dynamics across clients, ultimately hindering robust global performance. To transcend these challenges, we propose Mosaic, a novel data-free knowledge distillation framework tailored for heterogeneous distributed environments. Mosaic first trains local generative models to approximate each client's personalized distribution, enabling synthetic data generation that safeguards privacy through strict separation from real data. Subsequently, Mosaic forms a Mixture-of-Experts (MoE) from client models based on their specialized knowledge, and distills it into a global model using the generated data. To further enhance the MoE architecture, Mosaic integrates expert predictions via a lightweight meta model trained on a few representative prototypes. Extensive experiments on standard image and multimodal benchmarks demonstrate that Mosaic consistently outperforms state-of-the-art approaches under both model and data heterogeneity. The source code has been published at https://github.com/Wings-Of-Disaster/Mosaic.
arXiv:2602.00222v3 Announce Type: replace Abstract: Vision-Language Navigation (VLN) requires agents to follow natural language instructions in partially observed 3D environments, motivating map representations that aggregate spatial context beyond local perception. However, most existing approaches rely on hand-crafted maps constructed independently of the navigation policy. We argue that maps should instead be learned representations shaped directly by navigation objectives rather than exhaustive reconstructions. Based on this insight, we propose MapDream, a map-in-the-loop framework that formulates map construction as autoregressive bird's-eye-view (BEV) image synthesis. The framework jointly learns map generation and action prediction, distilling environmental context into a compact three-channel BEV map that preserves only navigation-critical affordances. Supervised pre-training bootstraps a reliable mapping-to-control interface, while the autoregressive design enables end-to-end joint optimization through reinforcement fine-tuning. Experiments on R2R-CE and RxR-CE achieve state-of-the-art monocular performance, validating task-driven generative map learning.
arXiv:2602.05779v2 Announce Type: replace Abstract: The Edge-of-Chaos (EoC) theory developed for the random initialization of deep networks allows more efficient training by both preserving information in the initial outputs of the network and minimising exploding or vanishing gradients through characterisation of the intermediate layers as Gaussian processes. This EoC theory provides formulae for the choice of the initialisation distribution variances of the weights and biases. For activations which are approximately linear around the origin, the EoC theory typically encourages the Gaussian process variance to converge towards zero with increasing depth. Here we consider the less studied setting of highly sparsity inducing activations where a large region of values near the origin are set to zero. In this setting we prove a new phenomenon whereby initialisations leading to larger fixed Gaussian processes are beneficial to training stability. This theory informs a new, yet simple, initialisation strategy that allows training DNNs and CNNs with as large as 90\% sparsity in the hidden layers.
arXiv:2602.07300v3 Announce Type: replace Abstract: This paper proposes distributed omniscient observers for both heterogeneous and homogeneous linear multi-agent systems, such that each agent can correctly estimate the states of all agents. The observer design is based on local input-output information available to each agent, and knowledge of the global communication graph among agents is not necessarily required. The proposed observers can contribute to distributed Nash equilibrium seeking in multi-player games and the emergence of self-organized social behaviors in artificial swarms. Simulation results demonstrate that artificial swarms can emulate animal social behaviors, including sheepdog herding and honeybee dance-based navigation.
arXiv:2606.06007v2 Announce Type: replace Abstract: Generating realistic synthetic sequential data is critical in real-world applications across operations research, finance, healthcare, energy systems, and scientific computing, where time-indexed observations are used for prediction, simulation, risk assessment, and data-driven decision-making. While diffusion models have achieved remarkable success in generating static data, their direct extensions to sequential settings often fail to capture temporal dependence and information structure. Designing diffusion models that can simulate sequential data in an adapted manner, and hence without anticipation of future information, therefore remains an open challenge. In this work, we propose a sequential forward-backward diffusion framework for adapted time series generation. Our approach progressively injects and removes noise along the sequence, conditioning on the previously generated history to ensure adaptiveness. A novel score-matching objective is introduced for efficient parallel training. We derive rigorous statistical guarantees under a generic framework, then establish score approximation, score estimation, and distribution estimation results with ReLU networks serving as a concrete instance. Empirically, we validate our method on synthetic data, including ARMA models and Gaussian processes, and demonstrate its effectiveness in constructing mean-variance optimal portfolios.
arXiv:2602.06490v2 Announce Type: replace-cross Abstract: Given a standard graded algebra over a field, we consider the relationship between G-quadraticity and the existence of a Koszul filtration. We show that having a quadratic Gr\"obner basis implies the existence of a Koszul filtration for algebras defined by generic determinantal ideals and for algebras defined by binomial edge ideals. We also resolve a conjecture of Ene, Herzog, and Hibi by constructing an example where this implication fails. These results are underpinned by algorithms we develop for constructing Koszul filtrations.
arXiv:2605.27284v2 Announce Type: replace Abstract: Vision-Language-Action (VLA) models are increasingly expected to not only complete robot tasks, but also follow human instructions about how those tasks should be executed. However, existing robot datasets usually pair trajectories with coarse goal-level language, leaving execution-critical details such as active arm, approach direction, and contact region unspecified. This limits steerable policy learning and robotic video understanding. We introduce FineVLA, an open framework for action-aligned fine-grained VLA supervision. The framework includes: (1) a data construction tool that unifies 972,247 trajectories across 85K tasks from 10 open-source robot datasets and builds FineVLA-Data, a human-verified dataset of 47,159 fine-grained trajectories; (2) a held-out benchmark with 500 videos, 11,631 atomic facts, and 1,030 VQA questions; (3) a robotics-specialized VLM annotator for scalable fine-grained annotation; and (4) a steerable VLA policy trained with controlled mixtures of fine-grained and raw goal-level instructions. Our experiments yield three findings. First, fine-grained supervision does not sacrifice goal-level success: FG-only improves over Raw-only by +1.4 to +8.1 success-rate points across settings. Second, fine-grained and raw instructions are complementary, following a consistent inverted-U trend peaking at FG:Raw = 1:2 to 1:1. The best mixed setting reaches 86.8%/82.5% in RoboTwin simulation and 62.7/100 in real-world dual-arm manipulation (vs. 49.9 Raw-only). Third, fine-grained supervision improves steerable control: the largest real-world gains appear on pose (+23), color (+18), and approach direction (+18)--factors where goal-level instructions provide no guidance. Overall, fine-grained language should augment goal-level instructions: specifying how to execute alongside what to achieve. Project page: https://finevla.xlang.ai/
arXiv:2606.04796v2 Announce Type: replace Abstract: Metasurfaces enable precise manipulation of light-matter interactions, and meta-atom coupling and scaling dominates their resonant properties and functional responses. Conventionally, although theories such as CMT and CDT can explain many metasurface phenomena, a unified mechanism that reveals how the deformation of metasurface affect its performance remains lacking. Here, by combining transformation optics and spatial perturbation, we proposed a universal deformation mechanism of metasurface, which establishes a general physical picture. Based on this mechanism, we interpret the resonance frequency drift caused by coupling of the meta-atoms, clarify the tuning law of resonant frequency via geometric scaling of unit structures, and further demonstrate the anisotropic shift of grating resonant peak. Theoretical predictions show consistency with full-wave simulation results in all three scenarios. Given the broad applicability of the transformation optics and perturbation theory, the universal mechanism with intuitive physical picture should be widely existed in diverse fields including photonics crystals, Bragg fibers, two-dimensional materials and crystalline optical properties.
arXiv:2602.00344v2 Announce Type: replace Abstract: While Retrieval-Augmented Generation (RAG) is one of the dominant paradigms for enhancing Large Vision-Language Models (LVLMs) on knowledge-based VQA tasks, recent work attributes RAG failures to insufficient attention towards the retrieved context, proposing to reduce the attention allocated to image tokens. In this work, we identify a distinct failure mode that previous study overlooked: Attention Distraction (AD). When the retrieved context is sufficient (highly relevant or including the correct answer), the retrieved text suppresses the visual attention globally, and the attention on image tokens shifts away from question-relevant regions. This leads to failures on questions the model could originally answer correctly without the retrieved text. To mitigate this issue, we propose MAD-RAG, a training-free intervention that decouples visual grounding from context integration through a dual-question formulation, combined with attention mixing to preserve image-conditioned evidence. Extensive experiments on OK-VQA, E-VQA, and InfoSeek demonstrate that MAD-RAG consistently outperforms existing baselines across different model families, yielding absolute gains of up to 4.76%, 9.20%, and 6.18% over the vanilla RAG baseline. Notably, MAD-RAG rectifies up to 74.68% of failure cases with negligible computational overhead.
arXiv:2602.03592v2 Announce Type: replace Abstract: Transcendental Liouvillian extensions are differential fields, in which one can model poly-logarithmic, hyperexponential, and trigonometric functions, logarithmic integrals, and their (nested) rational expressions. For such an extension, we construct, over the subfield of constants, a complement of the subspace of derivatives, and develop an algorithm that decomposes any element of the field into the sum of a derivative and a component lying in the complement. Consequently, an element is a derivative if and only if its complementary component vanishes. Moreover, the algorithm enables us to determine elementary integrability over the extension by computing parametric logarithmic parts, and leads to a reduction-based approach to constructing telescopers for elements in the extension, provided that an a priori order bound is given.
arXiv:2601.17322v2 Announce Type: replace Abstract: It is natural that we can extend Structural Operational Semantics (SOS) to SOS for true concurrency. From SOS to SOS for true concurrency, it is in nature to give the related concepts in SOS a truly concurrent semantics foundation, i.e., a transition occurs by executing a Partially Ordered Multi Set (pomset) of actions replacing just one single action. Under the framework of SOS, for the extension to the truly concurrent one, something are changing: Labelled Transition System (LTS) is generalized to Pomset LTS (PLTS), Transition System Specification (TSS) to Pomset TSS (PTSS), interleaving behavioural equivalences to truly concurrent ones, congruence formats of TSSs to those of PTSSs; something are remained, such as the concept of conservative extension, the meanings of TSSs and PTSSs.
arXiv:2604.00768v2 Announce Type: replace Abstract: Hand impairment following neurological disorders substantially limits independence in activities of daily living, motivating the development of effective assistive and rehabilitation strategies. Soft robotic gloves have attracted growing interest in this context, yet persistent challenges in customization, ergonomic fit, and user comfort constrain their clinical utility. Here, we present an ergonomic, customizable fabric-based soft robotic glove whose actuators can be tailored to individual finger-joint geometry. The glove comprises five dual-action actuators supporting finger flexion and extension, together with a dedicated thumb abduction actuator. Leveraging computer numerical control heat sealing technology, we fabricated symmetrical-chamber actuators that adopt a concave outer surface upon inflation, thereby increasing finger contact area and improving comfort. Characterization confirmed joint moment and grasping force sufficient for ADL-relevant tasks. In ten healthy subjects, active assistance significantly reduced forearm muscle activity during manipulation, and a pilot study in three individuals with cervical spinal cord injury showed more natural grasp patterns and reduced reliance on tenodesis grasp.
arXiv:2604.02343v2 Announce Type: replace Abstract: We study the compression of LLM-generated text across lossless and lossy regimes, characterizing a compression-compute frontier where more compression is possible at the cost of more compute. For lossless compression, domain-adapted LoRA adapters can improve LLM-based arithmetic coding by 2x over compression with the base LLM alone. For lossy compression, prompting a model for a succinct rewrite then applying arithmetic coding can achieve compression ratios of approximately 0.03, a 2x improvement over compressing the original response. We further introduce Question-Asking compression (QA), an interactive lossy protocol inspired by the game 'Twenty Questions'. A small model iteratively refines its response by asking yes/no questions to a stronger model, transferring exactly one bit per answer. On 8 benchmarks spanning math, science, and code, 10 binary questions recover 23% to 72% of the capability gap between a small and large model on standard benchmarks and 7% to 38% on harder benchmarks, achieving compression ratios of 0.0006 to 0.004. This is over 100x smaller than prior LLM-based compression (Deletang et al., 2024), suggesting that interactive protocols can transfer knowledge far more efficiently than transmitting full responses.
arXiv:2602.00482v2 Announce Type: replace Abstract: Reinforcement learning (RL)-based post-training for large language models (LLMs) is computationally expensive, as it generates many rollout sequences that frequently share long token prefixes. Existing RL frameworks usually process these sequences independently during policy training, i.e., repeatedly recomputing identical prefixes in both the forward and backward passes of policy gradient computation, leading to substantial inefficiencies in computation resources and memory usage. Although prefix sharing naturally induces a tree structure over rollouts, packed tree-mask approaches scale poorly in RL settings. In this paper, we introduce AReaL-DTA, which efficiently exploits prefix sharing in RL training. AReaL-DTA employs a depth-first search (DFS)-based execution strategy that dynamically traverses the rollout prefix tree during both forward and backward computation, materializing only a single root-to-leaf path at a time. To further improve scalability, AReaL-DTA incorporates a load-balanced distributed batching mechanism that dynamically constructs and processes prefix trees across multiple GPUs. On $\tau^2$-bench, AReaL-DTA improves training throughput by up to $8.31\times$ over dense training and up to $1.70\times$ over sparse training. Our code is available at https://github.com/areal-project/AReaL/tree/feat/dta.
arXiv:2602.08306v2 Announce Type: replace Abstract: Textual Gradient-style optimizers (TextGrad) enable gradient-like feedback propagation through compound AI systems. However, they do not work well for deep chains. The root cause of this limitation stems from the Semantic Entanglement problem in these extended workflows. In standard textual backpropagation, feedback signals mix local critiques with upstream contexts, leading to Attribution Ambiguity. To address this challenge, we propose TextResNet, a framework that reformulates the optimization process to achieve precise signal routing via four key innovations. Firstly, in the forward pass, it enforces Additive Semantic Deltas to preserve an Identity Highway for gradient flow. Secondly, in the backward pass, it introduces Semantic Gradient Decomposition via a Semantic Projector to disentangle feedback into causally independent subspaces. Thirdly, it implements Causal Routing, which routes projected signals to their specific components. Finally, it performs Density-Aware Optimization Scheduling to leverage the disentangled signals to dynamically allocate resources to key system bottlenecks. Our results show that TextResNet not only achieves superior performance compared to TextGrad, but also exhibits remarkable stability for agentic tasks in compound AI systems where baselines collapse. Code is available at https://github.com/JeanDiable/TextResNet.
arXiv:2506.19802v2 Announce Type: replace Abstract: Anomaly-based ML-NIDS (A-NIDS) model normal network behavior from benign data and classify deviations from this baseline as anomalies, theoretically enabling the detection of evolving attack variants without labeled attack data. The ability of A-NIDS to generalize critically depends on the quality of the feature space representing network behavior. However, the requirement for feature spaces that encode attack-relevant semantics has received little attention and remains poorly understood. As a consequence, these systems still struggle to meet practical operational constraints (low false positive rates without compromising detection performance and generalization to attack variants). We identify two limitations in the current feature spaces. First, Out-of-Dimension Blindness, where features do not capture essential attack mechanism properties. Second, Attack Strategy Aggregation Failure, where features cannot encode composite attack behaviors. Moreover, we demonstrate that two SotA data-driven generalization frameworks (based on incremental and contrastive learning) cannot compensate for these feature-level shortcomings. To bridge this gap, we present KnowML, a framework that encodes attack domain knowledge directly into the feature space. For each attack family, our method employs LLMs to construct a corresponding Knowledge Graph (KG) from attack implementations. Symbolic reasoning is then applied over the KG to enumerate potential attack strategies and their compositions. The resulting Knowledge-Augmented Feature Space enables effective generalization even when trained exclusively on benign traffic, a capability beyond current approaches. Systematic empirical evaluations show that KnowML achieves up to 99% detection rates while maintaining false positive rates at or below 0.0137%, substantially outperforming contemporary feature-based baselines across diverse attack variants.