Modern high-frequency trading (HFT) environments are characterized by sudden price spikes that present both risk and opportunity, but conventional financial models often fail to capture the required fine temporal structure. Spiking Neural Networks (SNNs) offer a biologically inspired framework well-suited to these challenges due to their natural ability to process discrete events and preserve millisecond-scale timing. This work investigates the application of SNNs to high-frequency price-spike forecasting, enhancing performance via robust hyperparameter tuning with Bayesian Optimization (BO). This work converts high-frequency stock data into spike trains and evaluates three architectures: an established unsupervised STDP-trained SNN, a novel SNN with explicit inhibitory competition, and a supervised backpropagation network. BO was driven by a novel objective, Penalized Spike Accuracy (PSA), designed to ensure a network's predicted price spike rate aligns with the empirical rate of price events. Simulated trading demonstrated that models optimized with PSA consistently outperformed their Spike Accuracy (SA)-tuned counterparts and baselines. Specifically, the extended SNN model with PSA achieved the highest cumulative return (76.8%) in simple backtesting, significantly surpassing the supervised alternative (42.54% return). These results validate the potential of spiking networks, when robustly tuned with task-specific objectives, for effective price spike forecasting in HFT.

Text-to-image diffusion models have revolutionized visual content generation, but current safety mechanisms apply uniform standards that often fail to account for individual user preferences. These models overlook the diverse safety boundaries shaped by factors like age, mental health, and personal beliefs. To address this, we propose Personalized Safety Alignment (PSA), a framework that allows user-specific control over safety behaviors in generative models. PSA integrates personalized user profiles into the diffusion process, adjusting the model's behavior to match individual safety preferences while preserving image quality. We introduce a new dataset, Sage, which captures user-specific safety preferences and incorporates these profiles through a cross-attention mechanism. Experiments show that PSA outperforms existing methods in harmful content suppression and aligns generated content better with user constraints, achieving higher Win Rate and Pass Rate scores.

The safety of large language models (LLMs) has garnered significant research attention. In this paper, we argue that previous empirical studies demonstrate LLMs exhibit a propensity to trust information from authoritative sources, such as academic papers, implying new possible vulnerabilities. To verify this possibility, a preliminary analysis is designed to illustrate our two findings. Based on this insight, a novel jailbreaking method, Paper Summary Attack (\llmname{PSA}), is proposed. It systematically synthesizes content from either attack-focused or defense-focused LLM safety paper to construct an adversarial prompt template, while strategically infilling harmful query as adversarial payloads within predefined subsections. Extensive experiments show significant vulnerabilities not only in base LLMs, but also in state-of-the-art reasoning model like Deepseek-R1. PSA achieves a 97\% attack success rate (ASR) on well-aligned models like Claude3.5-Sonnet and an even higher 98\% ASR on Deepseek-R1. More intriguingly, our work has further revealed diametrically opposed vulnerability bias across different base models, and even between different versions of the same model, when exposed to either attack-focused or defense-focused papers. This phenomenon potentially indicates future research clues for both adversarial methodologies and safety alignment.Code is available at https://github.com/233liang/Paper-Summary-Attack

Processing long contexts has become a critical capability for modern large language models (LLMs). However, serving long-context LLMs comes with significant inference costs due to the high memory overhead of the key-value (KV) cache. Existing work leverages dynamic sparse attention algorithms (DSAes) to mitigate the KV cache overhead, but these algorithms rely on top-$k$ KV cache selection, which results in a trade-off between accuracy and efficiency. A larger $k$ improves accuracy but decreases efficiency, while a smaller $k$ boosts efficiency but compromises accuracy. To overcome this trade-off, this paper presents PSA, a $\underline{P}$rogressive $\underline{S}$parse $\underline{A}$ttention mechanism that integrates algorithmic innovations with system co-design to achieve both high inference accuracy and improved efficiency in LLM serving. The PSA algorithm adaptively adjusts the KV cache budget of different tokens and layers according to their real attention weight distributions, rather than relying on a fixed budget $k$. This enables high accuracy while minimizing KV cache usage. To further enhance execution efficiency, we introduce a pipelined iteration scheme that reduces CPU-GPU interleaving and synchronization overhead during PSA computation. Additionally, we implement unified GPU memory management that optimizes PSA's memory utilization by accounting for uneven memory requirements across different model layers. Extensive experimental results demonstrate that PSA reduces KV cache usage for attention computation by up to 2.4$\times$ and 8.8$\times$, and increases end-to-end serving throughput by up to 1.4$\times$ and 2.0$\times$, compared to state-of-the-art DSAes and systems without sparse attention, respectively.

Many application domains, e.g., in medicine and manufacturing, can greatly benefit from pneumatic Soft Robots (SRs). However, the accurate control of SRs has remained a significant challenge to date, mainly due to their nonlinear dynamics and viscoelastic material properties. Conventional control design methods often rely on either complex system modeling or time-intensive manual tuning, both of which require significant amounts of human expertise and thus limit their practicality. In recent works, the data-driven method, Automatic Neural ODE Control (ANODEC) has been successfully used to -- fully automatically and utilizing only input-output data -- design controllers for various nonlinear systems in silico, and without requiring prior model knowledge or extensive manual tuning. In this work, we successfully apply ANODEC to automatically learn to perform agile, non-repetitive reference tracking motion tasks in a real-world SR and within a finite time horizon. To the best of the authors' knowledge, ANODEC achieves, for the first time, performant control of a SR with hysteresis effects from only 30 seconds of input-output data and without any prior model knowledge. We show that for multiple, qualitatively different and even out-of-training-distribution reference signals, a single feedback controller designed by ANODEC outperforms a manually tuned PID baseline consistently. Overall, this contribution not only further strengthens the validity of ANODEC, but it marks an important step towards more practical, easy-to-use SRs that can automatically learn to perform agile motions from minimal experimental interaction time.

Optimizing the fuel cycle cost through the optimization of nuclear reactor core loading patterns involves multiple objectives and constraints, leading to a vast number of candidate solutions that cannot be explicitly solved. To advance the state-of-the-art in core reload patterns, we have developed methods based on Deep Reinforcement Learning (DRL) for both single- and multi-objective optimization. Our previous research has laid the groundwork for these approaches and demonstrated their ability to discover high-quality patterns within a reasonable time frame. On the other hand, stochastic optimization (SO) approaches are commonly used in the literature, but there is no rigorous explanation that shows which approach is better in which scenario. In this paper, we demonstrate the advantage of our RL-based approach, specifically using Proximal Policy Optimization (PPO), against the most commonly used SO-based methods: Genetic Algorithm (GA), Parallel Simulated Annealing (PSA) with mixing of states, and Tabu Search (TS), as well as an ensemble-based method, Prioritized Replay Evolutionary and Swarm Algorithm (PESA). We found that the LP scenarios derived in this paper are amenable to a global search to identify promising research directions rapidly, but then need to transition into a local search to exploit these directions efficiently and prevent getting stuck in local optima. PPO adapts its search capability via a policy with learnable weights, allowing it to function as both a global and local search method. Subsequently, we compared all algorithms against PPO in long runs, which exacerbated the differences seen in the shorter cases. Overall, the work demonstrates the statistical superiority of PPO compared to the other considered algorithms.

More than a thousand 8" silicon sensors will be visually inspected to look for anomalies on their surface during the quality control preceding assembly into the High-Granularity Calorimeter for the CMS experiment at CERN. A deep learningbased algorithm that pre-selects potentially anomalous images of the sensor surface in real time has been developed to automate the visual inspection. The anomaly detection is done by an ensemble of independent deep convolutional neural networks: an autoencoder and a classifier. The performance is evaluated on images acquired in production. The pre-selection reduces the number of images requiring human inspection by 85%, with recall of 97%. Data gathered in production can be used for continuous learning to improve the accuracy incrementally. Keywords: Anomaly detection, autoencoder, convolutional deep neural networks, silicon sensors, quality control, visual inspection 1. Introduction Silicon sensors are used in high-energy physics experiments due to their sufficient radiation tolerance, energy resolution and cost-effectiveness. In the high radiation area, the active element of the High-Granularity Calorimeter (HGCAL) [1], which will replace the endcap calorimeters of the CMS [2] experiment at the Large Hadron Collider (LHC) [3], will consist of more than 27,000 hexagonal 8" silicon sensor wafers to achieve unprecedented transverse and longitudinal segmentation.

As social media becomes a hotbed for the spread of misinformation, the crucial task of rumor detection has witnessed promising advances fostered by open-source benchmark datasets. Despite being widely used, we find that these datasets suffer from spurious correlations, which are ignored by existing studies and lead to severe overestimation of existing rumor detection performance. The spurious correlations stem from three causes: (1) event-based data collection and labeling schemes assign the same veracity label to multiple highly similar posts from the same underlying event; (2) merging multiple data sources spuriously relates source identities to veracity labels; and (3) labeling bias. In this paper, we closely investigate three of the most popular rumor detection benchmark datasets (i.e., Twitter15, Twitter16 and PHEME), and propose event-separated rumor detection as a solution to eliminate spurious cues. Under the event-separated setting, we observe that the accuracy of existing state-of-the-art models drops significantly by over 40%, becoming only comparable to a simple neural classifier. To better address this task, we propose Publisher Style Aggregation (PSA), a generalizable approach that aggregates publisher posting records to learn writing style and veracity stance. Extensive experiments demonstrate that our method outperforms existing baselines in terms of effectiveness, efficiency and generalizability.

An early, accurate cancer diagnosis can dramatically improve a patient's outcome, but the tests used to diagnose some cancers are invasive at best and downright awful at worst. Prostate cancer is one of the most common cancers in men, with more than 1.2 million new cases every year. Doctors usually screen for it by looking at the levels of a protein called prostate-specific antigen (PSA) in a patient's blood. However, the PSA test isn't very accurate (70% of the people it flags as having prostate cancer don't), so people with high levels of PSA in their blood have to undergo a biopsy to confirm the diagnosis -- an invasive procedure that can lead to rectal bleeding, difficulty urinating, and other unpleasant side effects. Now, researchers at the Korea Institute of Science and Technology (KIST) have developed an AI that can make a nearly 100% accurate prostate cancer diagnosis from a urine sample -- and it may work for other types of cancers, too.

Complex Event Recognition (CER) systems have become popular in the past two decades due to their ability to "instantly" detect patterns on real-time streams of events. However, there is a lack of methods for forecasting when a pattern might occur before such an occurrence is actually detected by a CER engine. We present a formal framework that attempts to address the issue of Complex Event Forecasting (CEF). Our framework combines two formalisms: a) symbolic automata which are used to encode complex event patterns; and b) prediction suffix trees which can provide a succinct probabilistic description of an automaton's behavior. We compare our proposed approach against state-of-the-art methods and show its advantage in terms of accuracy and efficiency. In particular, prediction suffix trees, being variable-order Markov models, have the ability to capture long-term dependencies in a stream by remembering only those past sequences that are informative enough. Our experimental results demonstrate the benefits, in terms of accuracy, of being able to capture such long-term dependencies. This is achieved by increasing the order of our model beyond what is possible with full-order Markov models that need to perform an exhaustive enumeration of all possible past sequences of a given order. We also discuss extensively how CEF solutions should be best evaluated on the quality of their forecasts.