id
stringlengths
9
16
title
stringlengths
4
278
categories
stringlengths
5
104
abstract
stringlengths
6
4.09k
2502.07056
Autonomous Deep Agent
cs.AI cs.LG
This technical brief introduces Deep Agent, an advanced autonomous AI system designed to manage complex multi-phase tasks through a novel hierarchical task management architecture. The system's foundation is built on our Hierarchical Task DAG (HTDAG) framework, which dynamically decomposes high-level objectives into manageable sub-tasks while rigorously maintaining dependencies and execution coherence. Deep Agent advances beyond traditional agent systems through three key innovations: First, it implements a recursive two-stage planner-executor architecture that enables continuous task refinement and adaptation as circumstances change. Second, it features an Autonomous API & Tool Creation (AATC) system that automatically generates reusable components from UI interactions, substantially reducing operational costs for similar tasks. Third, it incorporates Prompt Tweaking Engine and Autonomous Prompt Feedback Learning components that optimize Large Language Model prompts for specific scenarios, enhancing both inference accuracy and operational stability. These components are integrated to form a service infrastructure that manages user contexts, handles complex task dependencies, and orchestrates end-to-end agentic workflow execution. Through this sophisticated architecture, Deep Agent establishes a novel paradigm in self-governing AI systems, demonstrating robust capability to independently handle intricate, multi-step tasks while maintaining consistent efficiency and reliability through continuous self-optimization.
2502.07057
Tokenization Standards for Linguistic Integrity: Turkish as a Benchmark
cs.CL
Tokenization is a fundamental preprocessing step in NLP, directly impacting large language models' (LLMs) ability to capture syntactic, morphosyntactic, and semantic structures. This paper introduces a novel framework for systematically evaluating tokenization strategies, addressing challenges in morphologically rich and low-resource languages. Using a Turkish dataset of 6,200 multiple-choice questions from the Massive Multitask Language Understanding (MMLU) benchmark, the framework assesses tokenizers across five key metrics: vocabulary size, token count, processing time, language-specific token percentages (\%TR), and token purity. These metrics provide a structured approach to evaluating how well tokenizers preserve linguistic structures. While \%TR measures the proportion of valid words in the target language, \%Pure assesses the alignment of tokens with meaningful linguistic units, such as roots and valid morphemes, minimizing semantic fragmentation. The findings reveal that \%TR, introduced as a critical metric, exhibits a stronger correlation with downstream performance (e.g., MMLU scores) than token purity, emphasizing its role in improving model accuracy. Additionally, larger model parameters do not necessarily yield better tokenization quality or enhanced results, highlighting the importance of tailored tokenization strategies that prioritize linguistic alignment. This framework sets a new standard for developing robust tokenization methods optimized for morphologically complex and low-resource languages. Future work will refine morphological analysis, explore domain-specific customizations, and conduct cross-linguistic evaluations to further enhance tokenization practices.
2502.07058
Using Contextually Aligned Online Reviews to Measure LLMs' Performance Disparities Across Language Varieties
cs.CL cs.HC
A language can have different varieties. These varieties can affect the performance of natural language processing (NLP) models, including large language models (LLMs), which are often trained on data from widely spoken varieties. This paper introduces a novel and cost-effective approach to benchmark model performance across language varieties. We argue that international online review platforms, such as Booking.com, can serve as effective data sources for constructing datasets that capture comments in different language varieties from similar real-world scenarios, like reviews for the same hotel with the same rating using the same language (e.g., Mandarin Chinese) but different language varieties (e.g., Taiwan Mandarin, Mainland Mandarin). To prove this concept, we constructed a contextually aligned dataset comprising reviews in Taiwan Mandarin and Mainland Mandarin and tested six LLMs in a sentiment analysis task. Our results show that LLMs consistently underperform in Taiwan Mandarin.
2502.07059
Federated Continual Learning: Concepts, Challenges, and Solutions
cs.LG cs.AI
Federated Continual Learning (FCL) has emerged as a robust solution for collaborative model training in dynamic environments, where data samples are continuously generated and distributed across multiple devices. This survey provides a comprehensive review of FCL, focusing on key challenges such as heterogeneity, model stability, communication overhead, and privacy preservation. We explore various forms of heterogeneity and their impact on model performance. Solutions to non-IID data, resource-constrained platforms, and personalized learning are reviewed in an effort to show the complexities of handling heterogeneous data distributions. Next, we review techniques for ensuring model stability and avoiding catastrophic forgetting, which are critical in non-stationary environments. Privacy-preserving techniques are another aspect of FCL that have been reviewed in this work. This survey has integrated insights from federated learning and continual learning to present strategies for improving the efficacy and scalability of FCL systems, making it applicable to a wide range of real-world scenarios.
2502.07064
Contextual Thompson Sampling via Generation of Missing Data
cs.LG cs.AI stat.ML
We introduce a framework for Thompson sampling contextual bandit algorithms, in which the algorithm's ability to quantify uncertainty and make decisions depends on the quality of a generative model that is learned offline. Instead of viewing uncertainty in the environment as arising from unobservable latent parameters, our algorithm treats uncertainty as stemming from missing, but potentially observable, future outcomes. If these future outcomes were all observed, one could simply make decisions using an "oracle" policy fit on the complete dataset. Inspired by this conceptualization, at each decision-time, our algorithm uses a generative model to probabilistically impute missing future outcomes, fits a policy using the imputed complete dataset, and uses that policy to select the next action. We formally show that this algorithm is a generative formulation of Thompson Sampling and prove a state-of-the-art regret bound for it. Notably, our regret bound i) depends on the probabilistic generative model only through the quality of its offline prediction loss, and ii) applies to any method of fitting the "oracle" policy, which easily allows one to adapt Thompson sampling to decision-making settings with fairness and/or resource constraints.
2502.07065
Active Inference through Incentive Design in Markov Decision Processes
eess.SY cs.SY
We present a method for active inference with partial observations in stochastic systems through incentive design, also known as the leader-follower game. Consider a leader agent who aims to infer a follower agent's type given a finite set of possible types. Different types of followers differ in either the dynamical model, the reward function, or both. We assume the leader can partially observe a follower's behavior in the stochastic system modeled as a Markov decision process, in which the follower takes an optimal policy to maximize a total reward. To improve inference accuracy and efficiency, the leader can offer side payments (incentives) to the followers such that different types of them, under the incentive design, can exhibit diverging behaviors that facilitate the leader's inference task. We show the problem of active inference through incentive design can be formulated as a special class of leader-follower games, where the leader's objective is to balance the information gain and cost of incentive design. The information gain is measured by the entropy of the estimated follower's type given partial observations. Furthermore, we demonstrate that this problem can be solved by reducing a single-level optimization through softmax temporal consistency between followers' policies and value functions. This reduction allows us to develop an efficient gradient-based algorithm. We utilize observable operators in the hidden Markov model (HMM) to compute the necessary gradients and demonstrate the effectiveness of our approach through experiments in stochastic grid world environments.
2502.07067
Repository-level Code Search with Neural Retrieval Methods
cs.IR
This paper presents a multi-stage reranking system for repository-level code search, which leverages the vastly available commit histories of large open-source repositories to aid in bug fixing. We define the task of repository-level code search as retrieving the set of files from the current state of a code repository that are most relevant to addressing a user's question or bug. The proposed approach combines BM25-based retrieval over commit messages with neural reranking using CodeBERT to identify the most pertinent files. By learning patterns from diverse repositories and their commit histories, the system can surface relevant files for the task at hand. The system leverages both commit messages and source code for relevance matching, and is evaluated in both normal and oracle settings. Experiments on a new dataset created from 7 popular open-source repositories demonstrate substantial improvements of up to 80% in MAP, MRR and P@1 over the BM25 baseline, across a diverse set of queries, demonstrating the effectiveness this approach. We hope this work aids LLM agents as a tool for better code search and understanding. Our code and results obtained are publicly available.
2502.07068
Specializing Large Language Models to Simulate Survey Response Distributions for Global Populations
cs.CL
Large-scale surveys are essential tools for informing social science research and policy, but running surveys is costly and time-intensive. If we could accurately simulate group-level survey results, this would therefore be very valuable to social science research. Prior work has explored the use of large language models (LLMs) for simulating human behaviors, mostly through prompting. In this paper, we are the first to specialize LLMs for the task of simulating survey response distributions. As a testbed, we use country-level results from two global cultural surveys. We devise a fine-tuning method based on first-token probabilities to minimize divergence between predicted and actual response distributions for a given question. Then, we show that this method substantially outperforms other methods and zero-shot classifiers, even on unseen questions, countries, and a completely unseen survey. While even our best models struggle with the task, especially on unseen questions, our results demonstrate the benefits of specialization for simulation, which may accelerate progress towards sufficiently accurate simulation in the future.
2502.07069
Semantics-Aware Updates from Remote IoT Devices to Interconnected LEO Satellites
cs.NI cs.IT math.IT
Providing timely and informative data in Integrated Terrestrial and Non-Terrestrial Networks (T-NTNs) is critical as data volume continues to grow while the resources available on devices remain limited. To address this, we adopt a semantics-aware approach to optimize the Version Age of Information (VAoI) in a status update system in which a remote Energy Harvesting (EH) Internet of Things (IoT) device samples data and transmits it to a network of interconnected Low Earth Orbit (LEO) satellites for dissemination and utilization. The optimal update policy is derived through stochastic modeling and optimization of the VAoI across the network. The results indicate that this policy reduces the frequency of updates by skipping stale or irrelevant data, significantly improving energy efficiency.
2502.07070
Comprehensive Analysis of Thermal Dissipation in Lithium-Ion Battery Packs
eess.SY cs.SY hep-ph
Effective thermal management is critical for lithium-ion battery packs' safe and efficient operations, particularly in applications such as drones, where compact designs and varying airflow conditions present unique challenges. This study investigates the thermal performance of a 16-cell lithium-ion battery pack by optimizing cooling airflow configurations and integrating phase change materials (PCMs) for enhanced heat dissipation. Seven geometric configurations were evaluated under airflow speeds ranging from 0 to 15 m/s, reflecting the operational conditions of civilian drones. A comprehensive 3D simulation approach was used to analyze the effects of inlet and outlet configurations, airflow dynamics, and PCM phase transition behavior. Results indicate that the trapezoidal (wide-base) configuration, paired with a 5-inlet and 1-outlet setup, achieves the most balanced performance, effectively maintaining optimal operating temperatures across low and high-speed airflow conditions. PCM integration further stabilized thermal behavior, with phase change durations extending to 12.5 min under tested conditions. These findings highlight the importance of geometric optimization and material integration in advancing compact and reliable thermal management systems for energy-dense battery packs. This study provides a foundation for designing efficient cooling strategies tailored to lightweight applications such as drones and portable energy storage systems.
2502.07071
TRADES: Generating Realistic Market Simulations with Diffusion Models
q-fin.TR cs.AI cs.LG q-fin.CP
Financial markets are complex systems characterized by high statistical noise, nonlinearity, and constant evolution. Thus, modeling them is extremely hard. We address the task of generating realistic and responsive Limit Order Book (LOB) market simulations, which are fundamental for calibrating and testing trading strategies, performing market impact experiments, and generating synthetic market data. Previous works lack realism, usefulness, and responsiveness of the generated simulations. To bridge this gap, we propose a novel TRAnsformer-based Denoising Diffusion Probabilistic Engine for LOB Simulations (TRADES). TRADES generates realistic order flows conditioned on the state of the market, leveraging a transformer-based architecture that captures the temporal and spatial characteristics of high-frequency market data. There is a notable absence of quantitative metrics for evaluating generative market simulation models in the literature. To tackle this problem, we adapt the predictive score, a metric measured as an MAE, by training a stock price predictive model on synthetic data and testing it on real data. We compare TRADES with previous works on two stocks, reporting an x3.27 and x3.47 improvement over SoTA according to the predictive score, demonstrating that we generate useful synthetic market data for financial downstream tasks. We assess TRADES's market simulation realism and responsiveness, showing that it effectively learns the conditional data distribution and successfully reacts to an experimental agent, giving sprout to possible calibrations and evaluations of trading strategies and market impact experiments. We developed DeepMarket, the first open-source Python framework for market simulation with deep learning. Our repository includes a synthetic LOB dataset composed of TRADES's generates simulations. We release the code at github.com/LeonardoBerti00/DeepMarket.
2502.07072
IRepair: An Intent-Aware Approach to Repair Data-Driven Errors in Large Language Models
cs.CL cs.AI cs.SE
Not a day goes by without hearing about the impressive feats of large language models (LLMs), and equally, not a day passes without hearing about their challenges. LLMs are notoriously vulnerable to biases in their dataset, leading to issues such as toxicity. While domain-adaptive training has been employed to mitigate these issues, these techniques often address all model parameters indiscriminately during the repair process, resulting in poor repair quality and reduced model versatility. In this paper, we introduce a novel dynamic slicing-based intent-aware LLM repair strategy, IRepair. This approach selectively targets the most error-prone sections of the model for repair. Specifically, we propose dynamically slicing the model's most sensitive layers that require immediate attention, concentrating repair efforts on those areas. This method enables more effective repairs with potentially less impact on the model's overall performance by altering a smaller portion of the model. We evaluated our technique on three models from the GPT2 and GPT-Neo families, with parameters ranging from 800M to 1.6B, in a toxicity mitigation setup. Our results show that IRepair repairs errors 43.6% more effectively while causing 46% less disruption to general performance compared to the closest baseline, direct preference optimization. Our empirical analysis also reveals that errors are more concentrated in a smaller section of the model, with the top 20% of layers exhibiting 773% more error density than the remaining 80\%. This highlights the need for selective repair. Additionally, we demonstrate that a dynamic selection approach is essential for addressing errors dispersed throughout the model, ensuring a robust and efficient repair.
2502.07076
On the use of neural networks for the structural characterization of polymeric porous materials
cond-mat.soft cond-mat.mtrl-sci cs.CV eess.IV
The structural characterization is an essential task in the study of porous materials. To achieve reliable results, it requires to evaluate images with hundreds of pores. Current methods require large time amounts and are subjected to human errors and subjectivity. A completely automatic tool would not only speed up the process but also enhance its reliability and reproducibility. Therefore, the main objective of this article is the study of a deep-learning-based technique for the structural characterization of porous materials, through the use of a convolutional neural network. Several fine-tuned Mask R CNN models are evaluated using different training configurations in four separate datasets each composed of numerous SEM images of diverse polymeric porous materials: closed-pore extruded polystyrene (XPS), polyurethane (PU), and poly(methyl methacrylate) (PMMA), and open-pore PU. Results prove the tool capable of providing very accurate results, equivalent to those achieved by time consuming manual methods, in a matter of seconds.
2502.07077
Multi-turn Evaluation of Anthropomorphic Behaviours in Large Language Models
cs.CL cs.CY cs.HC
The tendency of users to anthropomorphise large language models (LLMs) is of growing interest to AI developers, researchers, and policy-makers. Here, we present a novel method for empirically evaluating anthropomorphic LLM behaviours in realistic and varied settings. Going beyond single-turn static benchmarks, we contribute three methodological advances in state-of-the-art (SOTA) LLM evaluation. First, we develop a multi-turn evaluation of 14 anthropomorphic behaviours. Second, we present a scalable, automated approach by employing simulations of user interactions. Third, we conduct an interactive, large-scale human subject study (N=1101) to validate that the model behaviours we measure predict real users' anthropomorphic perceptions. We find that all SOTA LLMs evaluated exhibit similar behaviours, characterised by relationship-building (e.g., empathy and validation) and first-person pronoun use, and that the majority of behaviours only first occur after multiple turns. Our work lays an empirical foundation for investigating how design choices influence anthropomorphic model behaviours and for progressing the ethical debate on the desirability of these behaviours. It also showcases the necessity of multi-turn evaluations for complex social phenomena in human-AI interaction.
2502.07081
Fast Clustering of Categorical Big Data
cs.LG cs.DB
The K-Modes algorithm, developed for clustering categorical data, is of high algorithmic simplicity but suffers from unreliable performances in clustering quality and clustering efficiency, both heavily influenced by the choice of initial cluster centers. In this paper, we investigate Bisecting K-Modes (BK-Modes), a successive bisecting process to find clusters, in examining how good the cluster centers out of the bisecting process will be when used as initial centers for the K-Modes. The BK-Modes works by splitting a dataset into multiple clusters iteratively with one cluster being chosen and bisected into two clusters in each iteration. We use the sum of distances of data to their cluster centers as the selection metric to choose a cluster to be bisected in each iteration. This iterative process stops when K clusters are produced. The centers of these K clusters are then used as the initial cluster centers for the K-Modes. Experimental studies of the BK-Modes were carried out and were compared against the K-Modes with multiple sets of initial cluster centers as well as the best of the existing methods we found so far in our survey. Experimental results indicated good performances of BK-Modes both in the clustering quality and efficiency for large datasets.
2502.07082
"Once Upon a Time..." Literary Narrative Connectedness Progresses with Grade Level: Potential Impact on Reading Fluency and Literacy Skills
cs.CL
Selecting an appropriate book is crucial for fostering reading habits in children. While children exhibit varying levels of complexity when generating oral narratives, the question arises: do children's books also differ in narrative complexity? This study explores the narrative dynamics of literary texts used in schools, focusing on how their complexity evolves across different grade levels. Using Word-Recurrence Graph Analysis, we examined a dataset of 1,627 literary texts spanning 13 years of education. The findings reveal significant exponential growth in connectedness, particularly during the first three years of schooling, mirroring patterns observed in children's oral narratives. These results highlight the potential of literary texts as a tool to support the development of literacy skills.
2502.07085
Real-time Optimization for Wind-to-H2 Driven Critical Infrastructures Based on Active Constraints Identification and Integer Variables Prediction
eess.SY cs.SY
This paper proposes a concept of wind-to-hydrogen-driven critical infrastructure (W2H-CI) as an engineering solution for decarbonizing the power generation sector where it utilizes wind power to produce hydrogen through electrolysis and combines it with the carbon captured from fossil fuel power plants. First, a convex mathematical model of W2H-CI is developed. Then, an optimization model for optimal operation of W2H-CI, which is a large-scale mixed-integer convex program (MICP), is proposed. Moreover, we propose to solve this problem in real-time in order to hedge against the uncertainty of wind power. For this purpose, a novel solution method based on active constraints identification and integer variable prediction is introduced. This method can solve MICP problems very fast since it uses historical optimization data to predict the values of binary variables and a limited number of constraints which most likely contain all active constraints. We validate the effectiveness of the proposed fast solution method using two W2H-CI case studies.
2502.07087
Evaluating the Systematic Reasoning Abilities of Large Language Models through Graph Coloring
cs.LG
Contemporary large language models are powerful problem-solving tools, but they exhibit weaknesses in their reasoning abilities which ongoing research seeks to mitigate. We investigate graph coloring as a means of evaluating an LLM's capacities for systematic step-by-step reasoning and possibility space exploration, as well as effects of semantic problem framing. We test Claude 3.5 Sonnet, Llama 3.1 405B, Gemini 1.5 Pro, GPT-4o, o1-mini, and DeepSeek-R1 on a dataset of $k$-coloring problems with $2 \leq k \leq 4$ and vertex count $4 \leq n \leq 8$, using partial algorithmic solvers to further categorize problems by difficulty. In addition to substantial but varying framing effects, we find that all models except o1-mini and R1 exhibit $>60\%$ error rates on difficult problem types in all frames ($>15\%$ for o1-mini and $>10\%$ for R1), and no model achieves perfect accuracy even in the simple domain of 2-coloring 4-vertex graphs. Our results highlight both the considerable recent progress in LLM systematic reasoning and the limits of its reliability, especially in relation to increasing computational costs. We expect that more complex graph coloring problems, and procedural generation of arbitrary-complexity reasoning problems more broadly, offer further untapped potential for LLM benchmarking.
2502.07088
Kernels of Selfhood: GPT-4o shows humanlike patterns of cognitive consistency moderated by free choice
cs.CY cs.AI cs.CL cs.HC cs.LG
Large Language Models (LLMs) show emergent patterns that mimic human cognition. We explore whether they also mirror other, less deliberative human psychological processes. Drawing upon classical theories of cognitive consistency, two preregistered studies tested whether GPT-4o changed its attitudes toward Vladimir Putin in the direction of a positive or negative essay it wrote about the Russian leader. Indeed, GPT displayed patterns of attitude change mimicking cognitive consistency effects in humans. Even more remarkably, the degree of change increased sharply when the LLM was offered an illusion of choice about which essay (positive or negative) to write. This result suggests that GPT-4o manifests a functional analog of humanlike selfhood, although how faithfully the chatbot's behavior reflects the mechanisms of human attitude change remains to be understood.
2502.07090
Generative Distribution Prediction: A Unified Approach to Multimodal Learning
stat.ML cs.AI cs.LG
Accurate prediction with multimodal data-encompassing tabular, textual, and visual inputs or outputs-is fundamental to advancing analytics in diverse application domains. Traditional approaches often struggle to integrate heterogeneous data types while maintaining high predictive accuracy. We introduce Generative Distribution Prediction (GDP), a novel framework that leverages multimodal synthetic data generation-such as conditional diffusion models-to enhance predictive performance across structured and unstructured modalities. GDP is model-agnostic, compatible with any high-fidelity generative model, and supports transfer learning for domain adaptation. We establish a rigorous theoretical foundation for GDP, providing statistical guarantees on its predictive accuracy when using diffusion models as the generative backbone. By estimating the data-generating distribution and adapting to various loss functions for risk minimization, GDP enables accurate point predictions across multimodal settings. We empirically validate GDP on four supervised learning tasks-tabular data prediction, question answering, image captioning, and adaptive quantile regression-demonstrating its versatility and effectiveness across diverse domains.
2502.07096
Lotus: Creating Short Videos From Long Videos With Abstractive and Extractive Summarization
cs.HC cs.CV
Short-form videos are popular on platforms like TikTok and Instagram as they quickly capture viewers' attention. Many creators repurpose their long-form videos to produce short-form videos, but creators report that planning, extracting, and arranging clips from long-form videos is challenging. Currently, creators make extractive short-form videos composed of existing long-form video clips or abstractive short-form videos by adding newly recorded narration to visuals. While extractive videos maintain the original connection between audio and visuals, abstractive videos offer flexibility in selecting content to be included in a shorter time. We present Lotus, a system that combines both approaches to balance preserving the original content with flexibility over the content. Lotus first creates an abstractive short-form video by generating both a short-form script and its corresponding speech, then matching long-form video clips to the generated narration. Creators can then add extractive clips with an automated method or Lotus's editing interface. Lotus's interface can be used to further refine the short-form video. We compare short-form videos generated by Lotus with those using an extractive baseline method. In our user study, we compare creating short-form videos using Lotus to participants' existing practice.
2502.07101
SMAB: MAB based word Sensitivity Estimation Framework and its Applications in Adversarial Text Generation
cs.CL
To understand the complexity of sequence classification tasks, Hahn et al. (2021) proposed sensitivity as the number of disjoint subsets of the input sequence that can each be individually changed to change the output. Though effective, calculating sensitivity at scale using this framework is costly because of exponential time complexity. Therefore, we introduce a Sensitivity-based Multi-Armed Bandit framework (SMAB), which provides a scalable approach for calculating word-level local (sentence-level) and global (aggregated) sensitivities concerning an underlying text classifier for any dataset. We establish the effectiveness of our approach through various applications. We perform a case study on CHECKLIST generated sentiment analysis dataset where we show that our algorithm indeed captures intuitively high and low-sensitive words. Through experiments on multiple tasks and languages, we show that sensitivity can serve as a proxy for accuracy in the absence of gold data. Lastly, we show that guiding perturbation prompts using sensitivity values in adversarial example generation improves attack success rate by 15.58%, whereas using sensitivity as an additional reward in adversarial paraphrase generation gives a 12.00% improvement over SOTA approaches. Warning: Contains potentially offensive content.
2502.07102
Optimal Steady-State Secondary Control of MT-HVdc Grids with Reduced Communications
eess.SY cs.SY math.OC
In this paper, we propose a centralized secondary control for the real-time steady-state optimization of multi-terminal HVdc grids under voltage and current limits. First, we present the dynamic models of the grid components, including the modular multilevel converter (MMC) stations and their different control layers. We also derive the quasi-static input-output model of the system, which is suitable for the steady-state control design. Second, we formulate a general optimization problem using this quasi-static model and find the Karush-Kuhn-Tucker optimality conditions of its solutions. Third, we propose a secondary control based on primal-dual dynamics to adjust the voltage setpoints of the dispatchable MMCs, with which the system asymptotically converges to a steady state that satisfies these optimality conditions. Fourth, we provide a communication triggering mechanism to reduce the communication traffic between the secondary control unit and the MMC stations. Finally, we verify our proposal for different case studies by adapting it to an offshore multi-terminal HVdc grid composed of heterogeneous MMC stations simulated in the MATLAB/Simulink environment. The problems of proportional current minimization and loss reduction are two special case studies.
2502.07107
A Framework for Supervised and Unsupervised Segmentation and Classification of Materials Microstructure Images
stat.AP cs.CV stat.ML
Microstructure of materials is often characterized through image analysis to understand processing-structure-properties linkages. We propose a largely automated framework that integrates unsupervised and supervised learning methods to classify micrographs according to microstructure phase/class and, for multiphase microstructures, segments them into different homogeneous regions. With the advance of manufacturing and imaging techniques, the ultra-high resolution of imaging that reveals the complexity of microstructures and the rapidly increasing quantity of images (i.e., micrographs) enables and necessitates a more powerful and automated framework to extract materials characteristics and knowledge. The framework we propose can be used to gradually build a database of microstructure classes relevant to a particular process or group of materials, which can help in analyzing and discovering/identifying new materials. The framework has three steps: (1) segmentation of multiphase micrographs through a recently developed score-based method so that different microstructure homogeneous regions can be identified in an unsupervised manner; (2) {identification and classification of} homogeneous regions of micrographs through an uncertainty-aware supervised classification network trained using the segmented micrographs from Step $1$ with their identified labels verified via the built-in uncertainty quantification and minimal human inspection; (3) supervised segmentation (more powerful than the segmentation in Step $1$) of multiphase microstructures through a segmentation network trained with micrographs and the results from Steps $1$-$2$ using a form of data augmentation. This framework can iteratively characterize/segment new homogeneous or multiphase materials while expanding the database to enhance performance. The framework is demonstrated on various sets of materials and texture images.
2502.07109
Game of Coding With an Unknown Adversary
cs.IT cs.LG math.IT
Motivated by emerging decentralized applications, the \emph{game of coding} framework has been recently introduced to address scenarios where the adversary's control over coded symbols surpasses the fundamental limits of traditional coding theory. Still, the reward mechanism available in decentralized systems, motivates the adversary to act rationally. While the decoder, as the data collector (DC), has an acceptance and rejection mechanism, followed by an estimation module, the adversary aims to maximize its utility, as an increasing function of (1) the chance of acceptance (to increase the reward), and (2) estimation error. On the other hand, the decoder also adjusts its acceptance rule to maximize its own utility, as (1) an increasing function of the chance of acceptance (to keep the system functional), (2) decreasing function of the estimation error. Prior works within this framework rely on the assumption that the game is complete, that is, both the DC and the adversary are fully aware of each other's utility functions. However, in practice, the decoder is often unaware of the utility of the adversary. To address this limitation, we develop an algorithm enabling the DC to commit to a strategy that achieves within the vicinity of the equilibrium, without knowledge of the adversary's utility function. Our approach builds on an observation that at the equilibrium, the relationship between the probability of acceptance and the mean squared error (MSE) follows a predetermined curve independent of the specific utility functions of the players. By exploiting this invariant relationship, the DC can iteratively refine its strategy based on observable parameters, converging to a near-optimal solution. We provide theoretical guarantees on sample complexity and accuracy of the proposed scheme.
2502.07111
Likelihood-Free Estimation for Spatiotemporal Hawkes processes with missing data and application to predictive policing
cs.LG stat.AP stat.ME
With the growing use of AI technology, many police departments use forecasting software to predict probable crime hotspots and allocate patrolling resources effectively for crime prevention. The clustered nature of crime data makes self-exciting Hawkes processes a popular modeling choice. However, one significant challenge in fitting such models is the inherent missingness in crime data due to non-reporting, which can bias the estimated parameters of the predictive model, leading to inaccurate downstream hotspot forecasts, often resulting in over or under-policing in various communities, especially the vulnerable ones. Our work introduces a Wasserstein Generative Adversarial Networks (WGAN) driven likelihood-free approach to account for unreported crimes in Spatiotemporal Hawkes models. We demonstrate through empirical analysis how this methodology improves the accuracy of parametric estimation in the presence of data missingness, leading to more reliable and efficient policing strategies.
2502.07114
Online Covariance Matrix Estimation in Sketched Newton Methods
stat.ML cs.LG cs.NA math.NA math.OC stat.CO
Given the ubiquity of streaming data, online algorithms have been widely used for parameter estimation, with second-order methods particularly standing out for their efficiency and robustness. In this paper, we study an online sketched Newton method that leverages a randomized sketching technique to perform an approximate Newton step in each iteration, thereby eliminating the computational bottleneck of second-order methods. While existing studies have established the asymptotic normality of sketched Newton methods, a consistent estimator of the limiting covariance matrix remains an open problem. We propose a fully online covariance matrix estimator that is constructed entirely from the Newton iterates and requires no matrix factorization. Compared to covariance estimators for first-order online methods, our estimator for second-order methods is batch-free. We establish the consistency and convergence rate of our estimator, and coupled with asymptotic normality results, we can then perform online statistical inference for the model parameters based on sketched Newton methods. We also discuss the extension of our estimator to constrained problems, and demonstrate its superior performance on regression problems as well as benchmark problems in the CUTEst set.
2502.07115
Online Scheduling for LLM Inference with KV Cache Constraints
cs.LG cs.AI math.OC
Large Language Model (LLM) inference, where a trained model generates text one word at a time in response to user prompts, is a computationally intensive process requiring efficient scheduling to optimize latency and resource utilization. A key challenge in LLM inference is the management of the Key-Value (KV) cache, which reduces redundant computations but introduces memory constraints. In this work, we model LLM inference with KV cache constraints theoretically and propose novel batching and scheduling algorithms that minimize inference latency while effectively managing the KV cache's memory. We analyze both semi-online and fully online scheduling models, and our results are threefold. First, we provide a polynomial-time algorithm that achieves exact optimality in terms of average latency in the semi-online prompt arrival model. Second, in the fully online case with a stochastic prompt arrival, we introduce an efficient online scheduling algorithm with constant regret. Third, we prove that no algorithm (deterministic or randomized) can achieve a constant competitive ratio in fully online adversarial settings. Our empirical evaluations on a public LLM inference dataset, using the Llama-70B model on A100 GPUs, show that our approach significantly outperforms benchmark algorithms used currently in practice, achieving lower latency while reducing energy consumption. Overall, our results offer a path toward more sustainable and cost-effective LLM deployment.
2502.07117
Choroidal image analysis for OCT image sequences with applications in systemic health
eess.IV cs.CV cs.LG cs.MS
The choroid, a highly vascular layer behind the retina, is an extension of the central nervous system and has parallels with the renal cortex, with blood flow far exceeding that of the brain and kidney. Thus, there has been growing interest of choroidal blood flow reflecting physiological status of systemic disease. Optical coherence tomography (OCT) enables high-resolution imaging of the choroid, but conventional analysis methods remain manual or semi-automatic, limiting reproducibility, standardisation and clinical utility. In this thesis, I develop several new methods to analyse the choroid in OCT image sequences, with each successive method improving on its predecessors. I first develop two semi-automatic approaches for choroid region (Gaussian Process Edge Tracing, GPET) and vessel (Multi-scale Median Cut Quantisation, MMCQ) analysis, which improve on manual approaches but remain user-dependent. To address this, I introduce DeepGPET, a deep learning-based region segmentation method which improves on execution time, reproducibility, and end-user accessibility, but lacks choroid vessel analysis and automatic feature measurement. Improving on this, I developed Choroidalyzer, a deep learning-based pipeline to segment the choroidal space and vessels and generate fully automatic, clinically meaningful and reproducible choroidal features. I provide rigorous evaluation of these four approaches and consider their potential clinical value in three applications into systemic health: OCTANE, assessing choroidal changes in renal transplant recipients and donors; PREVENT, exploring choroidal associations with Alzheimer's risk factors at mid-life; D-RISCii, assessing choroidal variation and feasibility of OCT in critical care. In short, this thesis contributes many open-source tools for standardised choroidal measurement and highlights the choroid's potential as a biomarker in systemic health.
2502.07119
SAFE: Self-Supervised Anomaly Detection Framework for Intrusion Detection
cs.CR cs.LG
The proliferation of IoT devices has significantly increased network vulnerabilities, creating an urgent need for effective Intrusion Detection Systems (IDS). Machine Learning-based IDS (ML-IDS) offer advanced detection capabilities but rely on labeled attack data, which limits their ability to identify unknown threats. Self-Supervised Learning (SSL) presents a promising solution by using only normal data to detect patterns and anomalies. This paper introduces SAFE, a novel framework that transforms tabular network intrusion data into an image-like format, enabling Masked Autoencoders (MAEs) to learn robust representations of network behavior. The features extracted by the MAEs are then incorporated into a lightweight novelty detector, enhancing the effectiveness of anomaly detection. Experimental results demonstrate that SAFE outperforms the state-of-the-art anomaly detection method, Scale Learning-based Deep Anomaly Detection method (SLAD), by up to 26.2% and surpasses the state-of-the-art SSL-based network intrusion detection approach, Anomal-E, by up to 23.5% in F1-score.
2502.07120
Is Long Range Sequential Modeling Necessary For Colorectal Tumor Segmentation?
cs.CV
Segmentation of colorectal cancer (CRC) tumors in 3D medical imaging is both complex and clinically critical, providing vital support for effective radiation therapy planning and survival outcome assessment. Recently, 3D volumetric segmentation architectures incorporating long-range sequence modeling mechanisms, such as Transformers and Mamba, have gained attention for their capacity to achieve high accuracy in 3D medical image segmentation. In this work, we evaluate the effectiveness of these global token modeling techniques by pitting them against our proposed MambaOutUNet within the context of our newly introduced colorectal tumor segmentation dataset (CTS-204). Our findings suggest that robust local token interactions can outperform long-range modeling techniques in cases where the region of interest is small and anatomically complex, proposing a potential shift in 3D tumor segmentation research.
2502.07124
Structural Reformation of Large Language Model Neuron Encapsulation for Divergent Information Aggregation
cs.CL
Structured neuron encapsulation introduces a modular framework that enables more effective aggregation and specialization of information within deep learning architectures. A model modified through this framework demonstrated improved perplexity scores, greater lexical variability, and enhanced consistency in logical reasoning, suggesting that structured parameter distribution contributes to more efficient language representation. Statistical analyses of generated text highlighted a wider range of sentence structures and reduced redundancy in token selection, indicating that encapsulation fosters more adaptable language generation. A detailed evaluation of attention weight distributions revealed that the experimental model exhibited greater divergence in cross-layer activations, supporting the hypothesis that encapsulated neurons assume specialized processing roles. Logical consistency assessments further demonstrated that modular architectures mitigate contradictory outputs, reducing internal conflicts in inferred relationships between linguistic constructs. Computational trade-offs were analyzed, with results showing a minor increase in processing overhead, though improvements in parameter efficiency and structured decision-making compensated for the additional complexity. The mathematical formulation of the encapsulation mechanism confirmed that modular aggregation maintains stable convergence properties while promoting distinct functional roles for different neuron clusters.
2502.07128
Cardiverse: Harnessing LLMs for Novel Card Game Prototyping
cs.CL cs.AI cs.MM
The prototyping of computer games, particularly card games, requires extensive human effort in creative ideation and gameplay evaluation. Recent advances in Large Language Models (LLMs) offer opportunities to automate and streamline these processes. However, it remains challenging for LLMs to design novel game mechanics beyond existing databases, generate consistent gameplay environments, and develop scalable gameplay AI for large-scale evaluations. This paper addresses these challenges by introducing a comprehensive automated card game prototyping framework. The approach highlights a graph-based indexing method for generating novel game designs, an LLM-driven system for consistent game code generation validated by gameplay records, and a gameplay AI constructing method that uses an ensemble of LLM-generated action-value functions optimized through self-play. These contributions aim to accelerate card game prototyping, reduce human labor, and lower barriers to entry for game developers.
2502.07129
Fourier-enhanced Neural Networks For Systems Biology Applications
cs.LG q-bio.QM
In the field of systems biology, differential equations are commonly used to model biological systems, but solving them for large-scale and complex systems can be computationally expensive. Recently, the integration of machine learning and mathematical modeling has offered new opportunities for scientific discoveries in biology and health. The emerging physics-informed neural network (PINN) has been proposed as a solution to this problem. However, PINN can be computationally expensive and unreliable for complex biological systems. To address these issues, we propose the Fourier-enhanced Neural Networks for systems biology (SB-FNN). SB-FNN uses an embedded Fourier neural network with an adaptive activation function and a cyclic penalty function to optimize the prediction of biological dynamics, particularly for biological systems that exhibit oscillatory patterns. Experimental results demonstrate that SB-FNN achieves better performance and is more efficient than PINN for handling complex biological models. Experimental results on cellular and population models demonstrate that SB-FNN outperforms PINN in both accuracy and efficiency, making it a promising alternative approach for handling complex biological models. The proposed method achieved better performance on six biological models and is expected to replace PINN as the most advanced method in systems biology.
2502.07130
Unconstrained Body Recognition at Altitude and Range: Comparing Four Approaches
cs.CV cs.AI cs.LG
This study presents an investigation of four distinct approaches to long-term person identification using body shape. Unlike short-term re-identification systems that rely on temporary features (e.g., clothing), we focus on learning persistent body shape characteristics that remain stable over time. We introduce a body identification model based on a Vision Transformer (ViT) (Body Identification from Diverse Datasets, BIDDS) and on a Swin-ViT model (Swin-BIDDS). We also expand on previous approaches based on the Linguistic and Non-linguistic Core ResNet Identity Models (LCRIM and NLCRIM), but with improved training. All models are trained on a large and diverse dataset of over 1.9 million images of approximately 5k identities across 9 databases. Performance was evaluated on standard re-identification benchmark datasets (MARS, MSMT17, Outdoor Gait, DeepChange) and on an unconstrained dataset that includes images at a distance (from close-range to 1000m), at altitude (from an unmanned aerial vehicle, UAV), and with clothing change. A comparative analysis across these models provides insights into how different backbone architectures and input image sizes impact long-term body identification performance across real-world conditions.
2502.07131
TWICE: What Advantages Can Low-Resource Domain-Specific Embedding Model Bring? -- A Case Study on Korea Financial Texts
cs.CL q-fin.CP
Domain specificity of embedding models is critical for effective performance. However, existing benchmarks, such as FinMTEB, are primarily designed for high-resource languages, leaving low-resource settings, such as Korean, under-explored. Directly translating established English benchmarks often fails to capture the linguistic and cultural nuances present in low-resource domains. In this paper, titled TWICE: What Advantages Can Low-Resource Domain-Specific Embedding Models Bring? A Case Study on Korea Financial Texts, we introduce KorFinMTEB, a novel benchmark for the Korean financial domain, specifically tailored to reflect its unique cultural characteristics in low-resource languages. Our experimental results reveal that while the models perform robustly on a translated version of FinMTEB, their performance on KorFinMTEB uncovers subtle yet critical discrepancies, especially in tasks requiring deeper semantic understanding, that underscore the limitations of direct translation. This discrepancy highlights the necessity of benchmarks that incorporate language-specific idiosyncrasies and cultural nuances. The insights from our study advocate for the development of domain-specific evaluation frameworks that can more accurately assess and drive the progress of embedding models in low-resource settings.
2502.07132
Interactive Data Harmonization with LLM Agents
cs.AI cs.DB
Data harmonization is an essential task that entails integrating datasets from diverse sources. Despite years of research in this area, it remains a time-consuming and challenging task due to schema mismatches, varying terminologies, and differences in data collection methodologies. This paper presents the case for agentic data harmonization as a means to both empower experts to harmonize their data and to streamline the process. We introduce Harmonia, a system that combines LLM-based reasoning, an interactive user interface, and a library of data harmonization primitives to automate the synthesis of data harmonization pipelines. We demonstrate Harmonia in a clinical data harmonization scenario, where it helps to interactively create reusable pipelines that map datasets to a standard format. Finally, we discuss challenges and open problems, and suggest research directions for advancing our vision.
2502.07133
Cross-platform Learning-based Fault Tolerant Surfacing Controller for Underwater Robots
cs.RO
In this paper, we propose a novel cross-platform fault-tolerant surfacing controller for underwater robots, based on reinforcement learning (RL). Unlike conventional approaches, which require explicit identification of malfunctioning actuators, our method allows the robot to surface using only the remaining operational actuators without needing to pinpoint the failures. The proposed controller learns a robust policy capable of handling diverse failure scenarios across different actuator configurations. Moreover, we introduce a transfer learning mechanism that shares a part of the control policy across various underwater robots with different actuators, thus improving learning efficiency and generalization across platforms. To validate our approach, we conduct simulations on three different types of underwater robots: a hovering-type AUV, a torpedo shaped AUV, and a turtle-shaped robot (U-CAT). Additionally, real-world experiments are performed, successfully transferring the learned policy from simulation to a physical U-CAT in a controlled environment. Our RL-based controller demonstrates superior performance in terms of stability and success rate compared to a baseline controller, achieving an 85.7 percent success rate in real-world tests compared to 57.1 percent with a baseline controller. This research provides a scalable and efficient solution for fault-tolerant control for diverse underwater platforms, with potential applications in real-world aquatic missions.
2502.07135
One-Shot Learning for k-SAT
cs.DS cs.LG math.ST stat.ML stat.TH
Consider a $k$-SAT formula $\Phi$ where every variable appears at most $d$ times, and let $\sigma$ be a satisfying assignment of $\Phi$ sampled proportionally to $e^{\beta m(\sigma)}$ where $m(\sigma)$ is the number of variables set to true and $\beta$ is a real parameter. Given $\Phi$ and $\sigma$, can we learn the value of $\beta$ efficiently? This problem falls into a recent line of works about single-sample ("one-shot") learning of Markov random fields. The $k$-SAT setting we consider here was recently studied by Galanis, Kandiros, and Kalavasis (SODA'24) where they showed that single-sample learning is possible when roughly $d\leq 2^{k/6.45}$ and impossible when $d\geq (k+1) 2^{k-1}$. Crucially, for their impossibility results they used the existence of unsatisfiable instances which, aside from the gap in $d$, left open the question of whether the feasibility threshold for one-shot learning is dictated by the satisfiability threshold of $k$-SAT formulas of bounded degree. Our main contribution is to answer this question negatively. We show that one-shot learning for $k$-SAT is infeasible well below the satisfiability threshold; in fact, we obtain impossibility results for degrees $d$ as low as $k^2$ when $\beta$ is sufficiently large, and bootstrap this to small values of $\beta$ when $d$ scales exponentially with $k$, via a probabilistic construction. On the positive side, we simplify the analysis of the learning algorithm and obtain significantly stronger bounds on $d$ in terms of $\beta$. In particular, for the uniform case $\beta\rightarrow 0$ that has been studied extensively in the sampling literature, our analysis shows that learning is possible under the condition $d\lesssim 2^{k/2}$. This is nearly optimal (up to constant factors) in the sense that it is known that sampling a uniformly-distributed satisfying assignment is NP-hard for $d\gtrsim 2^{k/2}$.
2502.07136
A Safe Hybrid Control Framework for Car-like Robot with Guaranteed Global Path-Invariance using a Control Barrier Function
cs.RO cs.SY eess.SY
This work proposes a hybrid framework for car-like robots with obstacle avoidance, global convergence, and safety, where safety is interpreted as path invariance, namely, once the robot converges to the path, it never leaves the path. Given a priori obstacle-free feasible path where obstacles can be around the path, the task is to avoid obstacles while reaching the path and then staying on the path without leaving it. The problem is solved in two stages. Firstly, we define a ``tight'' obstacle-free neighborhood along the path and design a local controller to ensure convergence to the path and path invariance. The control barrier function technology is involved in the control design to steer the system away from its singularity points, where the local path invariant controller is not defined. Secondly, we design a hybrid control framework that integrates this local path-invariant controller with any global tracking controller from the existing literature without path invariance guarantee, ensuring convergence from any position to the desired path, namely, global convergence. This framework guarantees path invariance and robustness to sensor noise. Detailed simulation results affirm the effectiveness of the proposed scheme.
2502.07138
Towards a Robust Framework for Multimodal Hate Detection: A Study on Video vs. Image-based Content
cs.CV cs.CL cs.LG
Social media platforms enable the propagation of hateful content across different modalities such as textual, auditory, and visual, necessitating effective detection methods. While recent approaches have shown promise in handling individual modalities, their effectiveness across different modality combinations remains unexplored. This paper presents a systematic analysis of fusion-based approaches for multimodal hate detection, focusing on their performance across video and image-based content. Our comprehensive evaluation reveals significant modality-specific limitations: while simple embedding fusion achieves state-of-the-art performance on video content (HateMM dataset) with a 9.9% points F1-score improvement, it struggles with complex image-text relationships in memes (Hateful Memes dataset). Through detailed ablation studies and error analysis, we demonstrate how current fusion approaches fail to capture nuanced cross-modal interactions, particularly in cases involving benign confounders. Our findings provide crucial insights for developing more robust hate detection systems and highlight the need for modality-specific architectural considerations. The code is available at https://github.com/gak97/Video-vs-Meme-Hate.
2502.07139
Language-TPP: Integrating Temporal Point Processes with Language Models for Event Analysis
cs.CL cs.LG
Temporal Point Processes (TPPs) have been widely used for event sequence modeling, but they often struggle to incorporate rich textual event descriptions effectively. Conversely, while Large Language Models (LLMs) have been shown remarkable capabilities in processing textual data, they lack mechanisms for handling temporal dynamics. To bridge this gap, we introduce Language-TPP, a unified framework that integrates TPPs with LLMs for enhanced event sequence modeling. Language-TPP introduces a novel temporal encoding mechanism that converts continuous time intervals into specialized byte-tokens, enabling seamless integration with standard LLM architectures. This approach allows Language-TPP to achieve state-of-the-art performance across multiple TPP tasks, including event time prediction, type prediction, and intensity estimation, on five datasets. Additionally, we demonstrate that incorporating temporal information significantly improves the quality of generated event descriptions.
2502.07140
Few-Shot Multi-Human Neural Rendering Using Geometry Constraints
cs.CV cs.AI cs.GR
We present a method for recovering the shape and radiance of a scene consisting of multiple people given solely a few images. Multi-human scenes are complex due to additional occlusion and clutter. For single-human settings, existing approaches using implicit neural representations have achieved impressive results that deliver accurate geometry and appearance. However, it remains challenging to extend these methods for estimating multiple humans from sparse views. We propose a neural implicit reconstruction method that addresses the inherent challenges of this task through the following contributions: First, we propose to use geometry constraints by exploiting pre-computed meshes using a human body model (SMPL). Specifically, we regularize the signed distances using the SMPL mesh and leverage bounding boxes for improved rendering. Second, we propose a ray regularization scheme to minimize rendering inconsistencies, and a saturation regularization for robust optimization in variable illumination. Extensive experiments on both real and synthetic datasets demonstrate the benefits of our approach and show state-of-the-art performance against existing neural reconstruction methods.
2502.07141
Small steps no more: Global convergence of stochastic gradient bandits for arbitrary learning rates
cs.LG
We provide a new understanding of the stochastic gradient bandit algorithm by showing that it converges to a globally optimal policy almost surely using \emph{any} constant learning rate. This result demonstrates that the stochastic gradient algorithm continues to balance exploration and exploitation appropriately even in scenarios where standard smoothness and noise control assumptions break down. The proofs are based on novel findings about action sampling rates and the relationship between cumulative progress and noise, and extend the current understanding of how simple stochastic gradient methods behave in bandit settings.
2502.07143
Ask Patients with Patience: Enabling LLMs for Human-Centric Medical Dialogue with Grounded Reasoning
cs.CL
Accurate and efficient diagnosis in online medical consultations remains a challenge for current large language models. These models often rely on single-turn interactions and lack the ability to refine their predictions through follow-up questions. Additionally, their responses frequently contain complex medical terminology, making them less accessible to non-medical users and creating barriers to effective communication. In this paper, we introduce Ask Patients with Patience (APP), the first multi-turn dialogue that enables LLMs to iteratively refine diagnoses based on grounded reasoning. By integrating medical guidelines and entropy minimization, APP improves both diagnostic accuracy and efficiency. Furthermore, it features human-centric communication that bridges the gap between user comprehension and medical terminology, significantly enhancing user accessibility and engagement. We evaluated APP using a subset of the ReMeDi dataset, comparing it with single-turn and traditional multi-turn LLM baselines. APP achieved higher similarity scores in diagnosis predictions, demonstrating better alignment with ground truth diagnoses. Entropy analysis showed that APP reduces diagnostic uncertainty more rapidly across iterations, increasing confidence in its predictions. APP also excels in user accessibility and empathy, further bridging the gap between complex medical language and user understanding. Code will be released at: https://github.com/SuperMedIntel/AskPatients.
2502.07145
Mesh2SSM++: A Probabilistic Framework for Unsupervised Learning of Statistical Shape Model of Anatomies from Surface Meshes
cs.CV
Anatomy evaluation is crucial for understanding the physiological state, diagnosing abnormalities, and guiding medical interventions. Statistical shape modeling (SSM) is vital in this process. By enabling the extraction of quantitative morphological shape descriptors from MRI and CT scans, SSM provides comprehensive descriptions of anatomical variations within a population. However, the effectiveness of SSM in anatomy evaluation hinges on the quality and robustness of the shape models. While deep learning techniques show promise in addressing these challenges by learning complex nonlinear representations of shapes, existing models still have limitations and often require pre-established shape models for training. To overcome these issues, we propose Mesh2SSM++, a novel approach that learns to estimate correspondences from meshes in an unsupervised manner. This method leverages unsupervised, permutation-invariant representation learning to estimate how to deform a template point cloud into subject-specific meshes, forming a correspondence-based shape model. Additionally, our probabilistic formulation allows learning a population-specific template, reducing potential biases associated with template selection. A key feature of Mesh2SSM++ is its ability to quantify aleatoric uncertainty, which captures inherent data variability and is essential for ensuring reliable model predictions and robust decision-making in clinical tasks, especially under challenging imaging conditions. Through extensive validation across diverse anatomies, evaluation metrics, and downstream tasks, we demonstrate that Mesh2SSM++ outperforms existing methods. Its ability to operate directly on meshes, combined with computational efficiency and interpretability through its probabilistic framework, makes it an attractive alternative to traditional and deep learning-based SSM approaches.
2502.07148
Expressing entropy and cross-entropy in expansions of common meadows
cs.IT math.IT
A common meadow is an enrichment of a field with a partial division operation that is made total by assuming that division by zero takes the a default value, a special element $\bot$ adjoined to the field. To a common meadow of real numbers we add a binary logarithm $\log_2(-)$, which we also assume to be total with $\log_2(p) = \bot$ for $p \leq 0$. With these and other auxiliary operations, such as a sign function, we form algebras over which entropy and cross entropy can be defined for probability mass functions on a finite sample space by algebraic formulae that are simple terms built from the operations of the algebras and without case distinctions or conventions to avoid partiality. The discuss the advantages of algebras based on common meadows, whose theory is established, and alternate methods to define entropy and other information measures completely for all arguments using single terms.
2502.07151
Conditional Distribution Quantization in Machine Learning
cs.LG
Conditional expectation \mathbb{E}(Y \mid X) often fails to capture the complexity of multimodal conditional distributions \mathcal{L}(Y \mid X). To address this, we propose using n-point conditional quantizations--functional mappings of X that are learnable via gradient descent--to approximate \mathcal{L}(Y \mid X). This approach adapts Competitive Learning Vector Quantization (CLVQ), tailored for conditional distributions. It goes beyond single-valued predictions by providing multiple representative points that better reflect multimodal structures. It enables the approximation of the true conditional law in the Wasserstein distance. The resulting framework is theoretically grounded and useful for uncertainty quantification and multimodal data generation tasks. For example, in computer vision inpainting tasks, multiple plausible reconstructions may exist for the same partially observed input image X. We demonstrate the effectiveness of our approach through experiments on synthetic and real-world datasets.
2502.07153
Feature Importance Depends on Properties of the Data: Towards Choosing the Correct Explanations for Your Data and Decision Trees based Models
cs.LG cs.AI
In order to ensure the reliability of the explanations of machine learning models, it is crucial to establish their advantages and limits and in which case each of these methods outperform. However, the current understanding of when and how each method of explanation can be used is insufficient. To fill this gap, we perform a comprehensive empirical evaluation by synthesizing multiple datasets with the desired properties. Our main objective is to assess the quality of feature importance estimates provided by local explanation methods, which are used to explain predictions made by decision tree-based models. By analyzing the results obtained from synthetic datasets as well as publicly available binary classification datasets, we observe notable disparities in the magnitude and sign of the feature importance estimates generated by these methods. Moreover, we find that these estimates are sensitive to specific properties present in the data. Although some model hyper-parameters do not significantly influence feature importance assignment, it is important to recognize that each method of explanation has limitations in specific contexts. Our assessment highlights these limitations and provides valuable insight into the suitability and reliability of different explanatory methods in various scenarios.
2502.07154
Rethinking Fine-Tuning when Scaling Test-Time Compute: Limiting Confidence Improves Mathematical Reasoning
cs.LG cs.AI
Recent progress in large language models (LLMs) highlights the power of scaling test-time compute to achieve strong performance on complex tasks, such as mathematical reasoning and code generation. This raises a critical question: how should model training be modified to optimize performance under a subsequent test-time compute strategy and budget? To explore this, we focus on pass@N, a simple test-time strategy that searches for a correct answer in $N$ independent samples. We show, surprisingly, that training with cross-entropy (CE) loss can be ${\it misaligned}$ with pass@N in that pass@N accuracy ${\it decreases}$ with longer training. We explain the origins of this misalignment in terms of model overconfidence induced by CE, and experimentally verify our prediction of overconfidence as an impediment to scaling test-time compute via pass@N. Furthermore we suggest a principled, modified training loss that is better aligned to pass@N by limiting model confidence and rescuing pass@N test performance. Our algorithm demonstrates improved mathematical reasoning on MATH and MiniF2F benchmarks under several scenarios: (1) providing answers to math questions; and (2) proving theorems by searching over proof trees of varying shapes. Overall our work underscores the importance of co-designing two traditionally separate phases of LLM development: training-time protocols and test-time search and reasoning strategies.
2502.07156
Explaining 3D Computed Tomography Classifiers with Counterfactuals
cs.CV cs.AI
Counterfactual explanations in medical imaging are critical for understanding the predictions made by deep learning models. We extend the Latent Shift counterfactual generation method from 2D applications to 3D computed tomography (CT) scans. We address the challenges associated with 3D data, such as limited training samples and high memory demands, by implementing a slice-based approach. This method leverages a 2D encoder trained on CT slices, which are subsequently combined to maintain 3D context. We demonstrate this technique on two models for clinical phenotype prediction and lung segmentation. Our approach is both memory-efficient and effective for generating interpretable counterfactuals in high-resolution 3D medical imaging.
2502.07158
Early Risk Prediction of Pediatric Cardiac Arrest from Electronic Health Records via Multimodal Fused Transformer
cs.LG cs.AI
Early prediction of pediatric cardiac arrest (CA) is critical for timely intervention in high-risk intensive care settings. We introduce PedCA-FT, a novel transformer-based framework that fuses tabular view of EHR with the derived textual view of EHR to fully unleash the interactions of high-dimensional risk factors and their dynamics. By employing dedicated transformer modules for each modality view, PedCA-FT captures complex temporal and contextual patterns to produce robust CA risk estimates. Evaluated on a curated pediatric cohort from the CHOA-CICU database, our approach outperforms ten other artificial intelligence models across five key performance metrics and identifies clinically meaningful risk factors. These findings underscore the potential of multimodal fusion techniques to enhance early CA detection and improve patient care.
2502.07160
HDCompression: Hybrid-Diffusion Image Compression for Ultra-Low Bitrates
cs.CV cs.MM
Image compression under ultra-low bitrates remains challenging for both conventional learned image compression (LIC) and generative vector-quantized (VQ) modeling. Conventional LIC suffers from severe artifacts due to heavy quantization, while generative VQ modeling gives poor fidelity due to the mismatch between learned generative priors and specific inputs. In this work, we propose Hybrid-Diffusion Image Compression (HDCompression), a dual-stream framework that utilizes both generative VQ-modeling and diffusion models, as well as conventional LIC, to achieve both high fidelity and high perceptual quality. Different from previous hybrid methods that directly use pre-trained LIC models to generate low-quality fidelity-preserving information from heavily quantized latent, we use diffusion models to extract high-quality complimentary fidelity information from the ground-truth input, which can enhance the system performance in several aspects: improving indices map prediction, enhancing the fidelity-preserving output of the LIC stream, and refining conditioned image reconstruction with VQ-latent correction. In addition, our diffusion model is based on a dense representative vector (DRV), which is lightweight with very simple sampling schedulers. Extensive experiments demonstrate that our HDCompression outperforms the previous conventional LIC, generative VQ-modeling, and hybrid frameworks in both quantitative metrics and qualitative visualization, providing balanced robust compression performance at ultra-low bitrates.
2502.07161
A Survey on Mamba Architecture for Vision Applications
cs.CV cs.AI
Transformers have become foundational for visual tasks such as object detection, semantic segmentation, and video understanding, but their quadratic complexity in attention mechanisms presents scalability challenges. To address these limitations, the Mamba architecture utilizes state-space models (SSMs) for linear scalability, efficient processing, and improved contextual awareness. This paper investigates Mamba architecture for visual domain applications and its recent advancements, including Vision Mamba (ViM) and VideoMamba, which introduce bidirectional scanning, selective scanning mechanisms, and spatiotemporal processing to enhance image and video understanding. Architectural innovations like position embeddings, cross-scan modules, and hierarchical designs further optimize the Mamba framework for global and local feature extraction. These advancements position Mamba as a promising architecture in computer vision research and applications.
2502.07164
Does Training on Synthetic Data Make Models Less Robust?
cs.CL cs.AI cs.LG
An increasingly common practice is to train large language models (LLMs) using synthetic data. Often this synthetic data is produced by the same or similar LLMs as those it is being used to train. This raises the question of whether the synthetic data might in fact exacerbate certain "blindspots" by reinforcing heuristics that the LLM already encodes. In this paper, we conduct simulated experiments on the natural language inference (NLI) task with Llama-2-7B-hf models. We use MultiNLI as the general task and HANS, a targeted evaluation set designed to measure the presence of specific heuristic strategies for NLI, as our "blindspot" task. Our goal is to determine whether performance disparities between the general and blind spot tasks emerge. Our results indicate that synthetic data does not reinforce blindspots in the way we expected. Specifically, we see that, while fine-tuning with synthetic data doesn't necessarily reduce the use of the heuristic, it also does not make it worse as we hypothesized.
2502.07165
Don't Just Demo, Teach Me the Principles: A Principle-Based Multi-Agent Prompting Strategy for Text Classification
cs.CL cs.AI
We present PRINCIPLE-BASED PROMPTING, a simple but effective multi-agent prompting strategy for text classification. It first asks multiple LLM agents to independently generate candidate principles based on analysis of demonstration samples with or without labels, consolidates them into final principles via a finalizer agent, and then sends them to a classifier agent to perform downstream classification tasks. Extensive experiments on binary and multi-class classification datasets with different sizes of LLMs show that our approach not only achieves substantial performance gains (1.55% - 19.37%) over zero-shot prompting on macro-F1 score but also outperforms other strong baselines (CoT and stepback prompting). Principles generated by our approach help LLMs perform better on classification tasks than human crafted principles on two private datasets. Our multi-agent PRINCIPLE-BASED PROMPTING approach also shows on-par or better performance compared to demonstration-based few-shot prompting approaches, yet with substantially lower inference costs. Ablation studies show that label information and the multi-agent cooperative LLM framework play an important role in generating high-quality principles to facilitate downstream classification tasks.
2502.07166
Bayesian Optimization for Building Social-Influence-Free Consensus
cs.MA cs.GT cs.LG stat.ML
We introduce Social Bayesian Optimization (SBO), a vote-efficient algorithm for consensus-building in collective decision-making. In contrast to single-agent scenarios, collective decision-making encompasses group dynamics that may distort agents' preference feedback, thereby impeding their capacity to achieve a social-influence-free consensus -- the most preferable decision based on the aggregated agent utilities. We demonstrate that under mild rationality axioms, reaching social-influence-free consensus using noisy feedback alone is impossible. To address this, SBO employs a dual voting system: cheap but noisy public votes (e.g., show of hands in a meeting), and more accurate, though expensive, private votes (e.g., one-to-one interview). We model social influence using an unknown social graph and leverage the dual voting system to efficiently learn this graph. Our theoretical findigns show that social graph estimation converges faster than the black-box estimation of agents' utilities, allowing us to reduce reliance on costly private votes early in the process. This enables efficient consensus-building primarily through noisy public votes, which are debiased based on the estimated social graph to infer social-influence-free feedback. We validate the efficacy of SBO across multiple real-world applications, including thermal comfort, team building, travel negotiation, and energy trading collaboration.
2502.07169
Advancing Geological Carbon Storage Monitoring With 3d Digital Shadow Technology
physics.comp-ph cs.LG physics.geo-ph
Geological Carbon Storage (GCS) is a key technology for achieving global climate goals by capturing and storing CO2 in deep geological formations. Its effectiveness and safety rely on accurate monitoring of subsurface CO2 migration using advanced time-lapse seismic imaging. A Digital Shadow framework integrates field data, including seismic and borehole measurements, to track CO2 saturation over time. Machine learning-assisted data assimilation techniques, such as generative AI and nonlinear ensemble Bayesian filtering, update a digital model of the CO2 plume while incorporating uncertainties in reservoir properties. Compared to 2D approaches, 3D monitoring enhances the spatial accuracy of GCS assessments, capturing the full extent of CO2 migration. This study extends the uncertainty-aware 2D Digital Shadow framework by incorporating 3D seismic imaging and reservoir modeling, improving decision-making and risk mitigation in CO2 storage projects.
2502.07170
Effcient classical error correction for parity encoded spin systems
quant-ph cs.IT math.IT
Fast solvers for combinatorial optimization problems (COPs) have attracted engineering interest in various industrial and social applications. Quantum annealing (QA) has emerged as a promising candidate and significant efforts have been dedicated to its development. Since COP is encoded in the Ising interaction between logical spins, its realization requires a spin system with all-to-all connectivity, which poses technical difficulties in the physical implementation of large-scale QA devices. W. Lechner, P. Hauke, and P. Zoller proposed parity-encoding (PE) architecture, consisting of a larger system of physical spins with only local connectivities between them, to avoid this diffculty in the near future QA device development. They suggested that this architecture not only reduces implementation diffculties and improves scalability, but also has intrinsic fault tolerance because logical spins are redundantly and nonlocally encoded into the physical spins. Nevertheless, it remains unclear how these advantageous features can be exploited. This paper addresses how to correct errors in a spin readout of PE architecture. Our work is based on the close connection between PE architecture and classical low-density parity-check (LDPC) codes. We have shown that independent and identically distributed errors in a spin readout can be corrected by a very simple decoding algorithm that can be regarded as a bit flipping (BF) algorithm for the LDPC codes. The BF algorithm was shown to have comparable performance to the belief propagation (BP) decoding algorithm. Furthermore, it is suggested that the introduction of post-readout BF decoding reduces the total computational cost and improves the performance of the global optimal solution search using the PE architecture. We believe that our results indicate that the PE architecture is a promising platform for near-term QA devices.
2502.07171
Enhancing Robustness Of Digital Shadow For CO2 Storage Monitoring With Augmented Rock Physics Modeling
physics.comp-ph cs.LG physics.geo-ph
To meet climate targets, the IPCC underscores the necessity of technologies capable of removing gigatonnes of CO2 annually, with Geological Carbon Storage (GCS) playing a central role. GCS involves capturing CO2 and injecting it into deep geological formations for long-term storage, requiring precise monitoring to ensure containment and prevent leakage. Time-lapse seismic imaging is essential for tracking CO2 migration but often struggles to capture the complexities of multi-phase subsurface flow. Digital Shadows (DS), leveraging machine learning-driven data assimilation techniques such as nonlinear Bayesian filtering and generative AI, provide a more detailed, uncertainty-aware monitoring approach. By incorporating uncertainties in reservoir properties, DS frameworks improve CO2 migration forecasts, reducing risks in GCS operations. However, data assimilation depends on assumptions regarding reservoir properties, rock physics models, and initial conditions, which, if inaccurate, can compromise prediction reliability. This study demonstrates that augmenting forecast ensembles with diverse rock physics models mitigates the impact of incorrect assumptions and improves predictive accuracy, particularly in differentiating uniform versus patchy saturation models.
2502.07172
SemiHMER: Semi-supervised Handwritten Mathematical Expression Recognition using pseudo-labels
cs.CV cs.AI
In this paper, we study semi-supervised Handwritten Mathematical Expression Recognition (HMER) via exploring both labeled data and extra unlabeled data. We propose a novel consistency regularization framework, termed SemiHMER, which introduces dual-branch semi-supervised learning. Specifically, we enforce consistency between the two networks for the same input image. The pseudo-label, generated by one perturbed recognition network, is utilized to supervise the other network using the standard cross-entropy loss. The SemiHMER consistency encourages high similarity between the predictions of the two perturbed networks for the same input image and expands the training data by leveraging unlabeled data with pseudo-labels. We further introduce a weak-to-strong strategy by applying different levels of augmentation to each branch, effectively expanding the training data and enhancing the quality of network training. Additionally, we propose a novel module, the Global Dynamic Counting Module (GDCM), to enhance the performance of the HMER decoder by alleviating recognition inaccuracies in long-distance formula recognition and reducing the occurrence of repeated characters. The experimental results demonstrate that our work achieves significant performance improvements, with an average accuracy increase of 5.47% on CROHME14, 4.87% on CROHME16, and 5.25% on CROHME19, compared to our baselines.
2502.07175
Foreign-Object Detection in High-Voltage Transmission Line Based on Improved YOLOv8m
cs.CV cs.AI
The safe operation of high-voltage transmission lines ensures the power grid's security. Various foreign objects attached to the transmission lines, such as balloons, kites and nesting birds, can significantly affect the safe and stable operation of high-voltage transmission lines. With the advancement of computer vision technology, periodic automatic inspection of foreign objects is efficient and necessary. Existing detection methods have low accuracy because foreign objects at-tached to the transmission lines are complex, including occlusions, diverse object types, significant scale variations, and complex backgrounds. In response to the practical needs of the Yunnan Branch of China Southern Power Grid Co., Ltd., this paper proposes an improved YOLOv8m-based model for detecting foreign objects on transmission lines. Experiments are conducted on a dataset collected from Yunnan Power Grid. The proposed model enhances the original YOLOv8m by in-corporating a Global Attention Module (GAM) into the backbone to focus on occluded foreign objects, replacing the SPPF module with the SPPCSPC module to augment the model's multiscale feature extraction capability, and introducing the Focal-EIoU loss function to address the issue of high- and low-quality sample imbalances. These improvements accelerate model convergence and enhance detection accuracy. The experimental results demonstrate that our proposed model achieves a 2.7% increase in mAP_0.5, a 4% increase in mAP_0.5:0.95, and a 6% increase in recall.
2502.07176
MatrixKAN: Parallelized Kolmogorov-Arnold Network
cs.LG
Kolmogorov-Arnold Networks (KAN) are a new class of neural network architecture representing a promising alternative to the Multilayer Perceptron (MLP), demonstrating improved expressiveness and interpretability. However, KANs suffer from slow training and inference speeds relative to MLPs due in part to the recursive nature of the underlying B-spline calculations. This issue is particularly apparent with respect to KANs utilizing high-degree B-splines, as the number of required non-parallelizable recursions is proportional to B-spline degree. We solve this issue by proposing MatrixKAN, a novel optimization that parallelizes B-spline calculations with matrix representation and operations, thus significantly improving effective computation time for models utilizing high-degree B-splines. In this paper, we demonstrate the superior scaling of MatrixKAN's computation time relative to B-spline degree. Further, our experiments demonstrate speedups of approximately 40x relative to KAN, with significant additional speedup potential for larger datasets or higher spline degrees.
2502.07178
Online Aggregation of Trajectory Predictors
cs.RO
Trajectory prediction, the task of forecasting future agent behavior from past data, is central to safe and efficient autonomous driving. A diverse set of methods (e.g., rule-based or learned with different architectures and datasets) have been proposed, yet it is often the case that the performance of these methods is sensitive to the deployment environment (e.g., how well the design rules model the environment, or how accurately the test data match the training data). Building upon the principled theory of online convex optimization but also going beyond convexity and stationarity, we present a lightweight and model-agnostic method to aggregate different trajectory predictors online. We propose treating each individual trajectory predictor as an "expert" and maintaining a probability vector to mix the outputs of different experts. Then, the key technical approach lies in leveraging online data -- the true agent behavior to be revealed at the next timestep -- to form a convex-or-nonconvex, stationary-or-dynamic loss function whose gradient steers the probability vector towards choosing the best mixture of experts. We instantiate this method to aggregate trajectory predictors trained on different cities in the NUSCENES dataset and show that it performs just as well, if not better than, any singular model, even when deployed on the out-of-distribution LYFT dataset.
2502.07179
Improved YOLOv7 model for insulator defect detection
cs.CV cs.AI
Insulators are crucial insulation components and structural supports in power grids, playing a vital role in the transmission lines. Due to temperature fluctuations, internal stress, or damage from hail, insulators are prone to injury. Automatic detection of damaged insulators faces challenges such as diverse types, small defect targets, and complex backgrounds and shapes. Most research for detecting insulator defects has focused on a single defect type or a specific material. However, the insulators in the grid's transmission lines have different colors and materials. Various insulator defects coexist, and the existing methods have difficulty meeting the practical application requirements. Current methods suffer from low detection accuracy and mAP0.5 cannot meet application requirements. This paper proposes an improved YOLOv7 model for multi-type insulator defect detection. First, our model replaces the SPPCSPC module with the RFB module to enhance the network's feature extraction capability. Second, a CA mechanism is introduced into the head part to enhance the network's feature representation ability and to improve detection accuracy. Third, a WIoU loss function is employed to address the low-quality samples hindering model generalization during training, thereby improving the model's overall performance. The experimental results indicate that the proposed model exhibits enhancements across various performance metrics. Specifically, there is a 1.6% advancement in mAP_0.5, a corresponding 1.6% enhancement in mAP_0.5:0.95, a 1.3% elevation in precision, and a 1% increase in recall. Moreover, the model achieves parameter reduction by 3.2 million, leading to a decrease of 2.5 GFLOPS in computational cost. Notably, there is also an improvement of 2.81 milliseconds in single-image detection speed.
2502.07181
Tab2Visual: Overcoming Limited Data in Tabular Data Classification Using Deep Learning with Visual Representations
cs.LG cs.CV
This research addresses the challenge of limited data in tabular data classification, particularly prevalent in domains with constraints like healthcare. We propose Tab2Visual, a novel approach that transforms heterogeneous tabular data into visual representations, enabling the application of powerful deep learning models. Tab2Visual effectively addresses data scarcity by incorporating novel image augmentation techniques and facilitating transfer learning. We extensively evaluate the proposed approach on diverse tabular datasets, comparing its performance against a wide range of machine learning algorithms, including classical methods, tree-based ensembles, and state-of-the-art deep learning models specifically designed for tabular data. We also perform an in-depth analysis of factors influencing Tab2Visual's performance. Our experimental results demonstrate that Tab2Visual outperforms other methods in classification problems with limited tabular data.
2502.07182
Morphing Wing Designs in Commercial Aviation
eess.SY cs.SY
With increasing demands for fuel efficiency and operational adaptability in commercial aviation}, this paper provides a systematic review and classification of morphing wing technologies, analyzing their aerodynamic performance characteristics and atmospheric condition adaptability. We first develop a comprehensive classification framework for morphing wing designs based on their scale of morphing, actuation mechanisms, and intended purposes. Through analysis of historical developments and current implementations, we evaluate two significant case studies: the Mission Adaptive Compliant Wing (MACW) and Adaptive Aspect Ratio (AdAR) morphing wing, demonstrating performance improvements of up to 25% in drag reduction and 40% in control authority. Our investigation reveals critical trade-offs between full-span and partial morphing approaches, particularly regarding implementation complexity, certification requirements, and operational reliability. The study concludes with an assessment of technical barriers and opportunities, providing specific recommendations for advancing morphing wing technology in commercial aviation applications. Key findings indicate that while material science and control system advances enable practical implementation, certification pathways and maintenance considerations remain critical challenges for widespread adoption.
2502.07183
Space-Aware Instruction Tuning: Dataset and Benchmark for Guide Dog Robots Assisting the Visually Impaired
cs.RO cs.CV
Guide dog robots offer promising solutions to enhance mobility and safety for visually impaired individuals, addressing the limitations of traditional guide dogs, particularly in perceptual intelligence and communication. With the emergence of Vision-Language Models (VLMs), robots are now capable of generating natural language descriptions of their surroundings, aiding in safer decision-making. However, existing VLMs often struggle to accurately interpret and convey spatial relationships, which is crucial for navigation in complex environments such as street crossings. We introduce the Space-Aware Instruction Tuning (SAIT) dataset and the Space-Aware Benchmark (SA-Bench) to address the limitations of current VLMs in understanding physical environments. Our automated data generation pipeline focuses on the virtual path to the destination in 3D space and the surroundings, enhancing environmental comprehension and enabling VLMs to provide more accurate guidance to visually impaired individuals. We also propose an evaluation protocol to assess VLM effectiveness in delivering walking guidance. Comparative experiments demonstrate that our space-aware instruction-tuned model outperforms state-of-the-art algorithms. We have fully open-sourced the SAIT dataset and SA-Bench, along with the related code, at https://github.com/byungokhan/Space-awareVLM
2502.07184
Refine Knowledge of Large Language Models via Adaptive Contrastive Learning
cs.CL cs.AI
How to alleviate the hallucinations of Large Language Models (LLMs) has always been the fundamental goal pursued by the LLMs research community. Looking through numerous hallucination-related studies, a mainstream category of methods is to reduce hallucinations by optimizing the knowledge representation of LLMs to change their output. Considering that the core focus of these works is the knowledge acquired by models, and knowledge has long been a central theme in human societal progress, we believe that the process of models refining knowledge can greatly benefit from the way humans learn. In our work, by imitating the human learning process, we design an Adaptive Contrastive Learning strategy. Our method flexibly constructs different positive and negative samples for contrastive learning based on LLMs' actual mastery of knowledge. This strategy helps LLMs consolidate the correct knowledge they already possess, deepen their understanding of the correct knowledge they have encountered but not fully grasped, forget the incorrect knowledge they previously learned, and honestly acknowledge the knowledge they lack. Extensive experiments and detailed analyses on widely used datasets demonstrate the effectiveness of our method.
2502.07186
Perceived Confidence Scoring for Data Annotation with Zero-Shot LLMs
cs.CL cs.LG
Zero-shot LLMs are now also used for textual classification tasks, e.g., sentiment/emotion detection of a given input as a sentence/article. However, their performance can be suboptimal in such data annotation tasks. We introduce a novel technique Perceived Confidence Scoring (PCS) that evaluates LLM's confidence for its classification of an input by leveraging Metamorphic Relations (MRs). The MRs generate semantically equivalent yet textually mutated versions of the input. Following the principles of Metamorphic Testing (MT), the mutated versions are expected to have annotation labels similar to the input. By analyzing the consistency of LLM responses across these variations, PCS computes a confidence score based on the frequency of predicted labels. PCS can be used both for single LLM and multiple LLM settings (e.g., majority voting). We introduce an algorithm Perceived Differential Evolution (PDE) that determines the optimal weights assigned to the MRs and the LLMs for a classification task. Empirical evaluation shows PCS significantly improves zero-shot accuracy for Llama-3-8B-Instruct (4.96%) and Mistral-7B-Instruct-v0.3 (10.52%), with Gemma-2-9b-it showing a 9.39% gain. When combining all three models, PCS significantly outperforms majority voting by 7.75%.
2502.07187
Local Regularizers Are Not Transductive Learners
cs.LG stat.ML
We partly resolve an open question raised by Asilis et al. (COLT 2024): whether the algorithmic template of local regularization -- an intriguing generalization of explicit regularization, a.k.a. structural risk minimization -- suffices to learn all learnable multiclass problems. Specifically, we provide a negative answer to this question in the transductive model of learning. We exhibit a multiclass classification problem which is learnable in both the transductive and PAC models, yet cannot be learned transductively by any local regularizer. The corresponding hypothesis class, and our proof, are based on principles from cryptographic secret sharing. We outline challenges in extending our negative result to the PAC model, leaving open the tantalizing possibility of a PAC/transductive separation with respect to local regularization.
2502.07188
A Large-Scale Benchmark for Vietnamese Sentence Paraphrases
cs.CL
This paper presents ViSP, a high-quality Vietnamese dataset for sentence paraphrasing, consisting of 1.2M original-paraphrase pairs collected from various domains. The dataset was constructed using a hybrid approach that combines automatic paraphrase generation with manual evaluation to ensure high quality. We conducted experiments using methods such as back-translation, EDA, and baseline models like BART and T5, as well as large language models (LLMs), including GPT-4o, Gemini-1.5, Aya, Qwen-2.5, and Meta-Llama-3.1 variants. To the best of our knowledge, this is the first large-scale study on Vietnamese paraphrasing. We hope that our dataset and findings will serve as a valuable foundation for future research and applications in Vietnamese paraphrase tasks.
2502.07189
Exploring Neural Network Pruning with Screening Methods
cs.LG stat.ML
Deep neural networks (DNNs) such as convolutional neural networks (CNNs) for visual tasks, recurrent neural networks (RNNs) for sequence data, and transformer models for rich linguistic or multimodal tasks, achieved unprecedented performance on a wide range of tasks. The impressive performance of modern DNNs is partially attributed to their sheer scale. The latest deep learning models have tens to hundreds of millions of parameters which makes the inference processes resource-intensive. The high computational complexity of these networks prevents their deployment on resource-limited devices such as mobile platforms, IoT devices, and edge computing systems because these devices require energy-efficient and real-time processing capabilities. This paper proposes and evaluates a network pruning framework that eliminates non-essential parameters based on a statistical analysis of network component significance across classification categories. The proposed method uses screening methods coupled with a weighted scheme to assess connection and channel contributions for unstructured and structured pruning which allows for the elimination of unnecessary network elements without significantly degrading model performance. Extensive experimental validation on real-world vision datasets for both fully connected neural networks (FNNs) and CNNs has shown that the proposed framework produces competitive lean networks compared to the original networks. Moreover, the proposed framework outperforms state-of-art network pruning methods in two out of three cases.
2502.07190
Understanding LLMs' Fluid Intelligence Deficiency: An Analysis of the ARC Task
cs.AI
While LLMs have exhibited strong performance on various NLP tasks, it is noteworthy that most of these tasks rely on utilizing the vast amount of knowledge encoded in LLMs' parameters, rather than solving new problems without prior knowledge. In cognitive research, the latter ability is referred to as fluid intelligence, which is considered to be critical for assessing human intelligence. Recent research on fluid intelligence assessments has highlighted significant deficiencies in LLMs' abilities. In this paper, we analyze the challenges LLMs face in demonstrating fluid intelligence through controlled experiments, using the most representative ARC task as an example. Our study revealed three major limitations in existing LLMs: limited ability for skill composition, unfamiliarity with abstract input formats, and the intrinsic deficiency of left-to-right decoding. Our data and code can be found in https://wujunjie1998.github.io/araoc-benchmark.github.io/.
2502.07191
Bag of Tricks for Inference-time Computation of LLM Reasoning
cs.AI
With the advancement of large language models (LLMs), solving complex reasoning tasks has gained increasing attention. Inference-time computation methods (e.g., Best-of-N, beam search, et al.) are particularly valuable as they can enhance reasoning performance without modifying model parameters or requiring additional training. However, these techniques come with implementation challenges, and most existing methods remain at the proof-of-concept stage with limited practical adoption due to their computational complexity and varying effectiveness across different tasks. In this paper, we investigate and benchmark diverse inference-time computation strategies across reasoning tasks of varying complexity. Since most current methods rely on a proposer-verifier pipeline that first generates candidate solutions (e.g., reasoning solutions) and then selects the best one based on reward signals (e.g., RLHF rewards, process rewards), our research focuses on optimizing both candidate solution generation (e.g., instructing prompts, hyperparameters such as temperature and top-p) and reward mechanisms (e.g., self-evaluation, reward types). Through extensive experiments (more than 20,000 A100-80G GPU hours with over 1,000 experiments) across a variety of models (e.g., Llama, Qwen, and Mistral families) of various sizes, our ablation studies reveal that previously overlooked strategies can significantly enhance performance (e.g., tuning temperature can improve reasoning task performance by up to 5%). Furthermore, we establish a standardized benchmark for inference-time computation by systematically evaluating six representative methods across eight reasoning tasks. These findings provide a stronger foundation for future research. The code is available at https://github.com/usail-hkust/benchmark_inference_time_computation_LLM
2502.07192
OscNet: Machine Learning on CMOS Oscillator Networks
cs.CV
Machine learning and AI have achieved remarkable advancements but at the cost of significant computational resources and energy consumption. This has created an urgent need for a novel, energy-efficient computational fabric to replace the current computing pipeline. Recently, a promising approach has emerged by mimicking spiking neurons in the brain and leveraging oscillators on CMOS for direct computation. In this context, we propose a new and energy efficient machine learning framework implemented on CMOS Oscillator Networks (OscNet). We model the developmental processes of the prenatal brain's visual system using OscNet, updating weights based on the biologically inspired Hebbian rule. This same pipeline is then directly applied to standard machine learning tasks. OscNet is a specially designed hardware and is inherently energy-efficient. Its reliance on forward propagation alone for training further enhances its energy efficiency while maintaining biological plausibility. Simulation validates our designs of OscNet architectures. Experimental results demonstrate that Hebbian learning pipeline on OscNet achieves performance comparable to or even surpassing traditional machine learning algorithms, highlighting its potential as a energy efficient and effective computational paradigm.
2502.07193
Provably Efficient RLHF Pipeline: A Unified View from Contextual Bandits
cs.LG stat.ML
Reinforcement Learning from Human Feedback (RLHF) is a widely used approach for aligning Large Language Models (LLMs) with human preferences. While recent advancements have provided valuable insights into various stages and settings of RLHF, a comprehensive theoretical understanding of the entire RLHF pipeline remains lacking. Towards this end, we propose a unified framework for the RLHF pipeline from the view of contextual bandits and provide provable efficiency guarantees. In particular, we decompose the RLHF process into two distinct stages: (post-)training and deployment, exploring both passive and active data collection strategies during the training phase. By employing the Bradley-Terry preference model with a linearly parameterized reward function, we reformulate RLHF as a contextual preference bandit problem. We then develop novel algorithms for each stage, demonstrating significant improvements over existing approaches in both statistical and computational efficiency. Finally, we apply our method to train and deploy Llama-3-8B-Instruct on the Ultrafeedback-binarized dataset, and empirical results confirm the effectiveness of our approach.
2502.07194
Dense Object Detection Based on De-homogenized Queries
cs.CV cs.AI
Dense object detection is widely used in automatic driving, video surveillance, and other fields. This paper focuses on the challenging task of dense object detection. Currently, detection methods based on greedy algorithms, such as non-maximum suppression (NMS), often produce many repetitive predictions or missed detections in dense scenarios, which is a common problem faced by NMS-based algorithms. Through the end-to-end DETR (DEtection TRansformer), as a type of detector that can incorporate the post-processing de-duplication capability of NMS, etc., into the network, we found that homogeneous queries in the query-based detector lead to a reduction in the de-duplication capability of the network and the learning efficiency of the encoder, resulting in duplicate prediction and missed detection problems. To solve this problem, we propose learnable differentiated encoding to de-homogenize the queries, and at the same time, queries can communicate with each other via differentiated encoding information, replacing the previous self-attention among the queries. In addition, we used joint loss on the output of the encoder that considered both location and confidence prediction to give a higher-quality initialization for queries. Without cumbersome decoder stacking and guaranteeing accuracy, our proposed end-to-end detection framework was more concise and reduced the number of parameters by about 8% compared to deformable DETR. Our method achieved excellent results on the challenging CrowdHuman dataset with 93.6% average precision (AP), 39.2% MR-2, and 84.3% JI. The performance overperformed previous SOTA methods, such as Iter-E2EDet (Progressive End-to-End Object Detection) and MIP (One proposal, Multiple predictions). In addition, our method is more robust in various scenarios with different densities.
2502.07196
Parameter Optimization of Optical Six-Axis Force/Torque Sensor for Legged Robots
cs.RO
This paper introduces a novel six-axis force/torque sensor tailored for compact and lightweight legged robots. Unlike traditional strain gauge-based sensors, the proposed non-contact design employs photocouplers, enhancing resistance to physical impacts and reducing damage risk. This approach simplifies manufacturing, lowers costs, and meets the demands of legged robots by combining small size, light weight, and a wide force measurement range. A methodology for optimizing sensor parameters is also presented, focusing on maximizing sensitivity and minimizing error. Precise modeling and analysis of objective functions enabled the derivation of optimal design parameters. The sensor's performance was validated through extensive testing and integration into quadruped robots, demonstrating alignment with theoretical modeling. The sensor's precise measurement capabilities make it suitable for diverse robotic environments, particularly in analyzing interactions between robot feet and the ground. This innovation addresses existing sensor limitations while contributing to advancements in robotics and sensor technology, paving the way for future applications in robotic systems.
2502.07199
Fixed-Confidence Best Arm Identification with Decreasing Variance
cs.LG cs.IT math.IT math.ST stat.ML stat.TH
We focus on the problem of best-arm identification in a stochastic multi-arm bandit with temporally decreasing variances for the arms' rewards. We model arm rewards as Gaussian random variables with fixed means and variances that decrease with time. The cost incurred by the learner is modeled as a weighted sum of the time needed by the learner to identify the best arm, and the number of samples of arms collected by the learner before termination. Under this cost function, there is an incentive for the learner to not sample arms in all rounds, especially in the initial rounds. On the other hand, not sampling increases the termination time of the learner, which also increases cost. This trade-off necessitates new sampling strategies. We propose two policies. The first policy has an initial wait period with no sampling followed by continuous sampling. The second policy samples periodically and uses a weighted average of the rewards observed to identify the best arm. We provide analytical guarantees on the performance of both policies and supplement our theoretical results with simulations which show that our polices outperform the state-of-the-art policies for the classical best arm identification problem.
2502.07200
Color-Quality Invariance for Robust Medical Image Segmentation
eess.IV cs.CV
Single-source domain generalization (SDG) in medical image segmentation remains a significant challenge, particularly for images with varying color distributions and qualities. Previous approaches often struggle when models trained on high-quality images fail to generalize to low-quality test images due to these color and quality shifts. In this work, we propose two novel techniques to enhance generalization: dynamic color image normalization (DCIN) module and color-quality generalization (CQG) loss. The DCIN dynamically normalizes the color of test images using two reference image selection strategies. Specifically, the DCIN utilizes a global reference image selection (GRIS), which finds a universal reference image, and a local reference image selection (LRIS), which selects a semantically similar reference image per test sample. Additionally, CQG loss enforces invariance to color and quality variations by ensuring consistent segmentation predictions across transformed image pairs. Experimental results show that our proposals significantly improve segmentation performance over the baseline on two target domain datasets, despite being trained solely on a single source domain. Notably, our model achieved up to a 32.3-point increase in Dice score compared to the baseline, consistently producing robust and usable results even under substantial domain shifts. Our work contributes to the development of more robust medical image segmentation models that generalize across unseen domains. The implementation code is available at https://github.com/RaviShah1/DCIN-CQG.
2502.07202
Monte Carlo Tree Diffusion for System 2 Planning
cs.AI cs.LG
Diffusion models have recently emerged as a powerful tool for planning. However, unlike Monte Carlo Tree Search (MCTS)-whose performance naturally improves with additional test-time computation (TTC), standard diffusion-based planners offer only limited avenues for TTC scalability. In this paper, we introduce Monte Carlo Tree Diffusion (MCTD), a novel framework that integrates the generative strength of diffusion models with the adaptive search capabilities of MCTS. Our method reconceptualizes denoising as a tree-structured process, allowing partially denoised plans to be iteratively evaluated, pruned, and refined. By selectively expanding promising trajectories while retaining the flexibility to revisit and improve suboptimal branches, MCTD achieves the benefits of MCTS such as controlling exploration-exploitation trade-offs within the diffusion framework. Empirical results on challenging long-horizon tasks show that MCTD outperforms diffusion baselines, yielding higher-quality solutions as TTC increases.
2502.07203
Playmate: Flexible Control of Portrait Animation via 3D-Implicit Space Guided Diffusion
cs.CV
Recent diffusion-based talking face generation models have demonstrated impressive potential in synthesizing videos that accurately match a speech audio clip with a given reference identity. However, existing approaches still encounter significant challenges due to uncontrollable factors, such as inaccurate lip-sync, inappropriate head posture and the lack of fine-grained control over facial expressions. In order to introduce more face-guided conditions beyond speech audio clips, a novel two-stage training framework Playmate is proposed to generate more lifelike facial expressions and talking faces. In the first stage, we introduce a decoupled implicit 3D representation along with a meticulously designed motion-decoupled module to facilitate more accurate attribute disentanglement and generate expressive talking videos directly from audio cues. Then, in the second stage, we introduce an emotion-control module to encode emotion control information into the latent space, enabling fine-grained control over emotions and thereby achieving the ability to generate talking videos with desired emotion. Extensive experiments demonstrate that Playmate outperforms existing state-of-the-art methods in terms of video quality and lip-synchronization, and improves flexibility in controlling emotion and head pose. The code will be available at https://playmate111.github.io.
2502.07205
VINP: Variational Bayesian Inference with Neural Speech Prior for Joint ASR-Effective Speech Dereverberation and Blind RIR Identification
eess.AS cs.AI cs.LG
Reverberant speech, denoting the speech signal degraded by the process of reverberation, contains crucial knowledge of both anechoic source speech and room impulse response (RIR). This work proposes a variational Bayesian inference (VBI) framework with neural speech prior (VINP) for joint speech dereverberation and blind RIR identification. In VINP, a probabilistic signal model is constructed in the time-frequency (T-F) domain based on convolution transfer function (CTF) approximation. For the first time, we propose using an arbitrary discriminative dereverberation deep neural network (DNN) to predict the prior distribution of anechoic speech within a probabilistic model. By integrating both reverberant speech and the anechoic speech prior, VINP yields the maximum a posteriori (MAP) and maximum likelihood (ML) estimations of the anechoic speech spectrum and CTF filter, respectively. After simple transformations, the waveforms of anechoic speech and RIR are estimated. Moreover, VINP is effective for automatic speech recognition (ASR) systems, which sets it apart from most deep learning (DL)-based single-channel dereverberation approaches. Experiments on single-channel speech dereverberation demonstrate that VINP reaches an advanced level in most metrics related to human perception and displays unquestionable state-of-the-art (SOTA) performance in ASR-related metrics. For blind RIR identification, experiments indicate that VINP attains the SOTA level in blind estimation of reverberation time at 60 dB (RT60) and direct-to-reverberation ratio (DRR). Codes and audio samples are available online.
2502.07207
A Study on the Importance of Features in Detecting Advanced Persistent Threats Using Machine Learning
cs.CR cs.AI cs.LG
Advanced Persistent Threats (APTs) pose a significant security risk to organizations and industries. These attacks often lead to severe data breaches and compromise the system for a long time. Mitigating these sophisticated attacks is highly challenging due to the stealthy and persistent nature of APTs. Machine learning models are often employed to tackle this challenge by bringing automation and scalability to APT detection. Nevertheless, these intelligent methods are data-driven, and thus, highly affected by the quality and relevance of input data. This paper aims to analyze measurements considered when recording network traffic and conclude which features contribute more to detecting APT samples. To do this, we study the features associated with various APT cases and determine their importance using a machine learning framework. To ensure the generalization of our findings, several feature selection techniques are employed and paired with different classifiers to evaluate their effectiveness. Our findings provide insights into how APT detection can be enhanced in real-world scenarios.
2502.07209
Enhancing Physics-Informed Neural Networks Through Feature Engineering
cs.LG
Physics-Informed Neural Networks (PINNs) seek to solve partial differential equations (PDEs) with deep learning. Mainstream approaches that deploy fully-connected multi-layer deep learning architectures require prolonged training to achieve even moderate accuracy, while recent work on feature engineering allows higher accuracy and faster convergence. This paper introduces SAFE-NET, a Single-layered Adaptive Feature Engineering NETwork that achieves orders-of-magnitude lower errors with far fewer parameters than baseline feature engineering methods. SAFE-NET returns to basic ideas in machine learning, using Fourier features, a simplified single hidden layer network architecture, and an effective optimizer that improves the conditioning of the PINN optimization problem. Numerical results show that SAFE-NET converges faster and typically outperforms deeper networks and more complex architectures. It consistently uses fewer parameters -- on average, 65% fewer than the competing feature engineering methods -- while achieving comparable accuracy in less than 30% of the training epochs. Moreover, each SAFE-NET epoch is 95% faster than those of competing feature engineering approaches. These findings challenge the prevailing belief that modern PINNs effectively learn features in these scientific applications and highlight the efficiency gains possible through feature engineering.
2502.07211
Improve the Training Efficiency of DRL for Wireless Communication Resource Allocation: The Role of Generative Diffusion Models
cs.LG
Dynamic resource allocation in mobile wireless networks involves complex, time-varying optimization problems, motivating the adoption of deep reinforcement learning (DRL). However, most existing works rely on pre-trained policies, overlooking dynamic environmental changes that rapidly invalidate the policies. Periodic retraining becomes inevitable but incurs prohibitive computational costs and energy consumption-critical concerns for resource-constrained wireless systems. We identify three root causes of inefficient retraining: high-dimensional state spaces, suboptimal action spaces exploration-exploitation trade-offs, and reward design limitations. To overcome these limitations, we propose Diffusion-based Deep Reinforcement Learning (D2RL), which leverages generative diffusion models (GDMs) to holistically enhance all three DRL components. Iterative refinement process and distribution modelling of GDMs enable (1) the generation of diverse state samples to improve environmental understanding, (2) balanced action space exploration to escape local optima, and (3) the design of discriminative reward functions that better evaluate action quality. Our framework operates in two modes: Mode I leverages GDMs to explore reward spaces and design discriminative reward functions that rigorously evaluate action quality, while Mode II synthesizes diverse state samples to enhance environmental understanding and generalization. Extensive experiments demonstrate that D2RL achieves faster convergence and reduced computational costs over conventional DRL methods for resource allocation in wireless communications while maintaining competitive policy performance. This work underscores the transformative potential of GDMs in overcoming fundamental DRL training bottlenecks for wireless networks, paving the way for practical, real-time deployments.
2502.07213
Evaluation for Regression Analyses on Evolving Data Streams
cs.LG cs.AI
The paper explores the challenges of regression analysis in evolving data streams, an area that remains relatively underexplored compared to classification. We propose a standardized evaluation process for regression and prediction interval tasks in streaming contexts. Additionally, we introduce an innovative drift simulation strategy capable of synthesizing various drift types, including the less-studied incremental drift. Comprehensive experiments with state-of-the-art methods, conducted under the proposed process, validate the effectiveness and robustness of our approach.
2502.07214
Pareto Optimal Algorithmic Recourse in Multi-cost Function
cs.LG cs.AI cs.DS
In decision-making systems, algorithmic recourse aims to identify minimal-cost actions to alter an individual features, thereby obtaining a desired outcome. This empowers individuals to understand, question, or alter decisions that negatively affect them. However, due to the variety and sensitivity of system environments and individual personalities, quantifying the cost of a single function is nearly impossible while considering multiple criteria situations. Most current recourse mechanisms use gradient-based methods that assume cost functions are differentiable, often not applicable in real-world scenarios, resulting in sub-optimal solutions that compromise various criteria. These solutions are typically intractable and lack rigorous theoretical foundations, raising concerns regarding interpretability, reliability, and transparency from the explainable AI (XAI) perspective. To address these issues, this work proposes an algorithmic recourse framework that handles non-differentiable and discrete multi-cost functions. By formulating recourse as a multi-objective optimization problem and assigning weights to different criteria based on their importance, our method identifies Pareto optimal recourse recommendations. To demonstrate scalability, we incorporate the concept of epsilon-net, proving the ability to find approximated Pareto optimal actions. Experiments show the trade-off between different criteria and the methods scalability in large graphs. Compared to current heuristic practices, our approach provides a stronger theoretical foundation and better aligns recourse suggestions with real-world requirements.
2502.07215
PDV: Prompt Directional Vectors for Zero-shot Composed Image Retrieval
cs.CV
Zero-shot composed image retrieval (ZS-CIR) enables image search using a reference image and text prompt without requiring specialized text-image composition networks trained on large-scale paired data. However, current ZS-CIR approaches face three critical limitations in their reliance on composed text embeddings: static query embedding representations, insufficient utilization of image embeddings, and suboptimal performance when fusing text and image embeddings. To address these challenges, we introduce the Prompt Directional Vector (PDV), a simple yet effective training-free enhancement that captures semantic modifications induced by user prompts. PDV enables three key improvements: (1) dynamic composed text embeddings where prompt adjustments are controllable via a scaling factor, (2) composed image embeddings through semantic transfer from text prompts to image features, and (3) weighted fusion of composed text and image embeddings that enhances retrieval by balancing visual and semantic similarity. Our approach serves as a plug-and-play enhancement for existing ZS-CIR methods with minimal computational overhead. Extensive experiments across multiple benchmarks demonstrate that PDV consistently improves retrieval performance when integrated with state-of-the-art ZS-CIR approaches, particularly for methods that generate accurate compositional embeddings. The code will be publicly available.
2502.07216
SparseFormer: Detecting Objects in HRW Shots via Sparse Vision Transformer
cs.CV cs.AI
Recent years have seen an increase in the use of gigapixel-level image and video capture systems and benchmarks with high-resolution wide (HRW) shots. However, unlike close-up shots in the MS COCO dataset, the higher resolution and wider field of view raise unique challenges, such as extreme sparsity and huge scale changes, causing existing close-up detectors inaccuracy and inefficiency. In this paper, we present a novel model-agnostic sparse vision transformer, dubbed SparseFormer, to bridge the gap of object detection between close-up and HRW shots. The proposed SparseFormer selectively uses attentive tokens to scrutinize the sparsely distributed windows that may contain objects. In this way, it can jointly explore global and local attention by fusing coarse- and fine-grained features to handle huge scale changes. SparseFormer also benefits from a novel Cross-slice non-maximum suppression (C-NMS) algorithm to precisely localize objects from noisy windows and a simple yet effective multi-scale strategy to improve accuracy. Extensive experiments on two HRW benchmarks, PANDA and DOTA-v1.0, demonstrate that the proposed SparseFormer significantly improves detection accuracy (up to 5.8%) and speed (up to 3x) over the state-of-the-art approaches.
2502.07218
LUNAR: LLM Unlearning via Neural Activation Redirection
cs.LG cs.AI
Large Language Models (LLMs) benefit from training on ever larger amounts of textual data, but as a result, they increasingly incur the risk of leaking private information. The ability to selectively remove knowledge from LLMs is, therefore, a highly desirable capability. In this paper, we propose LUNAR, a novel unlearning methodology grounded in the Linear Representation Hypothesis. LUNAR operates by redirecting the representations of unlearned data to regions that trigger the model's inherent ability to express its inability to answer. LUNAR achieves state-of-the-art unlearning performance while significantly enhancing the controllability of the unlearned model during inference. Specifically, LUNAR achieves between 2.9x to 11.7x improvements on combined "unlearning efficacy" and "model utility" score ("Deviation Score") on the PISTOL dataset across various base models. We also demonstrate, through quantitative analysis and qualitative examples, LUNAR's superior controllability in generating coherent and contextually aware responses, mitigating undesired side effects of existing methods. Moreover, we demonstrate that LUNAR is robust against white-box adversarial attacks and versatile in handling real-world scenarios, such as processing sequential unlearning requests.
2502.07219
DOGR: Leveraging Document-Oriented Contrastive Learning in Generative Retrieval
cs.IR
Generative retrieval constitutes an innovative approach in information retrieval, leveraging generative language models (LM) to generate a ranked list of document identifiers (docid) for a given query. It simplifies the retrieval pipeline by replacing the large external index with model parameters. However, existing works merely learned the relationship between queries and document identifiers, which is unable to directly represent the relevance between queries and documents. To address the above problem, we propose a novel and general generative retrieval framework, namely Leveraging Document-Oriented Contrastive Learning in Generative Retrieval (DOGR), which leverages contrastive learning to improve generative retrieval tasks. It adopts a two-stage learning strategy that captures the relationship between queries and documents comprehensively through direct interactions. Furthermore, negative sampling methods and corresponding contrastive learning objectives are implemented to enhance the learning of semantic representations, thereby promoting a thorough comprehension of the relationship between queries and documents. Experimental results demonstrate that DOGR achieves state-of-the-art performance compared to existing generative retrieval methods on two public benchmark datasets. Further experiments have shown that our framework is generally effective for common identifier construction techniques.
2502.07221
MLLM4PUE: Toward Universal Embeddings in Computational Pathology through Multimodal LLMs
cs.CV
Pathology plays a critical role in diagnosing a wide range of diseases, yet existing approaches often rely heavily on task-specific models trained on extensive, well-labeled datasets. These methods face sustainability challenges due to the diversity of pathologies and the labor-intensive nature of data collection. To address these limitations, we highlight the need for universal multimodal embeddings that can support multiple downstream tasks. Previous approaches often involve fine-tuning CLIP-based models, which handle images and text separately, limiting their ability to capture complex multimodal relationships. Additionally, these models are evaluated across diverse datasets without a unified benchmark for assessing multimodal embeddings in pathology. To address these challenges, we propose MLLM4PUE, a novel framework that leverages Multimodal Large Language Models (MLLMs) to generate Pathology Universal Embeddings. The MLLM4PUE framework not only facilitates robust integration of images and text but also enhances understanding and fusion capabilities across various tasks. We further introduce the Pathology Multimodal Embedding Benchmark (PMEB), a comprehensive benchmark designed to assess the quality of pathology multimodal embeddings. PMEB comprises 15 original tasks drawn from 14 datasets, organized into three meta-tasks: retrieval, classification, and composed retrieval. Experimental results demonstrate the superiority of MLLM4PUE, illustrating MLLM-based models can effectively support a wide range of downstream tasks and unify the research direction for foundation models in pathology.
2502.07222
A Memory Efficient Randomized Subspace Optimization Method for Training Large Language Models
cs.LG
The memory challenges associated with training Large Language Models (LLMs) have become a critical concern, particularly when using the Adam optimizer. To address this issue, numerous memory-efficient techniques have been proposed, with GaLore standing out as a notable example designed to reduce the memory footprint of optimizer states. However, these approaches do not alleviate the memory burden imposed by activations, rendering them unsuitable for scenarios involving long context sequences or large mini-batches. Moreover, their convergence properties are still not well-understood in the literature. In this work, we introduce a Randomized Subspace Optimization framework for pre-training and fine-tuning LLMs. Our approach decomposes the high-dimensional training problem into a series of lower-dimensional subproblems. At each iteration, a random subspace is selected, and the parameters within that subspace are optimized. This structured reduction in dimensionality allows our method to simultaneously reduce memory usage for both activations and optimizer states. We establish comprehensive convergence guarantees and derive rates for various scenarios, accommodating different optimization strategies to solve the subproblems. Extensive experiments validate the superior memory and communication efficiency of our method, achieving performance comparable to GaLore and Adam.
2502.07223
Graph RAG-Tool Fusion
cs.CL
Recent developments in retrieval-augmented generation (RAG) for selecting relevant tools from a tool knowledge base enable LLM agents to scale their complex tool calling capabilities to hundreds or thousands of external tools, APIs, or agents-as-tools. However, traditional RAG-based tool retrieval fails to capture structured dependencies between tools, limiting the retrieval accuracy of a retrieved tool's dependencies. For example, among a vector database of tools, a "get stock price" API requires a "stock ticker" parameter from a "get stock ticker" API, and both depend on OS-level internet connectivity tools. In this paper, we address this limitation by introducing Graph RAG-Tool Fusion, a novel plug-and-play approach that combines the strengths of vector-based retrieval with efficient graph traversal to capture all relevant tools (nodes) along with any nested dependencies (edges) within the predefined tool knowledge graph. We also present ToolLinkOS, a new tool selection benchmark of 573 fictional tools, spanning over 15 industries, each with an average of 6.3 tool dependencies. We demonstrate that Graph RAG-Tool Fusion achieves absolute improvements of 71.7% and 22.1% over na\"ive RAG on ToolLinkOS and ToolSandbox benchmarks, respectively (mAP@10). ToolLinkOS dataset is available at https://github.com/EliasLumer/Graph-RAG-Tool-Fusion-ToolLinkOS
2502.07225
CAT: Contrastive Adversarial Training for Evaluating the Robustness of Protective Perturbations in Latent Diffusion Models
cs.CV
Latent diffusion models have recently demonstrated superior capabilities in many downstream image synthesis tasks. However, customization of latent diffusion models using unauthorized data can severely compromise the privacy and intellectual property rights of data owners. Adversarial examples as protective perturbations have been developed to defend against unauthorized data usage by introducing imperceptible noise to customization samples, preventing diffusion models from effectively learning them. In this paper, we first reveal that the primary reason adversarial examples are effective as protective perturbations in latent diffusion models is the distortion of their latent representations, as demonstrated through qualitative and quantitative experiments. We then propose the Contrastive Adversarial Training (CAT) utilizing adapters as an adaptive attack against these protection methods, highlighting their lack of robustness. Extensive experiments demonstrate that our CAT method significantly reduces the effectiveness of protective perturbations in customization configurations, urging the community to reconsider and enhance the robustness of existing protective perturbation methods. Code is available at \hyperlink{here}{https://github.com/senp98/CAT}.
2502.07226
Non-Iterative Coordination of Interconnected Power Grids via Dimension-Decomposition-Based Flexibility Aggregation
eess.SY cs.SY
The bulk power grid is divided into regional grids interconnected with multiple tie-lines for efficient operation. Since interconnected power grids are operated by different control centers, it is a challenging task to realize coordinated dispatch of multiple regional grids. A viable solution is to compute a flexibility aggregation model for each regional power grid, then optimize the tie-line schedule using the aggregated models to implement non-iterative coordinated dispatch. However, challenges such as intricate interdependencies and curse of dimensionality persist in computing the aggregated models in high-dimensional space. Existing methods like Fourier-Motzkin elimination, vertex search, and multi-parameter programming are limited by dimensionality and conservatism, hindering practical application. This paper presents a novel dimension-decomposition-based flexibility aggregation algorithm for calculating the aggregated models of multiple regional power grids, enabling non-iterative coordination in large-scale interconnected systems. Compared to existing methods, the proposed approach yields a significantly less conservative flexibility region. The derived flexibility aggregation model for each regional power grid has a well-defined physical counterpart, which facilitates intuitive analysis of multi-port regional power grids and provides valuable insights into their internal resource endowments. Numerical tests validate the feasibility of the aggregated model and demonstrate its accuracy in coordinating interconnected power grids.
2502.07230
Physics-Informed Recurrent Network for Gas Pipeline Network Parameters Identification
eess.SY cs.SY
As a part of the integrated energy system (IES), gas pipeline networks can provide additional flexibility to power systems through coordinated optimal dispatch. An accurate pipeline network model is critical for the optimal operation and control of IESs. However, inaccuracies or unavailability of accurate pipeline parameters often introduce errors in the mathematical models of such networks. This paper proposes a physics-informed recurrent network (PIRN) model to identify the state-space model of gas pipelines. The approach combines data-driven learning from measurement data with the fluid dynamics described by partial differential equations. By embedding the physical state-space model within the recurrent network, parameter identification is transformed into a training process for a PIRN. Similar to standard recurrent neural networks, this model can be implemented using the PyTorch framework and trained via backpropagation. Case studies demonstrate that our method accurately estimates gas pipeline models from sparse terminal node measurements, providing robust performance and significantly higher parameter efficiency. Furthermore, the identified models can be seamlessly integrated into optimization frameworks.
2502.07232
Simplifying Adversarially Robust PAC Learning with Tolerance
cs.LG
Adversarially robust PAC learning has proved to be challenging, with the currently best known learners [Montasser et al., 2021a] relying on improper methods based on intricate compression schemes, resulting in sample complexity exponential in the VC-dimension. A series of follow up work considered a slightly relaxed version of the problem called adversarially robust learning with tolerance [Ashtiani et al., 2023, Bhattacharjee et al., 2023, Raman et al., 2024] and achieved better sample complexity in terms of the VC-dimension. However, those algorithms were either improper and complex, or required additional assumptions on the hypothesis class H. We prove, for the first time, the existence of a simpler learner that achieves a sample complexity linear in the VC-dimension without requiring additional assumptions on H. Even though our learner is improper, it is "almost proper" in the sense that it outputs a hypothesis that is "similar" to a hypothesis in H. We also use the ideas from our algorithm to construct a semi-supervised learner in the tolerant setting. This simple algorithm achieves comparable bounds to the previous (non-tolerant) semi-supervised algorithm of Attias et al. [2022a], but avoids the use of intricate subroutines from previous works, and is "almost proper."