article stringlengths 507 295k | abstract stringlengths 417 1.92k | category listlengths 1 6 |
|---|---|---|
# 1 Introduction
Modern Artificial Neural Networks (ANNs) achieve remarkable results in recognizing patterns. However, due to their complexity and black-box character, their failures are hard to identify [13] which limits their use in safety-critical environments. Additionally, certain common training schemes encourag... | Understanding the decision-making and trusting the reliability of Deep
Machine Learning Models is crucial for adopting such methods to safety-relevant
applications. We extend self-explainable Prototypical Variational models with
autoencoder-based out-of-distribution (OOD) detection: A Variational
Autoencoder is applied... | [
"cs.LG",
"cs.CV"
] |
# 1 INTRODUCTION
Facial micro-expression recognition (MER) is a popular task in the fields of computer vision and affective computing [1]. It has applications in wide areas such as medicine, education, and criminal investigation. Micro-expressions (MEs) are subtle and involuntary that convey genuine emotions [2], and ... | Facial micro-expression recognition (MER) is a challenging problem, due to
transient and subtle micro-expression (ME) actions. Most existing methods
depend on hand-crafted features, key frames like onset, apex, and offset
frames, or deep networks limited by small-scale and low-diversity datasets. In
this paper, we prop... | [
"cs.CV"
] |
# I. INTRODUCTION
One of the primary goals of neuromorphic computing is to emulate the structure and dynamics of biological neuronal networks, achieving both brain-like energy efficiency and high computational accuracy. This is accomplished through the use of spiking neuron models implemented on neuromorphic chips. Ov... | Spiking Neural Networks (SNNs) are computational models inspired by the
structure and dynamics of biological neuronal networks. Their event-driven
nature enables them to achieve high energy efficiency, particularly when
deployed on neuromorphic hardware platforms. Unlike conventional Artificial
Neural Networks (ANNs), ... | [
"cs.NE",
"cs.AI"
] |
# 1 Introduction
Rapid development of LLM has revolutionized the way AI and humans interact. In particular, the development of GPT (Brown et al., 2020) and introduction of ChatGPT has provided a major turning point in the field of natural language processing, spawning a new specialization called ’prompt engineering’. ... | Since the advent of large language models, prompt engineering now enables the
rapid, low-effort creation of diverse autonomous agents that are already in
widespread use. Yet this convenience raises urgent concerns about the safety,
robustness, and behavioral consistency of the underlying prompts, along with
the pressin... | [
"cs.AI",
"cs.CR"
] |
# 1 Introduction
Generative models for visual content have achieved remarkable advancements and have been applied to various fields, including amateur entertainment and professional creation. However, several challenges persist, such as the model could generate outputs that conflict with human values, harmful content,... | An effective reward model plays a pivotal role in reinforcement learning for
post-training enhancement of visual generative models. However, current
approaches of reward modeling suffer from implementation complexity due to
their reliance on extensive human-annotated preference data or meticulously
engineered quality d... | [
"cs.CV",
"cs.AI",
"cs.LG"
] |
# 1 Introduction
The challenge of offensive language is a continually growing concern in the field of natural language processing (NLP). The extensive use of online social communities has created opportunities for heated discussions, which can quickly escalate to offensive or toxic levels. To minimize harmful social i... | Language detoxification involves removing toxicity from offensive language.
While a neutral-toxic paired dataset provides a straightforward approach for
training detoxification models, creating such datasets presents several
challenges: i) the need for human annotation to build paired data, and ii) the
rapid evolution ... | [
"cs.CL"
] |
# 1 Introduction
Motivation. Approximate Nearest Neighbor Search (ANNS) for high-dimensional vector databases [32, 89] is heavily used for semantic search in multiple application areas [31], including web search engines [17, 19], multimedia databases [35, 80], recommendation systems [20, 23, 77], image retrieval [95, ... | Approximate Nearest Neighbor Search (ANNS) presents an inherent tradeoff
between performance and recall (i.e., result quality). Each ANNS algorithm
provides its own algorithm-dependent parameters to allow applications to
influence the recall/performance tradeoff of their searches. This situation is
doubly problematic. ... | [
"cs.DB"
] |
# 1. Introduction
Text-to-3D generation—the task of creating 3D contents from natural language descriptions—has attracted enormous interest [21, 41, 48], due to its broad applications in vision and graphics. Recent advances, such as 3D representations[16, 37], large-scale pre-trained visionlanguage models[42], advance... | Distilling pre-trained 2D diffusion models into 3D assets has driven
remarkable advances in text-to-3D synthesis. However, existing methods
typically rely on Score Distillation Sampling (SDS) loss, which involves
asymmetric KL divergence--a formulation that inherently favors mode-seeking
behavior and limits generation ... | [
"cs.CV"
] |
# 1 Introduction
Negation is a fundamental and universal phenomenon found in languages worldwide. It is closely linked to various human communicative abilities, including denial, contradiction, deception, misrepresentation, and irony. Although affirmative statements are more common, negation is still prevalent in lang... | Negation is a fundamental linguistic phenomenon that poses persistent
challenges for Large Language Models (LLMs), particularly in tasks requiring
deep semantic understanding. Existing benchmarks often treat negation as a side
case within broader tasks like natural language inference, resulting in a lack
of benchmarks ... | [
"cs.CL"
] |
# 1 Introduction
Big data is now a key focus for both government and business leaders.[3]. However, buried within this immense data deluge lies an abundance of untapped potential and valuable insights, which has given rise to an innovative scientific paradigm known as dataintensive scientific discovery[4]. Researchers... | Large Language Models (LLMs) have shown remarkable proficiency in natural
language understanding (NLU), opening doors for innovative applications. We
introduce StreamLink - an LLM-driven distributed data system designed to
improve the efficiency and accessibility of data engineering tasks. We build
StreamLink on top of... | [
"cs.DB",
"cs.AI"
] |
# I. INTRODUCTION
Code-switching—switching between languages within the same conversation—is a common and natural way of speaking in many multilingual communities. This is especially true in Southeast Asia, where people often mix their native language with English in everyday conversations [1]. However, this kind of s... | Developing code-switched ASR systems is challenging due to language ambiguity
and limited exposure to multilingual, code-switched data, while collecting such
speech is costly. Prior work generates synthetic audio from text, but these
methods are computationally intensive and hard to scale. We introduce
AsyncSwitch, a n... | [
"cs.CL",
"cs.SD",
"eess.AS"
] |
# 1 Introduction
Large Language Models (LLMs) are rapidly evolving and demonstrating increasing capabilities in coding, fundamentally transforming the software development ecosystem. Recent LLMs such as ChatGPT [55] and Claude [4] exhibit remarkable code generation performance, producing highquality outputs in respons... | Large language models (LLMs) have become integral to modern software
development, producing vast amounts of AI-generated source code. While these
models boost programming productivity, their misuse introduces critical risks,
including code plagiarism, license violations, and the propagation of insecure
programs. As a r... | [
"cs.SE",
"cs.CL",
"cs.CY",
"cs.LG"
] |
# 1 Introduction
Many popular programming languages, including C#, Java, and Python, support exceptions [15, 17, 37]. Exceptions are thrown during program execution if an unwanted event happens, e.g., a method is invoked with an illegal argument value. Software developers write exceptional behavior tests (EBTs) to che... | Exceptional behavior tests (EBTs) are crucial in software development for
verifying that code correctly handles unwanted events and throws appropriate
exceptions. However, prior research has shown that developers often prioritize
testing "happy paths", e.g., paths without unwanted events over exceptional
scenarios. We ... | [
"cs.SE",
"cs.AI"
] |
# 1. Introduction
End-to-end (E2E) automatic speech recognition (ASR) [1, 2] has made significant strides in recent years, achieving remarkable performance on various benchmarks [3–5]. However, the challenge of recognizing overlapping speech in multi-talker scenarios remains a critical area of research. Traditional AS... | We extend the frameworks of Serialized Output Training (SOT) to address
practical needs of both streaming and offline automatic speech recognition
(ASR) applications. Our approach focuses on balancing latency and accuracy,
catering to real-time captioning and summarization requirements. We propose
several key improveme... | [
"eess.AS",
"cs.CL",
"cs.SD"
] |
# 1 Introduction
3D reconstruction from multi-view images is a cornerstone task in computer vision. Traditionally, this process has been achieved by assembling classical techniques such as keypoint detection $[ 1 -$ 3] and matching [4, 5], robust camera estimation [4, 6], Structure-from-Motion(SfM), Bundle Adjustment(... | Dense matching methods like DUSt3R regress pairwise pointmaps for 3D
reconstruction. However, the reliance on pairwise prediction and the limited
generalization capability inherently restrict the global geometric consistency.
In this work, we introduce Test3R, a surprisingly simple test-time learning
technique that sig... | [
"cs.CV"
] |
# 1 Introduction
As large language models (LLMs) rapidly evolve and are deployed across critical applications, there is a pressing need for reliable safety evaluation methods that can keep pace with new models and adversarial attacks, and uncover failure modes before harm occurs. One common paradigm is dynamic safety ... | Large language models (LLMs) are rapidly deployed in critical applications,
raising urgent needs for robust safety benchmarking. We propose Jailbreak
Distillation (JBDistill), a novel benchmark construction framework that
"distills" jailbreak attacks into high-quality and easily-updatable safety
benchmarks. JBDistill u... | [
"cs.CL",
"cs.CR",
"cs.SE"
] |
# 1. Introduction
Egocentric human-object interaction (Ego-HOI) detection aims to locate interacting human-object pairs and reason about their interaction relationships from the first-person vision. As a crucial task in human-centered advanced scene understanding, its precise outcomes can drive advancements in a wide ... | Understanding the interaction between humans and objects has gained much
attention in recent years. Existing human-object interaction (HOI) detection
methods mainly focus on the third-person perspectives, overlooking a more
intuitive way from the egocentric view of HOI, namely Ego-HOI. This paper
introduces an Ego-HOIB... | [
"cs.CV"
] |
# 1 Introduction
Open Source Software (OSS) is extensively used across various sectors and plays a critical role in powering modern technological systems, from critical infrastructure to innovative applications. Many foundational software systems, such as operating systems and database management systems, are written ... | Project building is pivotal to support various program analysis tasks, such
as generating intermediate rep- resentation code for static analysis and
preparing binary code for vulnerability reproduction. However, automating the
building process for C/C++ projects is a highly complex endeavor, involving
tremendous techni... | [
"cs.SE"
] |
# I. INTRODUCTION
poral boundaries of individual speakers within an audio stream and assign appropriate speaker identities, addresses the fundamental question of “who spoke when” [1]. It serves as a foundational component in numerous downstream speechrelated tasks, including automatic meeting summarization, conversati... | In this paper, we propose a novel neural speaker diarization system using
memory-aware multi-speaker embedding with sequence-to-sequence architecture
(NSD-MS2S), which integrates a memory-aware multi-speaker embedding module with
a sequence-to-sequence architecture. The system leverages a memory module to
enhance speak... | [
"cs.SD",
"cs.AI"
] |
# 1 Introduction
Recent advancements in generative AI have profoundly impacted autonomous driving, with diffusion models (DMs) emerging as pivotal tools for data synthesis and driving simulation. Some approaches utilize DMs as data machines, producing high-fidelity driving videos [1–14] or multi-modal synthetic data [... | Diffusion models are advancing autonomous driving by enabling realistic data
synthesis, predictive end-to-end planning, and closed-loop simulation, with a
primary focus on temporally consistent generation. However, the generation of
large-scale 3D scenes that require spatial coherence remains underexplored. In
this pap... | [
"cs.CV"
] |
# 1 INTRODUCTION
Index tuning aims to find the optimal index configuration (i.e., a set of indexes) for an input workload of SQL queries. It is often a time-consuming and resource-intensive process for large and complex workloads in practice. From user’s perspective, it is therefore desirable to constrain the index tu... | Index tuning aims to find the optimal index configuration for an input
workload. It is often a time-consuming and resource-intensive process, largely
attributed to the huge amount of "what-if" calls made to the query optimizer
during configuration enumeration. Therefore, in practice it is desirable to set
a budget cons... | [
"cs.DB"
] |
# 1 Introduction
The Lecture Video Visual Objects (LVVO) Dataset is designed as a benchmark for object detection in lecture video frames. It includes bounding box annotations for four visual categories: Table, Chart-Graph, Photographic-Image, and Visual-Illustration.
The dataset consists of 4,000 images (video frames... | We introduce the Lecture Video Visual Objects (LVVO) dataset, a new benchmark
for visual object detection in educational video content. The dataset consists
of 4,000 frames extracted from 245 lecture videos spanning biology, computer
science, and geosciences. A subset of 1,000 frames, referred to as LVVO_1k, has
been m... | [
"cs.CV",
"cs.LG"
] |
# 1. Introduction
Open Science practices are fostered by institutions and research funders as a way to make research more collaborative, transparent and closer to society. Among these practices we find the effort to make research data useful for reuse. To achieve this goal the FAIR principles were developed (Wilkinson... | Inspired by a proposal made almost ten years ago, this paper presents a model
for classifying per-sonal data for research to inform researchers on how to
manage them. The classification is based on the principles of the European
General Data Protection Regulation and its implementation under the Spanish
Law. The paper ... | [
"cs.CY",
"cs.DB"
] |
# I. INTRODUCTION
Requirements elicitation and specification is a continuous and fundamental activity in software development [1]. Despite its importance, the process is challenging, manual, and laborintensive. A significant barrier is tacit knowledge – information held by a stakeholder but not explicitly shared with ... | Requirements elicitation and specification remains a labor-intensive, manual
process prone to inconsistencies and gaps, presenting a significant challenge
in modern software engineering. Emerging studies underscore the potential of
employing large language models (LLMs) for automated requirements generation to
support ... | [
"cs.SE",
"cs.AI",
"cs.LG"
] |
# 1 Introduction
LLMs are now widely applied in the AI for materials science field for applications such as literature and materials database knowledge extraction [1–3], materials property prediction [4–6], alloy design [7, 8], discovering new physical laws [9, 10], and proposing new scientific hypotheses [11, 12].
R... | Large language models (LLMs) are increasingly applied to materials science
questions, including literature comprehension, property prediction, materials
discovery and alloy design. At the same time, a wide range of physics-based
computational approaches have been developed in which materials properties can
be calculate... | [
"cond-mat.mtrl-sci",
"cs.CL",
"cs.DB"
] |
# 1 Introduction
Large Language Models (LLMs) have unequivocally revolutionized the landscape of automated code generation. Models like GPT-4o [33], Google Gemini [15], Anthropic Claude [1], and GitHub Copilot [14] excel at generating functional code snippets and translating between languages. These capabilities are n... | Large language models (LLMs) have demonstrated remarkable progress in code
generation, but many existing benchmarks are approaching saturation and offer
little guarantee on the trustworthiness of the generated programs, offering
limited insight into deeper reasoning capabilities. We introduce
VerifyThisBench, a new ben... | [
"cs.SE"
] |
# 1 Introduction
In the era of computers and digital interactions, individuals are increasingly exposed to risks they do not anticipate or desire. Social media, instant messaging applications, and chatbots serve as digital platforms that enable individuals to interact without being physically present or revealing thei... | In the era of digitalization, as individuals increasingly rely on digital
platforms for communication and news consumption, various actors employ
linguistic strategies to influence public perception. While models have become
proficient at detecting explicit patterns, which typically appear in texts as
single remarks re... | [
"cs.CL"
] |
# 1. Introduction
In recent years, Large Language Models (LLMs) have emerged as transformative tools across a wide range of natural language processing (NLP) applications, including machine translation, question answering, summarization, and dialogue systems [1, 2, 3]. Their ability to model longrange dependencies and... | This paper presents our system for the MLC-SLM Challenge 2025, focusing on
multilingual speech recognition and language modeling with large language
models (LLMs). Our approach combines a fine-tuned Whisper-large-v3 encoder with
efficient projector architectures and various decoder configurations. We employ
a three-sta... | [
"cs.CL",
"cs.SD",
"eess.AS"
] |
# 1 Introduction
The efficient processing of analytic queries is an important issue in databases and has motivated considerable research work since the last three decades. The main purpose of analytic queries is to extract relevant ‘statistics’ from huge volumes of data, resulting from the integration of heterogeneous... | We present an approach to computing consistent answers to analytic queries in
data warehouses operating under a star schema and possibly containing missing
values and inconsistent data. Our approach is based on earlier work concerning
consistent query answering for standard, non-analytic queries in multi-table
database... | [
"cs.DB"
] |
# 1 INTRODUCTION
Recent developments in Large Language Models (LLMs) have spurred extensive research into automating many activities, including code generation. In this paper, we address text-to-SQL generation [QHWY $+ 2 2$ ].
Our research was informed by our experiences with large-scale industrial databases [DJMS02... | Large Language Models (LLMs) have recently become sophisticated enough to
automate many tasks ranging from pattern finding to writing assistance to code
generation. In this paper, we examine text-to-SQL generation. We have observed
from decades of experience that the most difficult part of query development
lies in und... | [
"cs.DB"
] |
# 1 Introduction
Large Language Models (LLMs) and Vision Language Models (VLMs) have achieved remarkable success across a wide range of natural language processing tasks (Dong et al., 2024; Sheng et al., 2024; Zhu et al., 2023b; Yin et al., 2024), demonstrating strong capabilities in reasoning, knowledge retrieval, an... | Model editing aims to efficiently update a pre-trained model's knowledge
without the need for time-consuming full retraining. While existing pioneering
editing methods achieve promising results, they primarily focus on editing
single-modal language models (LLMs). However, for vision-language models
(VLMs), which involv... | [
"cs.CV",
"cs.AI"
] |
# 1 Introduction
The concept of a foundation model was conceived based on substantial evidence suggesting that a pre-trained Transformer can solve many natural language tasks when properly adapted [3, 71]. Moreover, when scaled with sufficient parameters, these Transformers elicit emergent abilities that are not prese... | A foundation model like GPT elicits many emergent abilities, owing to the
pre-training with broad inclusion of data and the use of the powerful
Transformer architecture. While foundation models in natural languages are
prevalent, can we build similar models for graphs? This paper describes an
approach toward a graph fo... | [
"cs.LG",
"cs.AI"
] |
# 1 Introduction
Large language models (LLMs) have seen widespread adoption in software development for code generation, bug fixing, question answering, test generation, and related tasks. Recently, agentic assistants such as Cursor (Anysphere Inc. [2025])—an AI-powered IDE based on Visual Studio Code—have gained trac... | We present the Comprehensive Verilog Design Problems (CVDP) benchmark, a new
dataset and infrastructure to advance LLM and agent research in hardware design
and verification. CVDP includes 783 problems across 13 task categories,
covering RTL generation, verification, debugging, specification alignment, and
technical Q&... | [
"cs.LG",
"cs.AR"
] |
# 1 Introduction
Accompanying each new software version is a crucial document, the release note. Release notes summarise the new features, enhancements, bug fixes, and other changes in a software update, conveying the rationale and impact of changes to downstream developers and end users [Moreno et al. 2017; Wu et al.... | The release note is a crucial document outlining changes in new software
versions. Yet, many developers view the process of writing software release
notes as a tedious and dreadful task. Consequently, numerous tools have been
developed by researchers and practitioners to automate the generation of
software release note... | [
"cs.SE"
] |
# 1 Introduction
Recent open-source work like DeepSeek R1 [Guo et al., 2025] has highlighted Reinforcement Fine-Tuning (RFT) with verifiable rewards as a promising approach for improving large models such as OpenAI’s o1 [Jaech et al., 2024]. While these techniques have shown success in text-based models, their applica... | Reinforcement Fine-Tuning (RFT) with verifiable rewards has advanced large
language models but remains underexplored for Vision-Language (VL) models. The
Vision-Language Reward Model (VL-RM) is key to aligning VL models by providing
structured feedback, yet training effective VL-RMs faces two major challenges.
First, t... | [
"cs.CL",
"cs.CV"
] |
# 1 Introduction
Large Language Models (LLMs) have demonstrated impressive capabilities in SWE tasks, including code generation, debugging, and automated development workflows. Building on these capabilities, researchers have begun creating LLM-driven agents that interact with real codebases and development environmen... | LLM-based agents have shown promising capabilities in a growing range of
software engineering (SWE) tasks. However, advancing this field faces two
critical challenges. First, high-quality training data is scarce, especially
data that reflects real-world SWE scenarios, where agents must interact with
development environ... | [
"cs.SE",
"cs.CL"
] |
# 1 Introduction
Large language models (LLMs) are revolutionizing software engineering by offering exceptional capabilities in natural language understanding and generation. These models can significantly enhance productivity by, e.g., automating code generation [1], helping in the development of cyber-physical system... | Large Language Models (LLMs) are used for many different software engineering
tasks. In software architecture, they have been applied to tasks such as
classification of design decisions, detection of design patterns, and
generation of software architecture design from requirements. However, there is
little overview on ... | [
"cs.SE"
] |
# 1 Introduction
Code optimization refers to rewriting the code such that it performs the same task more efficiently [29]. The efficiency is predominantly measured by the runtime, but it could also refer to storage, energy, or other resource consumptions. Code optimization is a less explored coding task in the context... | Recent studies show that LLMs possess different skills and specialize in
different tasks. In fact, we observe that their varied performance occur in
several levels of granularity. For example, in the code optimization task, code
LLMs excel at different optimization categories and no one dominates others.
This observati... | [
"cs.AI",
"cs.LG",
"cs.MA",
"cs.SE"
] |
# I. INTRODUCTION
Installation of CCTV cameras has become a necessity nowadays almost everywhere including shopping malls, theatres, houses and restaurants. This is due to the increase in violent activities all over the world and the importance of monitoring it to reduce human and property loss. But simply keeping an ... | The increasing global crime rate, coupled with substantial human and property
losses, highlights the limitations of traditional surveillance methods in
promptly detecting diverse and unexpected acts of violence. Addressing this
pressing need for automatic violence detection, we leverage Machine Learning to
detect and c... | [
"cs.CV",
"cs.AI"
] |
# 1 Introduction
Retrieval systems play a pivotal role in many NLP applications, enabling models to utilize relevant information from large corpora such as document collections, web pages, or conversational histories (Lewis et al., 2020; Gao et al., 2023). Relevance in retrieval can be established through a range of c... | Retrieval systems are central to many NLP pipelines, but often rely on
surface-level cues such as keyword overlap and lexical semantic similarity. To
evaluate retrieval beyond these shallow signals, recent benchmarks introduce
reasoning-heavy queries; however, they primarily shift the burden to query-side
processing te... | [
"cs.CL",
"cs.AI"
] |
# 1 INTRODUCTION
Index tuning is a time-consuming process that may take hours to finish for large and complex workloads. Existing index tuners typically adopt a cost-based tuning architecture [7, 41], as illustrated in Figure 1. It consists of three main components: (1) workload parsing and analysis, which parses each... | Index tuning is a time-consuming process. One major performance bottleneck in
existing index tuning systems is the large amount of "what-if" query optimizer
calls that estimate the cost of a given pair of query and index configuration
without materializing the indexes. There has been recent work on budget-aware
index t... | [
"cs.DB"
] |
# 1 Introduction
Human mobility simulation is a critical real-world task with widespread applications across many domains [27], such as supporting the implementation of the 15-minute city concept in urban development by modeling residents’ daily activities [45], optimizing transportation strategies through travel beha... | Human mobility simulation plays a crucial role in various real-world
applications. Recently, to address the limitations of traditional data-driven
approaches, researchers have explored leveraging the commonsense knowledge and
reasoning capabilities of large language models (LLMs) to accelerate human
mobility simulation... | [
"cs.CL",
"cs.AI"
] |
# 1 Introduction
Object detection, a key computer vision task, aims to identify and locate specific objects in images or videos [1]. Deep learning, especially CNN-based methods, has significantly advanced this field. However, traditional visible-light detection algorithms, which rely on RGB images, struggle in complex... | Multispectral object detection, which integrates information from multiple
bands, can enhance detection accuracy and environmental adaptability, holding
great application potential across various fields. Although existing methods
have made progress in cross-modal interaction, low-light conditions, and model
lightweight... | [
"cs.CV"
] |
# 1 INTRODUCTION
The successful replication of long chain-of-thought (CoT) reasoning, similar to that in OpenAI’s o1 (OpenAI, 2024), by DeepSeek-R1 (Guo et al., 2025) using the Group Relative Policy Optimization (GRPO) algorithm (Shao et al., 2024), has sparked a surge of interest within the open research community. T... | Reinforcement Learning with Verifiable Rewards (RLVR) has emerged as a
promising paradigm for advancing the reasoning capabilities of Large Language
Models (LLMs). However, a critical paradox clouds its efficacy: RLVR-tuned
models often underperform their base models on the $Pass@K$ metric for
solution-finding, leading... | [
"cs.AI",
"cs.CL"
] |
# 1 Introduction
While autoregressive models are effective, they also involve highly sequential sampling, cannot take bidirectional context, and lead to constrained architectures by requiring a decoder mask. In contrast, discrete diffusion models (Lou et al.; Sahoo et al., 2024) can parallelize generation by denoising... | Discrete diffusion models are a new class of text generators that offer
advantages such as bidirectional context use, parallelizable generation, and
flexible prompting compared to autoregressive models. However, a critical
limitation of discrete diffusion models is their inability to perform
flexible-length or flexible... | [
"cs.LG",
"cs.AI",
"cs.CL",
"cs.CV"
] |
# I. INTRODUCTION
In recent years, the prevalence of psychological disorders such as depression, anxiety, and stress has increased significantly due to the fast-paced nature of modern life. As a result, this has driven considerable interest in applying machine learning (ML) techniques for early detection, accurate dia... | Depression is a significant mental health concern, particularly in
professional environments where work-related stress, financial pressure, and
lifestyle imbalances contribute to deteriorating well-being. Despite increasing
awareness, researchers and practitioners face critical challenges in developing
accurate and gen... | [
"cs.LG"
] |
# 1 Introduction
Large language models (LLMs) have fundamentally reshaped the landscape of software engineering [7], powering tools such as Cursor [4] and GitHub Copilot [5] that are now integral to modern development workflows. These models have transformed key stages of the software development lifecycle—automated c... | The issue-resolving task, where a model generates patches to fix real-world
bugs, has emerged as a critical benchmark for evaluating the capabilities of
large language models (LLMs). While SWE-bench and its variants have become
standard in this domain, they suffer from key limitations: they have not been
updated since ... | [
"cs.SE",
"cs.AI",
"cs.CL"
] |
# 1 Introduction
Anomaly detection (AD) — the task of identifying data that deviate from expected behavior — is central in many domains, from daily usage in manufacturing [3] and content moderation [8] to high stakes domains like cybersecurity [43, 23] and healthcare [31, 14]. Despite its broad applicability, most AD ... | Anomaly detection (AD) is a critical task across domains such as
cybersecurity and healthcare. In the unsupervised setting, an effective and
theoretically-grounded principle is to train classifiers to distinguish normal
data from (synthetic) anomalies. We extend this principle to semi-supervised
AD, where training data... | [
"stat.ML",
"cs.CR",
"cs.LG",
"stat.AP"
] |
# 1 Introduction
Text-to-SQL (Tai et al., 2023; Li et al., 2024b; Shi et al., 2025) aims to translate natural language into structured database queries, playing a crucial role in democratizing data access by enabling nontechnical users to interact effectively with relational databases. A significant amount of work is ... | Reinforcement learning (RL) has been widely adopted to enhance the
performance of large language models (LLMs) on Text-to-SQL tasks. However,
existing methods often rely on execution-based or LLM-based Bradley-Terry
reward models. The former suffers from high execution latency caused by
repeated database calls, whereas... | [
"cs.LG",
"cs.DB",
"cs.PL"
] |
# 1 Introduction
Recent advancements in large language models (LLMs) have been significantly influenced by the Mixture-of-Experts (MoE) architecture [20, 13, 5, 27, 28, 6, 7], which leverages dynamic routing mechanisms to scale model parameters efficiently while maintaining sub-linear increases in computational requir... | Mixture-of-Experts (MoE) models have emerged as a cornerstone of large-scale
deep learning by efficiently distributing computation and enhancing
performance. However, their unique architecture-characterized by sparse expert
activation and dynamic routing mechanisms-introduces inherent complexities that
challenge conven... | [
"cs.CL"
] |
# 1 INTRODUCTION
The $k$ nearest neighbor (KNN) search over high-dimensional vectors has become a fundamental operator in modern data systems, powering applications including recommendation systems [40], data mining [6], face recognition [46], product search [46], and retrievalargument generation (RAG) for large langu... | Real-world vector embeddings are usually associated with extra labels, such
as attributes and keywords. Many applications require the nearest neighbor
search that contains specific labels, such as searching for product image
embeddings restricted to a particular brand. A straightforward approach is to
materialize all p... | [
"cs.DB"
] |
# 1 Introduction
Large language models (LLMs) have shown strong performance in programming [1–3] and are widely adopted in tools like Cursor and GitHub Copilot to boost developer productivity [4]. LLM-generated code is becoming prevalent in commercial software [5] and may eventually form a substantial portion of the w... | Large language models (LLMs) are increasingly integrated in software
development, but ensuring correctness in LLM-generated code remains challenging
and often requires costly manual review. Verifiable code generation -- jointly
generating code, specifications, and proofs of code-specification alignment --
offers a prom... | [
"cs.LG",
"cs.AI",
"cs.LO",
"cs.PL",
"cs.SE"
] |
# 1 Introduction
Large Language Models (LLMs) have rapidly transformed digital communication, enabling machines to generate text that closely mimics human writing. Their utility in domains such as virtual assistants, content production, and automated dialogue is evident. However, their widespread and unregulated use i... | The increasing sophistication of large language models (LLMs) has sparked
growing concerns regarding their potential role in exacerbating ideological
polarization through the automated generation of persuasive and biased content.
This study explores the extent to which fine-tuned LLMs can replicate and
amplify polarizi... | [
"cs.CL",
"cs.CY"
] |
# 1 INTRODUCTION
The convergence of high-resolution structural biology and generative AI has ushered in a new era of rational drug design. Breakthroughs in cryo-EM and geometric deep learning now permit direct generation of ligand molecules within three-dimensional protein pockets, effectively unifying structural insi... | Breakthroughs in high-accuracy protein structure prediction, such as
AlphaFold, have established receptor-based molecule design as a critical driver
for rapid early-phase drug discovery. However, most approaches still struggle
to balance pocket-specific geometric fit with strict valence and synthetic
constraints. To re... | [
"q-bio.BM",
"cs.LG"
] |
# 1 Introduction
The chase is a fundamental algorithm in database theory that is applied to address a wide range of problems. For instance, it is used to check containment of queries under constraints, in data exchange settings, or to solve ontology-based query answering; see the introductions of [11, 13] for more inf... | The chase is a fundamental algorithm with ubiquitous uses in database theory.
Given a database and a set of existential rules (aka tuple-generating
dependencies), it iteratively extends the database to ensure that the rules are
satisfied in a most general way. This process may not terminate, and a major
problem is to d... | [
"cs.LO",
"cs.DB"
] |
# 1. Introduction
Large language models (LLMs) [3] [6] have emerged as powerful tools across a wide range of applications, from content generation to customer service [2, 18, 22]. As the technology behind these models advances, there is growing interest in their ability to integrate with external tools [7, 29, 43, 50]... | Large language models have demonstrated exceptional performance, yet struggle
with complex tasks such as numerical reasoning, plan generation. Integrating
external tools, such as calculators and databases, into large language models
(LLMs) is crucial for enhancing problem-solving capabilities. Current methods
assign a ... | [
"cs.CL",
"cs.AI"
] |
# I. Introduction
of egocentric video understanding, driven by improvements in foundation models [1]–[4], pretraining strategies [5]–[7], loss functions [8], [9], and data augmentations [10]. Despite significant performance gains, the increasing scale of models, prolonged training pipelines, and ever-larger datasets h... | Egocentric video-language understanding demands both high efficiency and
accurate spatial-temporal modeling. Existing approaches face three key
challenges: 1) Excessive pre-training cost arising from multi-stage
pre-training pipelines, 2) Ineffective spatial-temporal encoding due to
manually split 3D rotary positional ... | [
"cs.CV",
"cs.AI"
] |
# 1 Introduction
Language models (LMs) have become powerful tools for automating software engineering tasks (Hou et al., 2024). A natural extension is software vulnerability detection, which plays a critical role in the early stages of the software lifecycle. Researchers typically fine-tune general or codespecific LMs... | Language models (LMs) show promise for vulnerability detection but struggle
with long, real-world code due to sparse and uncertain vulnerability locations.
These issues, exacerbated by token limits, often cause models to miss
vulnerability-related signals, thereby impairing effective learning. A key
intuition is to enh... | [
"cs.SE"
] |
# 1 Introduction
Spatial data has become integral to modern applications due to the evergrowing amount of Internet-of-Things (IoT) devices. These devices produce geospatial information used in services such as real-time traffic-aware routing, incident hotspot detection, and weather updates and predictions, with variou... | The growing number of moving Internet-of-Things (IoT) devices has led to a
surge in moving object data, powering applications such as traffic routing,
hotspot detection, or weather forecasting. When managing such data, spatial
database systems offer various index options and data formats, e.g.,
point-based or trajector... | [
"cs.DB",
"cs.DC"
] |
# I. INTRODUCTION
representations limit their feasibility for real-time rendering applications.
More recently, Gaussian-based approaches [14], [15] have emphasized modeling the intrinsic features of each Gaussian point, and simultaneously combining the appearance embeddings to predict affine transformations of base c... | Reconstructing 3D scenes from unconstrained image collections poses
significant challenges due to variations in appearance. In this paper, we
propose Scalable Micro-macro Wavelet-based Gaussian Splatting (SMW-GS), a novel
method that enhances 3D reconstruction across diverse scales by decomposing
scene representations ... | [
"cs.CV"
] |
# I. INTRODUCTION
The rapid advancement in digital technologies has significantly transformed scientific research, facilitating the collection, processing, and analysis of extensive datasets. However, the growing diversity and complexity of research data present substantial challenges, particularly in terms of findabi... | FAIR Digital Objects support research data management aligned with the FAIR
principles. To be machine-actionable, they must support operations that
interact with their contents. This can be achieved by associating operations
with FAIR-DO data types. However, current typing models and Data Type
Registries lack support f... | [
"cs.DL",
"cs.DB"
] |
# 1 INTRODUCTION
Statutory Interpretation (SI) is an important task in the scope of Artificial Intelligence (AI) and Law. It is concerned with the interpretation of legal concepts made by judges in the court rulings. The classical example is the sentence “No vehicles in the park” with respect to the vehicles concept. ... | One of the elements of legal research is looking for cases where judges have
extended the meaning of a legal concept by providing interpretations of what a
concept means or does not mean. This allow legal professionals to use such
interpretations as precedents as well as laymen to better understand the legal
concept. T... | [
"cs.CL"
] |
# 1. Introduction
As neural networks grow following established scaling laws (Kaplan et al., 2020; Hoffmann et al., 2022), they become increasingly inaccessible to much of the research community. Training models with hundreds of billions of parameters requires computational resources available only to select instituti... | Training large neural networks with end-to-end backpropagation creates
significant memory bottlenecks, limiting accessibility to state-of-the-art AI
research. We propose $\textit{DiffusionBlocks}$, a novel training framework
that interprets neural network blocks as performing denoising operations in a
continuous-time d... | [
"cs.LG",
"cs.AI",
"stat.ML"
] |
# 1 Introduction and Related Work
Effective decision-making in networks—such as in communication networks, social networks, and transportation networks— often relies on graph-structured data representations. Among the techniques developed for learning from such data, Graph Neural Networks (GNNs) have become widely ado... | Effective decision-making on networks often relies on learning from
graph-structured data, where Graph Neural Networks (GNNs) play a central role,
but they take efforts to configure and tune. In this demo, we propose LLMNet,
showing how to design GNN automated through Large Language Models. Our system
develops a set of... | [
"cs.LG"
] |
# 1 Introduction
Large language models have achieved remarkable progress in a wide range of natural language understanding and generation tasks, largely powered by next word prediction (NWP) algorithm and supervised finetuning during pretraining and posttraining (Devlin et al., 2019; Raffel et al., 2020; Brown et al.,... | Large language models (LLMs) are typically trained via next-word prediction
(NWP), which provides strong surface-level fluency but often lacks support for
robust reasoning. We propose BOttlenecked next Word exploration (BOW), a novel
RL framework that rethinks NWP by introducing a reasoning bottleneck where a
policy mo... | [
"cs.CL"
] |
# 1 Introduction
We are interested in studying the effect of treatment e.g., different policies and drugs, on rare yet impactful events such as large wildfires, hurricanes, tsunamis and climate change.
These kinds of events happen at an extremely low frequency, but they can cause considerable damage to properties and... | Causal effect estimation seeks to determine the impact of an intervention
from observational data. However, the existing causal inference literature
primarily addresses treatment effects on frequently occurring events. But what
if we are interested in estimating the effects of a policy intervention whose
benefits, whil... | [
"stat.ML",
"cs.LG",
"math.ST",
"stat.ME",
"stat.TH"
] |
# 1. Introduction
Computational models of the interstellar medium help us to understand the physical structure and chemical content that we observe in astronomical regions such as the Orion bar (Peeters et al.; 2024). In order to understand the transition from the low density medium into the high density medium, three... | Computational astrochemical models are essential for helping us interpret and
understand the observations of different astrophysical environments. In the age
of high-resolution telescopes such as JWST and ALMA, the substructure of many
objects can be resolved, raising the need for astrochemical modeling at these
smalle... | [
"astro-ph.GA",
"cs.LG"
] |
# 1 INTRODUCTION
In scientific research, datasets play a crucial role in method validation, model training, and result evaluation. Currently, research in many fields relies heavily on datasets, such as disease prediction in the medical field [18] and climate forecasting meteorology [44]. These studies often involve di... | Scientific research heavily depends on suitable datasets for method
validation, but existing academic platforms with dataset management like
PapersWithCode suffer from inefficiencies in their manual workflow. To overcome
this bottleneck, we present a system, called ChatPD, that utilizes Large
Language Models (LLMs) to ... | [
"cs.DB",
"cs.AI",
"cs.IR"
] |
# 1 Introduction
The field of medicine increasingly relies on machine learning tools for clinical decision support. In diagnosis and prognosis, probabilistic scores capture uncertainty about patient outcomes. Combined with value judgments, they produce expected values that guide clinical decisions. It is less often em... | Machine learning-based decision support systems are increasingly deployed in
clinical settings, where probabilistic scoring functions are used to inform and
prioritize patient management decisions. However, widely used scoring rules,
such as accuracy and AUC-ROC, fail to adequately reflect key clinical
priorities, incl... | [
"cs.LG",
"cs.AI"
] |
# 1 Introduction
Online social media platforms, despite their limitations (Ivan et al., 2015) and potential risks (Bert et al., 2016; Abolfathi et al., 2022), have revolutionized how individuals connect and communicate with others who share similar interests. The rapid growth in their usage can be attributed to the ub... | Search engines play a crucial role as digital gatekeepers, shaping the
visibility of Web and social media content through algorithmic curation. This
study investigates how search engines like Google selectively promotes or
suppresses certain hashtags and subreddits, impacting the information users
encounter. By compari... | [
"cs.CL"
] |
# 1 Introduction
Tabular data is a prevalent modality in numerous real-world applications [39], including critical areas such as disease diagnosis [6], credit scoring [13], census data analysis [35], and cybersecurity [7]. Consequently, developing effective tabular foundation models [48, 21, 22] has become a pressing ... | Tabular data, ubiquitous and rich in informational value, is an increasing
focus for deep representation learning, yet progress is hindered by studies
centered on single tables or isolated databases, which limits model
capabilities due to data scale. While collaborative learning approaches such as
federated learning, t... | [
"cs.DB",
"cs.LG"
] |
# I. INTRODUCTION
Pedestrian trajectory prediction is a fundamental task in domains such as robotics and surveillance systems aimed at enhancing public safety and services efficiency. Accurate forecasting of human movement allows these systems to operate safely and efficiently in dynamic and crowded environments. Alth... | Accurate prediction of pedestrian trajectories is essential for applications
in robotics and surveillance systems. While existing approaches primarily focus
on social interactions between pedestrians, they often overlook the rich
environmental context that significantly shapes human movement patterns. In
this paper, we... | [
"cs.CV",
"cs.AI"
] |
# 1 Introduction
Image restoration (IR) addresses the reconstruction of high-quality images from degraded inputs, with super-resolution and inpainting representing its fundamental tasks. Traditional IR techniques, such as bicubic or B-spline[5] methods, often produce blurry results with compromised details, whereas mo... | Image restoration faces challenges including ineffective feature fusion,
computational bottlenecks and inefficient diffusion processes. To address
these, we propose DiffRWKVIR, a novel framework unifying Test-Time Training
(TTT) with efficient diffusion. Our approach introduces three key innovations:
(1) Omni-Scale 2D ... | [
"cs.CV",
"I.4.9"
] |
# Introduction
Reasoning about the presence and properties of occluded objects is a fundamental aspect of human intelligence, crucial for navigating and interacting with a complex world. This ability is based on the integration of various sensory cues and contextual information to form probabilistic judgments. Conside... | People regularly make inferences about objects in the world that they cannot
see by flexibly integrating information from multiple sources: auditory and
visual cues, language, and our prior beliefs and knowledge about the scene. How
are we able to so flexibly integrate many sources of information to make sense
of the w... | [
"cs.AI"
] |
# 1 INTRODUCTION
Software testing plays a crucial role in ensuring software quality, primarily through the selection of test data that evaluates whether a program behaves as intended. Ideally, testing would cover all possible inputs to verify that software functions correctly. However, covering all inputs is generally... | Software systems exhibit distinct behaviors based on input characteristics,
and failures often occur at the boundaries between input domains. Traditional
Boundary Value Analysis (BVA) relies on manual heuristics, while automated
Boundary Value Exploration (BVE) methods typically optimize a single quality
metric, riskin... | [
"cs.SE",
"D.2.5"
] |
# 1. Introduction
Recent “long-thought” Large Reasoning Models (LRMs), such as OpenAI’s O1 (Jaech et al., 2024) and Deepseek-R1 (DeepSeek-AI et al., 2025), represent a significant paradigm extension of foundational Chain-of-Thought (CoT) techniques (Wei et al., 2023). Fine-tuned with Reinforcement Learning (RL), these... | Large Reasoning Models (LRMs) have achieved remarkable success, yet they
often suffer from producing unnecessary and verbose reasoning chains. We
identify a core aspect of this issue as "invalid thinking" -- models tend to
repeatedly double-check their work after having derived the correct answer. To
address this speci... | [
"cs.AI",
"cs.CL"
] |
# 1 Introduction
Large language models (LLMs) have shown remarkable performances on various NLP tasks, however, due to their huge sizes of model parameters, LLMs require significant GPU memory for inference, substantially limiting the throughput and latency time. To address these challenges, quantization methods (Yao ... | Quantization has been widely studied as an effective technique for reducing
the memory requirement of large language models (LLMs), potentially improving
the latency time as well. Utilizing the characteristic of rotational invariance
of transformer, we propose the rotation-based saliency-aware weight
quantization (ROSA... | [
"cs.CL",
"cs.AI"
] |
# INTRODUCTION
Artificial Intelligence (AI) software has become a critical component in numerous applications, ranging from autonomous driving [1] and healthcare diagnostics [2] to financial decisionmaking and public service automation [3]. The rapid advancement and adoption of AI technologies have brought profound be... | As Artificial Intelligence (AI) systems, particularly those based on machine
learning (ML), become integral to high-stakes applications, their probabilistic
and opaque nature poses significant challenges to traditional verification and
validation methods. These challenges are exacerbated in regulated sectors
requiring ... | [
"cs.SE",
"cs.CR"
] |
# 1. INTRODUCTION
Multimodal data combining structural, functional, and genetic information has been increasingly used in medical imaging research for brain-related predictions and neurological disorder analysis, with advanced models like vision transformers (ViT) predicting diseases and identifying biomarkers [1, 2, ... | Multimodal medical imaging integrates diverse data types, such as structural
and functional neuroimaging, to provide complementary insights that enhance
deep learning predictions and improve outcomes. This study focuses on a
neuroimaging prediction framework based on both structural and functional
neuroimaging data. We... | [
"eess.IV",
"cs.CV"
] |
# 1 Introduction
Relevant and up-to-date documentation is useful for software maintenance (Stapleton et al., 2020; Misra et al., 2020; de Souza et al., 2006)/ To support one important form of documentation, researchers have developed models that generate one-line summaries of functions (Hu et al., 2018a; LeClair et al... | Code documentation is useful, but writing it is time-consuming. Different
techniques for generating code summaries have emerged, but comparing them is
difficult because human evaluation is expensive and automatic metrics are
unreliable. In this paper, we introduce a simple new baseline in which we ask
an LLM to give an... | [
"cs.CL",
"cs.AI",
"cs.SE",
"68T50",
"I.2.7"
] |
1 Introduction 3
2 Related Work 4
2.1 Contributions 4
# 3 Taxonomy 5
3.1 Formal Definition 5
3.2 Desiderata 5
3.3 Development Methodology 5
3.4 Selected Categories 6
# 4 Downstream Results 6
4.1 Experimental Protocol 6
4.2 Math 8
4.3 Code 9
4.4 Medical 9
4.5 STEM 10
# 5 Developing Taxonomy 11
5.1 Measuring Ortho... | Data plays the most prominent role in how language models acquire skills and
knowledge. The lack of massive, well-organized pre-training datasets results in
costly and inaccessible data pipelines. We present Essential-Web v1.0, a
24-trillion-token dataset in which every document is annotated with a
twelve-category taxo... | [
"cs.CL",
"cs.AI",
"cs.LG"
] |
# 1 Introduction
Conversational Recommendation Systems (CRSs) leverage multi-turn natural language interactions to gradually uncover user interests and subsequently recommend items aligned with their preferences (Jannach et al., 2021; Gao et al., 2021). Powered by the advanced text generation and toolcalling capabilit... | Recent advancements in Large Language Models (LLMs) have significantly
propelled the development of Conversational Recommendation Agents (CRAs).
However, these agents often generate short-sighted responses that fail to
sustain user guidance and meet expectations. Although preference optimization
has proven effective in... | [
"cs.CL"
] |
# 1 Introduction
Computational science communities have proposed numerous learning-based approaches for solving PDE-governed systems for simulation, optimization, or scientific discovery. These methods provide tradeoffs across accuracy, applicability, and speed. Physics-informed neural networks (PINNs) [59, 60] can be... | We present a unified framework for solving partial differential equations
(PDEs) using video-inpainting diffusion transformer models. Unlike existing
methods that devise specialized strategies for either forward or inverse
problems under full or partial observation, our approach unifies these tasks
under a single, flex... | [
"cs.LG",
"cs.AI",
"cs.CV"
] |
# 1 Introduction
While diffusion models have set a new standard for photorealism in generative art [1], their operational costs remain a major challenge. The generation of a single image can involve many denoising steps, each utilizes a learned denoiser model with potentially over a billion parameters [2]. This makes ... | Diffusion models are well known for their ability to generate a high-fidelity
image for an input prompt through an iterative denoising process.
Unfortunately, the high fidelity also comes at a high computational cost due
the inherently sequential generative process. In this work, we seek to
optimally balance quality an... | [
"cs.CV",
"cs.LG"
] |
# 1 Introduction
Document intelligence systems are transforming modern business operations by automating the extraction of critical information from unstructured documents [35,36]. Within this domain, Table Extraction (TE) plays a pivotal role in converting tabular data into structured formats for downstream applicati... | Automating table extraction (TE) from business documents is critical for
industrial workflows but remains challenging due to sparse annotations and
error-prone multi-stage pipelines. While semi-supervised learning (SSL) can
leverage unlabeled data, existing methods rely on confidence scores that poorly
reflect extracti... | [
"cs.AI"
] |
# I. INTRODUCTION
practice for describing interfaces for integrating systems. It contains formal elements like paths and natural language constituents such as descriptions. For integrating these systems automatically, automated service composition using Large Language Models (LLMs) has been recently proposed [1], [2],... | Integrating multiple (sub-)systems is essential to create advanced
Information Systems. Difficulties mainly arise when integrating dynamic
environments, e.g., the integration at design time of not yet existing
services. This has been traditionally addressed using a registry that provides
the API documentation of the en... | [
"cs.SE",
"cs.AI"
] |
# 1 Introduction
Imitation learning for sensorimotor skills has made significant strides in recent years, propelled by the increasing scale and diversity of robotic datasets. From controlled tabletop environments to open-world household settings, large-scale data has been shown to improve the generalization of robotic... | Data-driven approaches struggle with precise manipulation; imitation learning
requires many hard-to-obtain demonstrations, while reinforcement learning
yields brittle, non-generalizable policies. We introduce VisuoTactile Local
(ViTaL) policy learning, a framework that solves fine-grained manipulation
tasks by decompos... | [
"cs.RO",
"cs.CV"
] |
# 1 Introduction
Invariant manifolds (IMs) and attendant methods remain powerful, versatile, and insightful tools in the broader nonlinear dynamical systems theory. An invariant manifold is a subset embedded in the system’s state space with the property that any trajectory starting on it remains on it for all time. Th... | We propose a hybrid machine learning scheme to learn -- in physics-informed
and numerical analysis-informed fashion -- invariant manifolds (IM) of discrete
maps for constructing reduced-order models (ROMs) for dynamical systems. The
proposed scheme combines polynomial series with shallow neural networks,
exploiting the... | [
"math.NA",
"cs.LG",
"cs.NA",
"math.DS",
"37M10, 37M21, 37N30, 37N35, 65P99, 65D15, 68T07",
"G.1.0; F.1.1; I.2.6"
] |
# 1. Introduction
In recent years, Multi-Agent Reinforcement Learning (MARL) has emerged as a critical area of research within the broader field of machine learning. While traditional reinforcement learning focuses on training a single agent to interact with an environment and maximize cumulative rewards, MARL extends... | This paper investigates multi-agent reinforcement learning (MARL) in a
partially observable, cooperative-competitive combat environment known as LAG.
We describe the environment's setup, including agent actions, hierarchical
controls, and reward design across different combat modes such as No Weapon and
ShootMissile. T... | [
"cs.LG",
"cs.MA"
] |
# 1 Introduction
3D scence reconstruction remains a longstanding challenge in computer vision and graphics. A significant advancement in this domain is the Neural Radiance Field (NeRF) [36], which effectively represents geometry and view-dependent appearance using multi-layer perceptrons (MLPs), demonstrating signific... | 3D Gaussian Splatting (3DGS) has made significant strides in real-time 3D
scene reconstruction, but faces memory scalability issues in high-resolution
scenarios. To address this, we propose Hierarchical Gaussian Splatting (HRGS),
a memory-efficient framework with hierarchical block-level optimization. First,
we generat... | [
"cs.CV",
"cs.AI"
] |
# 1 Introduction
As LLMs become increasingly integrated into real-world applications, hallucination—where LLMs generate unsupported or misleading information—remains a fundamental limitation. Ensuring that their responses remain factual is a critical challenge. This attention has led to the development of factuality e... | Fact verification is essential for ensuring the reliability of LLM
applications. In this study, we evaluate 12 pre-trained LLMs and one
specialized fact-verifier, including frontier LLMs and open-weight reasoning
LLMs, using a collection of examples from 14 fact-checking benchmarks. We share
three findings intended to ... | [
"cs.AI",
"cs.CL",
"cs.LG"
] |
# 1 Introduction
Developers spend a significant amount of time reading and comprehending code [20, 47], and identifier names play a central role in this process, accounting for roughly $7 0 \%$ of all code characters [21]. Prior work shows that the quality of identifier names significantly impacts comprehension [10,15... | Identifier names are crucial components of code, serving as primary clues for
developers to understand program behavior. This paper investigates the
linguistic structure of identifier names by extending the concept of grammar
patterns; representations of the part-of-speech (PoS) sequences that underlie
identifier phras... | [
"cs.SE"
] |
# Introduction
Ensuring computational reproducibility is increasingly recognized as a cornerstone of credible scientific research (Peng 2011; Seibold et al. 2021; National Academies of Sciences and Medicine 2019). Several works, beginning in the 1990s (Claerbout and Karrenbach 1992), have highlighted the importance of... | Computational reproducibility is fundamental to scientific research, yet many
published code supplements lack the necessary documentation to recreate their
computational environments. While researchers increasingly share code alongside
publications, the actual reproducibility of these materials remains poorly
understoo... | [
"cs.CY",
"cs.SE"
] |
# 1 Introduction
Tokenization is a crucial component of nearly all modern language models: it allows them to consume and produce arbitrary streams of text using only finite vocabularies. The vast majority of tokenizers in use today, such as those based on Byte-Pair Encoding (BPE) [57] or Unigram [27], feature tokens s... | Tokenization is used almost universally by modern language models, enabling
efficient text representation using multi-byte or multi-character tokens.
However, prior work has shown that tokenization can introduce distortion into
the model's generations. For example, users are often advised not to end their
prompts with ... | [
"cs.CL",
"cs.FL",
"cs.LG"
] |
# 1 INTRODUCTION
Vector similarity search [25, 33, 34, 43–45] aims to identify the most similar vectors from a large dataset given a query vector. Approximate nearest neighbor (ANN) search has emerged as a practical alternative for exact top- $k$ nearest neighbor search, offering significant speedups with minimal accu... | Efficient vector query processing is critical to enable AI applications at
scale. Recent solutions struggle with growing vector datasets that exceed
single-machine memory capacity, forcing unnecessary data movement and resource
underutilization in monolithic architectures. We present d-HNSW, the first
disaggregated vec... | [
"cs.DB"
] |
# 1 Introduction
Neuro-symbolic AI aims for an integration of symbolic and subsymbolic AI approaches, to overcome the limitations of either. Symbolic AI is useful for the communication between humans and machines, for guiding the reasoning (in terms of logical and mathematical reasoning, but also in terms of common-se... | The goal of neuro-symbolic AI is to integrate symbolic and subsymbolic AI
approaches, to overcome the limitations of either. Prominent systems include
Logic Tensor Networks (LTN) or DeepProbLog, which offer neural predicates and
end-to-end learning. The versatility of systems like LTNs and DeepProbLog,
however, makes t... | [
"cs.AI",
"cs.LO"
] |
# 1 Introduction
The Background of Geo-localization. The rapid growth of visual content on social media and mobile devices, has made image geo-localization (determining where an image was taken) increasingly important for downstream applications such as autonomous navigation [15] and crisis response [18]. Given that m... | Previous methods for image geo-localization have typically treated the task
as either classification or retrieval, often relying on black-box decisions
that lack interpretability. The rise of large vision-language models (LVLMs)
has enabled a rethinking of geo-localization as a reasoning-driven task
grounded in visual ... | [
"cs.CV"
] |
# 1 Introduction
With the recent success of deep-thinking large language models (LLMs) – such as OpenAI O1 [15], DeepSeek R1 [7], and Qwen3 [25, 26], which are capable of generating long sequences of thoughts to achieve better performance – there has been a growing need to control the reasoning length of these models ... | Recent deep-thinking large language models often reason extensively to
improve performance, but such lengthy reasoning is not always desirable, as it
incurs excessive inference costs with disproportionate performance gains.
Controlling reasoning length without sacrificing performance is therefore
important, but remains... | [
"cs.CL",
"cs.AI"
] |
# Introduction
Psychophysics experiments that require the physical presence of participants in a laboratory suffer from limitations in sample size and recruitment options. For example, they often underrepresent participants with accessibility and mobility limitations and overrepresent student populations (Druckman & K... | In recent years, there is a growing need and opportunity to use online
platforms for psychophysics research. Online experiments make it possible to
evaluate large and diverse populations remotely and quickly, complementing
laboratory-based research. However, developing and running online psychophysics
experiments poses... | [
"q-bio.NC",
"cs.HC",
"cs.SE"
] |
# Introduction
Zero-shot learning (ZSL) has emerged as a promising paradigm for recognizing novel categories without requiring labeled examples (Lampert, Nickisch, and Harmeling 2009; Han et al. 2021). In particular, Transductive ZSL, where the entire unlabeled test set is available at inference time, offers a valuabl... | Transductive zero-shot learning (ZSL) aims to classify unseen categories by
leveraging both semantic class descriptions and the distribution of unlabeled
test data. While Vision-Language Models (VLMs) such as CLIP excel at aligning
visual inputs with textual semantics, they often rely too heavily on
class-level priors ... | [
"cs.CV"
] |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.