Daily Arxiv

This page organizes papers related to artificial intelligence published around the world.
This page is summarized using Google Gemini and is operated on a non-profit basis.
The copyright of the paper belongs to the author and the relevant institution. When sharing, simply cite the source.

LLMs for Automated Unit Test Generation and Assessment in Java: The AgoneTest Framework

BengaliFig: A Low-Resource Challenge for Figurative and Culturally Grounded Reasoning in Bengali

R3A: Reliable RTL Repair Framework with Multi-Agent Fault Localization and Stochastic Tree-of-Thoughts Patch Generation

EmoFeedback$^2$: Reinforcement of Continuous Emotional Image Generation via LVLM-based Reward and Textual Feedback

Agent0-VL: Exploring Self-Evolving Agent for Tool-Integrated Vision-Language Reasoning

A Systematic Analysis of Large Language Models with RAG-enabled Dynamic Prompting for Medical Error Detection and Correction

TiCT: A Synthetically Pre-Trained Foundation Model for Time Series Classification

TREASURE: A Transformer-Based Foundation Model for High-Volume Transaction Understanding

Human Experts' Evaluation of Generative AI for Contextualizing STEAM Education in the Global South

Pistachio: Towards Synthetic, Balanced, and Long-Form Video Anomaly Benchmarks

PrefixGPT: Prefix Adder Optimization by a Generative Pre-trained Transformer

UniGame: Turning a Unified Multimodal Model Into Its Own Adversary

DR Tulu: Reinforcement Learning with Evolving Rubrics for Deep Research

Adversarial Attack-Defense Co-Evolution for LLM Safety Alignment via Tree-Group Dual-Aware Search and Optimization

Uncertainty-Aware Deep Learning Framework for Remaining Useful Life Prediction in Turbofan Engines with Learned Aleatoric Uncertainty

ConceptGuard: Proactive Safety in Text-and-Image-to-Video Generation through Multimodal Risk Detection

CLLMRec: LLM-powered Cognitive-Aware Concept Recommendation via Semantic Alignment and Prerequisite Knowledge Distillation

TimeViper: A Hybrid Mamba-Transformer Vision-Language Model for Efficient Long Video Understanding

Think Visually, Reason Textually: Vision-Language Synergy in ARC

Cheating Stereo Matching in Full-scale: Physical Adversarial Attack against Binocular Depth Estimation in Autonomous Driving

Passive Dementia Screening via Facial Temporal Micro-Dynamics Analysis of In-the-Wild Talking-Head Video

Lost in Serialization: Invariance and Generalization of LLM Graph Reasoners

DensiCrafter: Physically-Constrained Generation and Fabrication of Self-Supporting Hollow Structures

Diffusion Models at the Drug Discovery Frontier: A Review on Generating Small Molecules versus Therapeutic Peptides

SARVLM: A Vision Language Foundation Model for Semantic Understanding and Target Recognition in SAR Imagery

Weak-to-Strong Generalization under Distribution Shifts

LightMem: Lightweight and Efficient Memory-Augmented Generation

SaFiRe: Saccade-Fixation Reiteration with Mamba for Referring Image Segmentation

Automated Neural Architecture Design for Industrial Defect Detection

PointNSP: Autoregressive 3D Point Cloud Generation with Next-Scale Level-of-Detail Prediction

Uncovering Implicit Bias in Large Language Models with Concept Learning Dataset

The Impossibility of Inverse Permutation Learning in Transformer Models

Mechanism of Task-oriented Information Removal in In-context Learning

A Unified Noise-Curvature View of Loss of Trainability

Modular, On-Site Solutions with Lightweight Anomaly Detection for Sustainable Nutrient Management in Agriculture

Not All Splits Are Equal: Rethinking Attribute Generalization Across Unrelated Categories

Equivariant Flow Matching for Symmetry-Breaking Bifurcation Problems

Consistent Opponent Modeling of Static Opponents in Imperfect-Information Games

Special-Character Adversarial Attacks on Open-Source Language Model

Where to Start Alignment? Diffusion Large Language Model May Demand a Distinct Position

WeatherDiffusion: Controllable Weather Editing in Intrinsic Space

Empowering Time Series Forecasting with LLM-Agents

Contrast-Prior Enhanced Duality for Mask-Free Shadow Removal

AutoDiscovery: Open-ended Scientific Discovery via Bayesian Surprise

Fair Algorithms with Probing for Multi-Agent Multi-Armed Bandits

Rigor in AI: Doing Rigorous AI Work Requires a Broader, Responsible AI-Informed Conception of Rigor

UITron-Speech: Towards Automated GUI Agents Based on Speech Instructions

Flow Matching Meets PDEs: A Unified Framework for Physics-Constrained Generation

Physics-Constrained Flow Matching: Sampling Generative Models with Hard Constraints

IVY-FAKE: A Unified Explainable Framework and Benchmark for Image and Video AIGC Detection

LCB-CV-UNet: Enhanced Detector for High Dynamic Range Radar Signals

Characterizing Pattern Matching and Its Limits on Compositional Task Structures

Force Prompting: Video Generation Models Can Learn and Generalize Physics-based Control Signals

How Do Companies Manage the Environmental Sustainability of AI? An Interview Study About Green AI Efforts and Regulations

MigGPT: Harnessing Large Language Models for Automated Migration of Out-of-Tree Linux Kernel Patches Across Versions

Reasoning Transfer for an Extremely Low-Resource and Endangered Language: Bridging Languages Through Sample-Efficient Language Understanding

OuroMamba: A Data-Free Quantization Framework for Vision Mamba

TS-RAG: Retrieval-Augmented Generation based Time Series Foundation Models are Stronger Zero-Shot Forecaster

Bridging Critical Gaps in Convergent Learning: How Representational Alignment Evolves Across Layers, Training, and Distribution Shifts

Personalized Image Generation for Recommendations Beyond Catalogs

Without Paired Labeled Data: End-to-End Self-Supervised Learning for Drone-view Geo-Localization

Meursault as a Data Point

Adaptive Object Detection for Indoor Navigation Assistance: A Performance Evaluation of Real-Time Algorithms

CroMe: Multimodal Fake News Detection using Cross-Modal Tri-Transformer and Metric Learning

BoundingDocs: a Unified Dataset for Document Question Answering with Spatial Annotations

Beyond Introspection: Reinforcing Thinking via Externalist Behavioral Feedback

Multi-PA: A Multi-perspective Benchmark on Privacy Assessment for Large Vision-Language Models

Inference-Aware Fine-Tuning for Best-of-N Sampling in Large Language Models

CoxKAN: Kolmogorov-Arnold Networks for Interpretable, High-Performance Survival Analysis

A Gray-box Attack against Latent Diffusion Model-based Image Editing by Posterior Collapse

SOAP: Enhancing Spatio-Temporal Relation and Motion Information Capturing for Few-Shot Action Recognition

Data-Driven Lipschitz Continuity: A Cost-Effective Approach to Improve Adversarial Robustness

Safety Control of Service Robots with LLMs and Embodied Knowledge Graphs

Data Valuation by Fusing Global and Local Statistical Information

Natural Strategic Ability in Stochastic Multi-Agent Systems

Dual-Balancing for Multi-Task Learning

ProtoPFormer: Concentrating on Prototypical Parts in Vision Transformers for Interpretable Image Recognition

PaTAS: A Parallel System for Trust Propagation in Neural Networks Using Subjective Logic

FRAGMENTA: End-to-end Fragmentation-based Generative Model with Agentic Tuning for Drug Lead Optimization

Universe of Thoughts: Enabling Creative Reasoning with Large Language Models

Failure Modes in LLM Systems: A System-Level Taxonomy for Reliable AI Applications

Simulated Self-Assessment in Large Language Models: A Psychometric Approach to AI Self-Efficacy

KRAL: Knowledge and Reasoning Augmented Learning for LLM-assisted Clinical Antimicrobial Therapy

Step-Audio-R1 Technical Report

Heterogeneous Multi-Agent Proximal Policy Optimization for Power Distribution System Restoration

Operationalizing Pluralistic Values in Large Language Model Alignment Reveals Trade-offs in Safety, Inclusivity, and Model Behavior

Augur: Modeling Covariate Causal Associations in Time Series via Large Language Models

CoMind: Towards Community-Driven Agents for Machine Learning Engineering

Safe and Economical UAV Trajectory Planning in Low-Altitude Airspace: A Hybrid DRL-LLM Approach with Compliance Awareness

Learning Individual Behavior in Agent-Based Models with Graph Diffusion Networks

Co-PatcheR: Collaborative Software Patching with Component(s)-specific Small Reasoning Models

Earth Observation Satellite Scheduling with Graph Neural Networks and Monte Carlo Tree Search

Revisiting Generalization Across Difficulty Levels: It's Not So Easy

ToolOrchestra: Elevating Intelligence via Efficient Model and Tool Orchestration

G$^2$VLM: Geometry Grounded Vision Language Model with Unified 3D Reconstruction and Spatial Reasoning

Matrix: Peer-to-Peer Multi-Agent Synthetic Data Generation Framework

Through the telecom lens: Are all training samples important?

Escaping the Verifier: Learning to Reason via Demonstrations

Attention-Guided Patch-Wise Sparse Adversarial Attacks on Vision-Language-Action Models

Continual Error Correction on Low-Resource Devices

Escaping the Verifier: Learning to Reason via Demonstrations

Created by

Haebom

Category

Empty

저자

Locke Cai, Ivan Provilkov

개요

본 논문은 task-specific verifier 없이 전문가 시연만으로 강력한 추론 능력을 학습하는 방법을 제시한다. RARO (Relativistic Adversarial Reasoning Optimization)는 정책(생성자)과 상대주의적 비평가(판별자) 간의 적대적 상호 작용을 통해 Inverse Reinforcement Learning을 사용한다. 정책은 전문가의 답변을 모방하고, 비평가는 정책과 전문가 답변을 비교 및 구별하도록 학습된다. RARO는 정책과 비평가를 RL을 통해 지속적으로 함께 훈련시키며, 안정적인 학습을 위한 주요 기술을 식별한다.

시사점, 한계점

•

시사점:

◦

전문가 시연만으로도 강력한 추론 성능을 달성할 수 있음을 입증.

◦

task-specific verifier가 없는 경우에도 효과적인 추론 학습을 가능하게 함.

◦

Countdown, DeepMath, Poetry Writing과 같은 다양한 평가 과제에서 기존의 verifier-free baseline보다 우수한 성능을 보임.

◦

verifier 기반 RL과 유사한 견고한 scaling 추세를 보임.

•

한계점:

◦

논문에 구체적인 한계점에 대한 언급은 없음.

Made with Slashpage