Daily Arxiv

전 세계에서 발간되는 인공지능 관련 논문을 정리하는 페이지 입니다.
본 페이지는 Google Gemini를 활용해 요약 정리하며, 비영리로 운영 됩니다.
논문에 대한 저작권은 저자 및 해당 기관에 있으며, 요약본 공유 시 출처만 명기하면 됩니다.
This service is supported by Google Gemini.

Task Abstention for Large Language Models in Code Generation

PARALLAX: Separating Genuine Hallucination Detection from Benchmark Construction Artifacts

When Dynamics Shift, Robust Task Inference Wins: Offline Imitation Learning with Behavior Foundation Models Revisited

Algorithmic Cultivation: How Social Media Feeds Shape User Language

The IsalProgram Programming Language

Learning-Zone Energy: Online Data Selection for Efficient RL Post-Training

BoLT: A Benchmark to Democratize Black-box Optimization Research for Expensive LLM Tasks

Adversarial Fragility and Language Vulnerability in Clinical AI: A Systematic Audit of Diagnostic Collapse Under Imperceptible Perturbations and Cross-Lingual Drift in Low-Resource Healthcare Settings

Response-free item difficulty modelling for multiple-choice items with fine-tuned transformers: Component-wise representation and multi-task learning

Skills on the Fly: Test-Time Adaptive Skill Synthesis for LLM Agents

Extending Pretrained 10-Second ECG Foundation Models to Longer Horizons

WhiteTesseract: Reframing the Interpretation of Cultural Heritage through XR and Conversational AI

OmniVL-Guard Pro: A Tool-Augmented Agent for Omnibus Vision-Language Forensics

Latent Action Control for Reasoning-Guided Unified Image Generation

Effort as Ceiling, Not Dial: Reasoning Budget Does Not Modulate Cognitive Cost Alignment Between Humans and Large Reasoning Models

Full Attention Strikes Back: Transferring Full Attention into Sparse within Hundred Training Steps

The Alpha Illusion: Reported Alpha from LLM Trading Agents Should Not Be Treated as Deployment Evidence

DriveSafe: A Framework for Risk Detection and Safety Suggestions in Driving Scenarios

Some[Body] Must Receive That Pain for Agent Accountability

Metric-Guided Feature Fusion of Visual Foundation Models for Segmentation Tasks

Plan First, Diffuse Later: Extrinsic Graph Guidance for Long-Horizon Diffusion Planning

Prefix-Adaptive Block Diffusion for Efficient Document Recognition

PhysioSeq2Seq: A Hybrid Physiological Digital Twin and Sequence-to-Sequence LSTM for Long-Horizon Glucose Forecasting in Type 1 Diabetes

VGGT-CD: Training-Free Robust Registration for 3D Change Detection

Pedestrian-Aware LLM-Driven Behavioral Planning for Autonomous Vehicles

Thinking with Patterns: Breaking the Perceptual Bottleneck in Visual Planning via Pattern Induction

Learning Relative Representations for Fine-Grained Multimodal Alignment with Limited Data

Prediction-Intervention Games and Invariant Sets

Decoupling KL and Trajectories: A Unified Perspective for SFT, DAgger, Offline RL, and OPD in LLM Distillation

Echoes in Filter Bubble: Diagnosing and Curing Popularity Bias in Generative Recommenders

AgentKernelArena: Generalization-Aware Benchmarking of GPU Kernel Optimization Agents

Observation-Aligned Mask Priors for Learning Physical Dynamics from Authentic Occlusions

Cross-modal Affinity-aligned Multimodal Learning Analytics for Predicting Student Collaboration Satisfaction in Game-Based Learning

Cross-Domain Molecular Relational Learning: Leveraging Chemical Structure-Activity Analysis

3DPhysVideo: Consistency-Guided Flow SDE for Video Generation via 3D Scene Reconstruction and Physical Simulation

TIER: Trajectory-Invariant Execution Rewards for Multi-Step Tool Composition

Encoding Robust Topological Signatures for Hyperdimensional Computing

A Holistic Method for Superquadric Fitting Using Unsupervised Clustering Analysis

Distinguishable Deletion: Unifying Knowledge Erasure and Refusal for Large Language Model Unlearning

VolTA-3D: Self-Supervised Learning for Brain MRI using 3D Volumetric Token Alignment

CANSURF: An ASV-View Can Dataset and Benchmark for Detection and Tracking of Surface-Level Debris

Exploring Lightweight Large Language Models for Court View Generation

Learning Unbiased Permutations via Flow Matching

UniER: A Unified Benchmark for Item-level and Path-level Exercise Recommendation

Genflow Ad Studio: A Compound AI Architecture for Brand-Aligned, Self-Correcting Video Generation

EmoMind: Decoding Affective Captions from Human Brain fMRI

Universal Dynamics of Punctuated Progress

MAVEN A Multi-Agent Framework for Multicultural Text-to-Video Generation

GeoWorld-VLM: Geometry from World Models for Vision-Language Models

EfficientTDMPC: Improved MPC Objectives for Sample-Efficient Continuous Control

CHI-Bench: Can AI Agents Automate End-to-End, Long-Horizon, Policy-Rich Healthcare Workflows?

Multi-Object Tracking Consistently Improves Wildlife Inference

GraViti: Graph-Level Variational Autoencoders with Relaxed Permutation Invariance

A Scalable Tool for Measuring Manner and Result Verbs in Developmental Language Research

SKG-Eval: Stateful Evaluation of Multi-Turn Dialogue via Incremental Semantic Knowledge Graphs

Learning How to Cube

\textsc{PrivScope}: Task-scoped Disclosure Control for Hybrid Agentic Systems

SLEIGHT-Bench: A Benchmark of Evasion Attacks Against Agent Monitors

To Trust or Not to Trust: Authors' Response to AI-based Reviews

PromptDecipher: Supporting AI Tutor Authoring Through Editable Simulated Interactions

Why Modeling Human Haptic Material Perception with AI Is Difficult

Where Pretraining writes and Alignment reads: the asymmetry of Transformer weight space

GRASP: Graph Agentic Search over Propositions for Multi-hop Question Answering

How Few-Shot Examples Add Up: A Causal Decomposition of Function Vectors in In-Context Learning

Voice ''Cloning'' is Style Transfer

Wavelet Flow Matching for Multi-Scale Physics Emulation

Isotonic Survival Regression: Calibrated Survival Distributions from Deep Cox Models

Automatic Unsupervised Ensemble Outlier Model Selection--Extended Version

Symphony for Speech-to-Text: Supporting Real-Time Medical Voice Interfaces

RAPT: Retrieval-Augmented Post-hoc Thresholding for Multi-Label Classification

Inventorship in AI-Assisted Inventions: Designing an Experiment to Shape Case Law

Hypergraph Pattern Machine: Compositional Tokenization for Higher-Order Interactions

Toward Template-Free Explainability for Monte Carlo Tree Search

Alignment Drift in Long-Term Human-LLM Interaction: A Mechanism-Oriented Framework

No Plan, Yet Human: A Reactive Robotics Model Predicts Human Planning Failures on a Clinical Task

The Scaling Laws of Skills in LLM Agent Systems

Visual Agentic Memory: Enabling Online Long Video Understanding via Online Indexing, Hierarchical Memory, and Agentic Retrieval

MoleCode unlocks structural intelligence in large language models

Policy-Grounded Dynamic Facet Suggestions for Job Search

LERA: LLM-Enhanced RAG for Ad Auction in Generative Chatbots

Strategic Over-Parameterization for Generalizable Low-Rank Adaptation

Mechanistically Interpretable Neural Encoding Reveals Fine-Grained Functional Selectivity in Human Visual Cortex

MHMamba: Multi-Head Mamba for 3D Brain Tumor Segmentation

Asking Back: Interaction-Layer Antidistillation Watermarks

Conservative AI for Safety-Sensitive Medical Image Restoration: Residual-Bounded CT-CTA Enhancement for Intracranial Aneurysm-Relevant Signal Recovery

Identifiable Token Correspondence for World Models

Peak-Detector: Explainable Peak Detection via Instruction-Tuned Large Language Models in Physiological Sign

Physics-Guided Geometric Diffusion for Macro Placement Generation

PESD-TSF: A Period-Aware and Explicit Structured Decomposition Framework for Long-Term Time Series Forecasting

Nested Spatio-Temporal Time Series Forecasting

Avoiding Structural Failure Modes in Tabular Fair SSL: Online Primal-Dual Allocation under Confidence Gating

Membership Inference Attacks on Discrete Diffusion Language Models

Diffusion Attention Expert Model for Predicting and Semi-automatic Localizing STAS in Lung Cancer Histopathological Images

Two-Valued Symmetric Circulant Matrices: Applications in Deep Learning

Hierarchical Two-Stage Framework for Environment-Aware Long-Horizon Vessel Trajectory Prediction

DeepArrhythmia: Segment-Contextualized ECG Arrhythmia Classification via Selective Evidence Acquisition

Semantic Smoothing via Novel View Synthesis for Robust SAR Image Classification

KVCapsule: Efficient Sequential KV Cache Compression for Vision-Language Models with Asymmetric Redundancy

Byzantine-Resilient Federated Learning via QUBO-Based Client Selection on Quantum Annealers

The End of Trust: How Agentic AI Breaks Security Assumptions

MPS-Prover: Advancing Stepwise Theorem Proving by Multi-Perspective Search and Data Curation

Created by

Haebom

저자

Zhenwen Liang, Linfeng Song, Yang Li, Tao Yang, Feng Zhang, Haitao Mi, Dong Yu

개요

본 논문은 자동 정리 증명(ATP)에서 대규모 언어 모델(LLM)의 한계를 극복하기 위해 다중 관점 탐색 증명기(MPS-Prover)를 제안한다. MPS-Prover는 훈련 데이터의 40%를 제거하는 데이터 정제 전략과 학습된 평가 모델과 휴리스틱 규칙을 통합한 다중 관점 트리 탐색 메커니즘을 통해 효율적인 정리 증명을 수행한다. miniF2F와 ProofNet 벤치마크에서 기존 7B 파라미터 모델들을 능가하는 성능을 보이며, 더 짧고 다양한 증명을 생성하는 것을 실험적으로 보였다.