new

Get trending papers in your email inbox once a day!

Get trending papers in your email inbox!

Daily Papers

byAK and the research community

Jul 3

Submitted by

yuntian-deng

Program-as-Weights: A Programming Paradigm for Fuzzy Functions

UWaterloo

University of Waterloo

Submitted by

MikaStars39

EvoPolicyGym: Evaluating Autonomous Policy Evolution in Interactive Environments

·
16 authors

Submitted by

kpzhang996

AgenticSTS: A Bounded-Memory Testbed for Long-Horizon LLM Agents

AlayaLab

Submitted by

landisen

Morphing into Hybrid Attention Models

ByteDance-Seed

Submitted by

curtis-sun

AgenticDataBench: A Comprehensive Benchmark for Data Agents

Tsinghua University

Submitted by

Xingyu-Zheng

Multi-Resolution Flow Matching: Training-Free Diffusion Acceleration via Staged Sampling

·
7 authors

Submitted by

taesiri

WorldDirector: Building Controllable World Simulators with Persistent Dynamic Memory

·
13 authors

Submitted by

Minbyul

Breaking Failure Cascades: Step-Aware Reinforcement Learning for Medical Multimodal Reasoning

·
8 authors

Submitted by

taesiri

SkillCoach: Self-Evolving Rubrics for Evaluating and Enhancing Agentic Skill-Use

·
7 authors

Submitted by

rhli

Optimizing Visual Generative Models via Distribution-wise Rewards

Tencent-Hunyuan

Tencent Hunyuan

Submitted by

pminervini

Logit-Contribution Scoring Identifies Non-Literal Retrieval Heads

·
3 authors

Submitted by

DyJiang

From SRA to Self-Flow: Data Augmentation or Self-Supervision?

·
4 authors

Submitted by

Xin1118

AGVBench: A Reliability-Oriented Benchmark of Data Augmentation for Vein Recognition

CTBU-University

Chongqing Technology and Business University

Submitted by

xiaoyu1104

InstanceControl: Controllable Complex Image Generation without Instance Labeling

·
7 authors

Submitted by

rhachiuma

AnyGroundBench: A Specialized-Domain Benchmark for Video Grounding in Vision-Language Models

·
9 authors

Submitted by

zzh068

When Search Agents Should Ask: DiscoBench for Clarification-Aware Deep Search

Tencent-Hunyuan

Tencent Hunyuan

1

Submitted by

danielwusg

AutoMem: Automated Learning of Memory as a Cognitive Skill

·
5 authors

Submitted by

taesiri

PACE: A Proxy for Agentic Capability Evaluation

·
11 authors

Submitted by

taesiri

Representation Distribution Matching for One-Step Visual Generation

·
5 authors

Submitted by

taesiri

Learning to Move Before Learning to Do: Task-Agnostic pretraining for VLAs

·
6 authors

Submitted by

Moenupa

Denser neq Better: Limits of On-Policy Self-Distillation for Continual Post-Training

·
10 authors

Submitted by

mxvp

Discrete Diffusion Language Models for Interactive Radiology Report Drafting

gevaertlab

Submitted by

obohdal

DuoMem: Towards Capable On-Device Memory Agents via Dual-Space Distillation

·
9 authors

Submitted by

Dragonjinny

Transferability for General Reasoning: An Automated Curriculum for Multi-Domain RLVR

uoft

University of Toronto

Submitted by

zihengh1

WARP: Weight-Space Analysis for Recovering Training Data Portfolios

sprocket-lab

Submitted by

Jim137

Parameter-Efficient Quantum-Inspired Fast Weight Programmers for Traffic-Matrix Forecasting

·
6 authors

Submitted by

noaflynn

Scaling Laws for Grid-Based Approximate Nearest Neighbor Search in High Dimensions

UCBerkeley

University of California, Berkeley