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

May 12

Submitted by

amphora

Soohak: A Mathematician-Curated Benchmark for Evaluating Research-level Math Capabilities of LLMs

EleutherAI

Submitted by

lhjiang

Qwen-Image-2.0 Technical Report

Qwen

Submitted by

unclegeorge

TMAS: Scaling Test-Time Compute via Multi-Agent Synergy

IQuestLab

Submitted by

taesiri

CollabVR: Collaborative Video Reasoning with Vision-Language and Video Generation Models

kaist-ai

Submitted by

chengtan9907

PaperFit: Vision-in-the-Loop Typesetting Optimization for Scientific Documents

opendatalab

Submitted by

wyy-code

Model Merging Scaling Laws in Large Language Models

PolyUHK

The Hong Kong Polytechnic University

Submitted by

wyy-code

Geometry Conflict: Explaining and Controlling Forgetting in LLM Continual Post-Training

PolyUHK

The Hong Kong Polytechnic University

Submitted by

wukeming11

WorldReasonBench: Human-Aligned Stress Testing of Video Generators as Future World-State Predictors

UniX-Lab

Submitted by

Juanxi

Auto-Rubric as Reward: From Implicit Preferences to Explicit Multimodal Generative Criteria

OpenEnvisionLab

Submitted by

eggplant95

X-OmniClaw Technical Report: A Unified Mobile Agent for Multimodal Understanding and Interaction

OppoAI

Submitted by

fvmassoli

Memory-Efficient Looped Transformer: Decoupling Compute from Memory in Looped Language Models

qualcomm

Submitted by

dd12345789

SEIF: Self-Evolving Reinforcement Learning for Instruction Following

Fudan-University

Fudan University

Submitted by

shenjunhao

Dynamic Skill Lifecycle Management for Agentic Reinforcement Learning

CUHKDBGROUP

The Chinese University of Hong Kong - Database Group

Submitted by

thuzhaowang

Pixal3D: Pixel-Aligned 3D Generation from Images

TencentARC

ARC Lab, Tencent PCG

Submitted by

beanie00

Rebellious Student: Reversing Teacher Signals for Reasoning Exploration with Self-Distilled RLVR

MicrosoftResearch

Microsoft Research

Submitted by

ChengsongHuang

G-Zero: Self-Play for Open-Ended Generation from Zero Data

·
10 authors

Submitted by

ngocbh

Make Each Token Count: Towards Improving Long-Context Performance with KV Cache Eviction

YaleUniversity

Yale University

Submitted by

Blaze7451

Mela: Test-Time Memory Consolidation based on Transformation Hypothesis

MusubiAI

Submitted by

Shengkun

SlimQwen: Exploring the Pruning and Distillation in Large MoE Model Pre-training

·
10 authors

Submitted by

frankzydou

RigidFormer: Learning Rigid Dynamics using Transformers

MIT

Massachusetts Institute of Technology

Submitted by

astrlrd

SlimSpec: Low-Rank Draft LM-Head for Accelerated Speculative Decoding

nebius

Submitted by

visity

Shaping Schema via Language Representation as the Next Frontier for LLM Intelligence Expanding

·
6 authors

Submitted by

Yeongtak

Omni-Persona: Systematic Benchmarking and Improving Omnimodal Personalization

SeoulNatlUniv

Seoul National University

Submitted by

lr10260

Reinforcing Multimodal Reasoning Against Visual Degradation

Tencent-Hunyuan

Tencent Hunyuan

Submitted by

lr10260

DeltaRubric: Generative Multimodal Reward Modeling via Joint Planning and Verification

Tencent-Hunyuan

Tencent Hunyuan

1

Submitted by

taesiri

NanoResearch: Co-Evolving Skills, Memory, and Policy for Personalized Research Automation

·
15 authors

Submitted by

Yirany

LLaVA-UHD v4: What Makes Efficient Visual Encoding in MLLMs?

·
6 authors

Submitted by

vicgalle

Metal-Sci: A Scientific Compute Benchmark for Evolutionary LLM Kernel Search on Apple Silicon

·
1 authors

Submitted by

ZBox008003

AgentForesight: Online Auditing for Early Failure Prediction in Multi-Agent Systems

RutgersU

Rutgers University

Submitted by

zhenwang9102

FlashEvolve: Accelerating Agent Self-Evolution with Asynchronous Stage Orchestration

UCSanDiego

University of California at San Diego

1

Submitted by

Raincleared

DECO: Sparse Mixture-of-Experts with Dense-Comparable Performance on End-Side Devices

TsinghuaNLP

Tsinghua NLP Group

Submitted by

ChanganYao

Injecting Distributional Awareness into MLLMs via Reinforcement Learning for Deep Imbalanced Regression

·
3 authors

1

Submitted by

yihongLiu

Crosslingual On-Policy Self-Distillation for Multilingual Reasoning

cis-lmu

CIS, LMU Munich

Submitted by

linzheng

Scratchpad Patching: Decoupling Compute from Patch Size in Byte-Level Language Models

·
6 authors

Submitted by

Franck-Dernoncourt

FORTIS: Benchmarking Over-Privilege in Agent Skills

·
11 authors

Submitted by

eehaojiezhang

MuSS: A Large-Scale Dataset and Cinematic Narrative Benchmark for Multi-Shot Subject-to-Video Generation

·
8 authors

Submitted by

lballore

LLiMba: Sardinian on a Single GPU -- Adapting a 3B Language Model to a Vanishing Romance Language

·
1 authors

Submitted by

XinwuYe

Pushing Biomolecular Utility-Diversity Frontiers with Supergroup Relative Policy Optimization

Submitted by

cjerzak

Queryable LoRA: Instruction-Regularized Routing Over Shared Low-Rank Update Atoms

JerzakLabs

1

Submitted by

Jeasco

SplatWeaver: Learning to Allocate Gaussian Primitives for Generalizable Novel View Synthesis

hitcdl-edu

Harbin Institute of Technology

Submitted by

Acruxos

Uncovering Entity Identity Confusion in Multimodal Knowledge Editing

CASIA

Chinese Academic of Science Institute of Automation

1

Submitted by

Jessogreat

Can Muon Fine-tune Adam-Pretrained Models?

MBZUAI

Mohamed Bin Zayed University of Artificial Intelligence

Submitted by

Wenxuan123

CapVector: Learning Transferable Capability Vectors in Parametric Space for Vision-Language-Action Models

OpenHelix-Team

Submitted by

Wenxuan123

RoboMemArena: A Comprehensive and Challenging Robotic Memory Benchmark

·
13 authors

Submitted by

yermandy

The Alpha Blending Hypothesis: Compositing Shortcut in Deepfake Detection

·
4 authors

Submitted by

taesiri

Shepherd: A Runtime Substrate Empowering Meta-Agents with a Formalized Execution Trace

·
7 authors

Submitted by

k-zhai

Dystruct: Dynamically Structured Diffusion Language Model Decoding via Bayesian Inference

·
4 authors

1

Submitted by

taesiri

SimWorld Studio: Automatic Environment Generation with Evolving Coding Agent for Embodied Agent Learning

·
8 authors

Submitted by

kaizhao

Sub-JEPA: Subspace Gaussian Regularization for Stable End-to-End World Models

shanghai

shanghai university

Submitted by

pranamanam

TD3B: Transition-Directed Discrete Diffusion for Allosteric Binder Generation

ChatterjeeLab

Programmable Biology Group

Submitted by

yeshpanovrustem

100,000+ Movie Reviews from Kazakhstan: Russian, Kazakh, and Code-Switched Texts

·
1 authors

1

Submitted by

tangzhy

Safe, or Simply Incapable? Rethinking Safety Evaluation for Phone-Use Agents

·
21 authors

Submitted by

dqj5182

Training-Free Dense Hand Contact Estimation with Multi-Modal Large Language Models

SeoulNatlUniv

Seoul National University