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

Oct 13

Submitted by

Skywalker0410

GroundingPI: A Grounding Foundation Model towards Physical Intelligence with Visual Primitives

XPENG-AI

Submitted by

DianQiao

Thinking Inertia: LLMs Keep Thinking When Told Not To

StanfordUniversity

Stanford University

Submitted by

TommyChien

AgentSWE: Can Coding Agents Build the Agent You Actually Want?

·
5 authors

Submitted by

seashell11

Scaling Agentic Data for Long-Horizon Terminal Intelligence

alibabagroup

Submitted by

bx6d

VIS-Ground: Video Interactive Storytelling with Contextual Grounding

GoogIe

Submitted by

Skywalker0410

GroundAnything: Reconciling Parallel Decoding with Precise Visual Grounding at Flash Speed

XPENG-AI

Submitted by

galaxy-1

Multi-Stream Attention Residuals

SJTU

Shanghai Jiao Tong University

Submitted by

fyzzzzzz

Relevance Does Not Imply Applicability: Experience Activation for Personal GUI Agents

NanyangTechnologicalUniversity

Nanyang Technological University

Submitted by

Balibata

PR-OPD: Privileged Representation On-policy Self-Distillation for Agentic Reinforcement Learning

·
8 authors

Submitted by

LuHanwen

CrossTimeEdit: A Decade-Spanning Cross-View Dataset and Reward-Guided Editing for Historical Street-View Generation

SunYatsen

Sun Yat-Sen University

Submitted by

YuhongLiu

SpatialASM: Internalizing Active Mentalization for Spatial Reasoning

Shanghai-AI-Laboratory

Shanghai AI Laboratory

Submitted by

anogu

T3lescope: Arbitrary-Resolution High-Fidelity Generative Surface Reconstruction from Images

pfnet

Preferred Networks, Inc.

Submitted by

xyz9911

Systematic Multi-Agent Vision-and-Language Navigation: Formulation, Benchmark, and Method

SJTU

Shanghai Jiao Tong University

Submitted by

yanlinf

ARCS: Towards Precise Text-to-SQL via Structured Disambiguation

megagonlabs

Submitted by

Jackbrocp

Learning More by Training Less: Uncertainty-Guided Progressive Contraction for Efficient Reasoning Post-Training

NJU

Nanjing University

Submitted by

mciccone

Initialization Improves LLM-Driven Discovery

vector-institute

Vector Institute

Submitted by

DongyangChen

Branching On-Policy Distillation

Submitted by

Erfan-Shayegani

Visual Grounding Safety in Vision-Language Models

apple

Submitted by

zhangjiaxin2012

Prospective Hindsight: Self-Calibrating Reinforcement Learning via Prediction-Reality Gaps

Salesforce

Salesforce AI Research

Submitted by

ZachW

Evaluating LLM Creativity through Long-Tail Performance

Uchicago

University of Chicago

Submitted by

ParthKulshreshtha

Same Output, Different Gold: Measuring How Reference Choice Moves a Multilingual Benchmark Score

CentificAIResearch

Centific Applied AI Research

1

Submitted by

RobbinWang123

UNITAS: A 3D-Native World Action Model for Embodied Manipulation

CUHKSZ

Chinese University of Hong Kong, Shenzhen

Submitted by

sj68

SPLATIFY: Reproduce, Discover, Innovate! From Papers and Ideas to Trainable 3DGS Code

UCSanDiego

University of California at San Diego

Submitted by

Jacklu0831

HyperThink: Text-to-Parameter Hypernetworks for Efficient Reasoning

·
5 authors