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 22

Submitted by

taesiri

ABot-World-0: Infinite Interactive World Rollout on a Single Desktop GPU

acvlab

Alibaba AMAP CV Lab

Submitted by

lhpku20010120

DataFlow-Harness: A Grounded Code-Agent Platform for Constructing Editable LLM Data Pipelines

PekingUniversity

Peking University

Submitted by

Met4physics

Text Template Tokens Are Implicit Semantic Registers in Diffusion Transformers

RTP-LLM

Submitted by

wangzx1994

Generative World Renderer at the Speed of Play

AlayaLab

Submitted by

Xinjie-Q

Mage-Flow: An Efficient Native-Resolution Foundation Model for Image Generation and Editing

microsoft

Submitted by

kpzhang996

AlayaWorld: Interactive Long-Horizon World Modeling -- Full Technical Report

AlayaLab

Submitted by

maximorulli

Subliminal Clocks: Latent Time Modelling in Diffusion Language Models

Sapienza

Sapienza University of Rome

2

Submitted by

zli12321

Stale but Stable: Staleness-Adaptive Trust Regions for Stabilizing Asynchronous Reinforcement Learning

Tencent-Hunyuan

Tencent Hunyuan

Submitted by

Leozkl

AgentDebugX: An Open-Source Toolkit for Failure Observability, Attribution, and Recovery in LLM Agents

UIUC-CS

University of Illinois at Urbana-Champaign

Submitted by

LoYuXrqw

SciForma: Structure-Faithful Generation of Scientific Diagrams

PekingUniversity

Peking University

2

Submitted by

ChengCui

HPD-Parsing: Hierarchical Parallel Document Parsing

PaddlePaddle

2

Submitted by

mrdrozdov

AutoIndex: Learning Representation Programs for Retrieval

·
6 authors

Submitted by

ccsasuke

Two-Level Meta-Rubrics for Evaluating Open-Ended Generation: GAMUT, a Benchmark for Factual Completeness

Submitted by

FOGmiaow

ISO: An RLVR-Native Optimization Stack

UTEXAS

University of Texas at Austin

Submitted by

taesiri

H^2SD: Hybrid Hindsight Self-Distillation

·
10 authors

Submitted by

Laurin-myreha

Transcription Policy as a Latent Variable: Activating Controllable Verbatim ASR with Word-Level Timing

nyralabs

Submitted by

taesiri

Masked Visual Actions for Unified World Modeling

·
11 authors

Submitted by

anonnumaan01

Where Should Optimizer State Live? Tiered State Allocation for Memory-Efficient Mixture-of-Experts Training

·
1 authors

Submitted by

research-WM

Trajectory-aware Cross-view Geo-localization with Sequential Observations

MVRL

Multimodal Vision Research Laboratory @ WashU

Submitted by

MykolaL

Delineate Anything v2: A Global Foundation Model for Field Delineation

·
7 authors

2

Submitted by

Snooow1029

EduPanel: A Three-Agent LLM Judge for Teaching Videos -- Reliability, Complementarity, and Human Trust Calibration

nationaltaiwan

Submitted by

mjbuehler

Reading and Steering Representations of Materials-Science Mechanisms in an Open-Weight Language Model

lamm-mit

LAMM: MIT Laboratory for Atomistic and Molecular Mechanics

Submitted by

taesiri

Appearance Pointers -- Multimodal Region Control of Diffusion Transformers

·
5 authors

Submitted by

zhehuderek

Computational Humor with Multimodal LLMs: Methods, Datasets, Evaluation, and Challenges

·
5 authors

Submitted by

Believe0029

ConsiSpace: Learning Geometric Consistency Matters for Video Spatial Reasoning

·
5 authors

4