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

Apr 21

Submitted by

xiaochonglinghu

Extending One-Step Image Generation from Class Labels to Text via Discriminative Text Representation

GD-ML

Submitted by

ldkong

OneVL: One-Step Latent Reasoning and Planning with Vision-Language Explanation

xiaomi-research

Xiaomi Research

Submitted by

dongguanting

Agent-World: Scaling Real-World Environment Synthesis for Evolving General Agent Intelligence

ByteDance-Seed

3

Submitted by

csuhan

OpenGame: Open Agentic Coding for Games

·
11 authors

Submitted by

Haoyuwu

MultiWorld: Scalable Multi-Agent Multi-View Video World Models

hkuhk

The University of Hong Kong

Submitted by

iieycx

EasyVideoR1: Easier RL for Video Understanding

·
9 authors

Submitted by

zwq2018

GFT: From Imitation to Reward Fine-Tuning with Unbiased Group Advantages and Dynamic Coefficient Rectification

OmniAI-ZJU

Submitted by

salmannyu

When Can LLMs Learn to Reason with Weak Supervision?

·
6 authors

Submitted by

taesiri

WebCompass: Towards Multimodal Web Coding Evaluation for Code Language Models

·
19 authors

Submitted by

taesiri

ClawEnvKit: Automatic Environment Generation for Claw-Like Agents

umd-zhou-lab

Submitted by

YuZeng260

SkillFlow:Benchmarking Lifelong Skill Discovery and Evolution for Autonomous Agents

·
16 authors

Submitted by

yixuantt

Crowded in B-Space: Calibrating Shared Directions for LoRA Merging

·
2 authors

Submitted by

zhangjiaxin2012

The Illusion of Certainty: Decoupling Capability and Calibration in On-Policy Distillation

Salesforce

Salesforce AI Research

Submitted by

eunwooim

Concrete Jungle: Towards Concreteness Paved Contrastive Negative Mining for Compositional Understanding

·
3 authors

2

Submitted by

chalo2000

On the Reliability of Computer Use Agents

simular-ai

Submitted by

taesiri

MathNet: a Global Multimodal Benchmark for Mathematical Reasoning and Retrieval

MIT

Massachusetts Institute of Technology

Submitted by

liangjiaqing

GenericAgent: A Token-Efficient Self-Evolving LLM Agent via Contextual Information Density Maximization (V1.0)

Fudan-University

Fudan University

Submitted by

leungtianle

VoxMind: An End-to-End Agentic Spoken Dialogue System

·
10 authors

Submitted by

Jevin754

OmniScript: Towards Audio-Visual Script Generation for Long-Form Cinematic Video

TencentARC

ARC Lab, Tencent PCG

Submitted by

lenglaender

Agents Explore but Agents Ignore: LLMs Lack Environmental Curiosity

Cohere

2

Submitted by

taesiri

Training LLM Agents for Spontaneous, Reward-Free Self-Evolution via World Knowledge Exploration

·
8 authors

Submitted by

xiachongfeng

Stratagem: Learning Transferable Reasoning via Trajectory-Modulated Game Self-Play

hkuhk

The University of Hong Kong

Submitted by

aluo-x

Meta-learning In-Context Enables Training-Free Cross Subject Brain Decoding

hkuhk

The University of Hong Kong

Submitted by

cjerzak

Multiplication in Multimodal LLMs: Computation with Text, Image, and Audio Inputs

JerzakLabs

2

Submitted by

jacklanda

Revisiting a Pain in the Neck: A Semantic Reasoning Benchmark for Language Models

ustb-edu-cn

University of Science and Technology Beijing

Submitted by

doctorcoal

River-LLM: Large Language Model Seamless Exit Based on KV Share

SJTU

Shanghai Jiao Tong University

4

Submitted by

Bill1235813

Precise Debugging Benchmark: Is Your Model Debugging or Regenerating?

USC

University of Southern California

Submitted by

Bowieee

Beyond Text-Dominance: Understanding Modality Preference of Omni-modal Large Language Models

·
7 authors

Submitted by

harvardairobotics

MedConclusion: A Benchmark for Biomedical Conclusion Generation from Structured Abstracts

Harvard

Harvard University

Submitted by

gabTriv

MARCO: Navigating the Unseen Space of Semantic Correspondence

PolitecnicoDiTorino

Politecnico di Torino

Submitted by

stillwithyou

Latent Preference Modeling for Cross-Session Personalized Tool Calling

HYU-NLP

2

Submitted by

amazingj

Modeling Multiple Support Strategies within a Single Turn for Emotional Support Conversations

DianJin

2

Submitted by

pcr2120

The Geometric Canary: Predicting Steerability and Detecting Drift via Representational Stability

·
1 authors

Submitted by

Agcs12

When Background Matters: Breaking Medical Vision Language Models by Transferable Attack

MBZUAI

Mohamed Bin Zayed University of Artificial Intelligence

Submitted by

taesiri

EvoMaster: A Foundational Agent Framework for Building Evolving Autonomous Scientific Agents at Scale

·
23 authors

Submitted by

liboaccn

MNAFT: modality neuron-aware fine-tuning of multimodal large language models for image translation

·
6 authors

2

Submitted by

cedricbonhomme

Modeling Sparse and Bursty Vulnerability Sightings: Forecasting Under Data Constraints

CIRCL

Computer Incident Response Center Luxembourg

Submitted by

pcr2120

Geometric coherence of single-cell CRISPR perturbations reveals regulatory architecture and predicts cellular stress

·
1 authors

Submitted by

Jeff0918

MTR-DuplexBench: Towards a Comprehensive Evaluation of Multi-Round Conversations for Full-Duplex Speech Language Models

·
8 authors

Submitted by

Satyamk098

Forge-UGC: FX optimization and register-graph engine for universal graph compiler

·
2 authors

Submitted by

ibercovich

Terminal Wrench: A Dataset of 331 Reward-Hackable Environments and 3,632 Exploit Trajectories

few-sh

Submitted by

SamuelTanguturi

The Continuity Layer: Why Intelligence Needs an Architecture for What It Carries Forward

Kenotic-Labs

Submitted by

kadiroezer

Back to Repair: A Minimal Denoising Network\ for Time Series Anomaly Detection

Mercedes-Benz

Mercedes-Benz AG

Submitted by

SteveZeyuZhang

HSG: Hyperbolic Scene Graph

PekingUniversity

Peking University

Submitted by

liyongkang

On the Robustness of LLM-Based Dense Retrievers: A Systematic Analysis of Generalizability and Stability

uva-irlab

IRLab, University Of Amsterdam

Submitted by

ankitclio

KWBench: Measuring Unprompted Problem Recognition in Knowledge Work

clio-ai

Submitted by

hyn0027D

Symbolic Guardrails for Domain-Specific Agents: Stronger Safety and Security Guarantees Without Sacrificing Utility

CarnegieMellonU

Carnegie Mellon University

Submitted by

xXiaobuding

Protecting Language Models Against Unauthorized Distillation through Trace Rewriting

·
4 authors

Submitted by

Q-M-E

Significance and Stability Analysis of Gene-Environment Interaction using RGxEStat

HenanUniversity

Henan University