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

Sep 16

Submitted by

cozzyde

Continual Learning Mechanisms Compose for Long-Horizon Memorization

JohnsHopkins

Johns Hopkins University

Submitted by

Eureka-Leo

AI for Games in the Foundation Model Era

NationalUniversityofSingapore

National University of Singapore

Submitted by

yanchaomars

StepAudio 3 Realtime Technical Report

stepfun-ai

2

Submitted by

yanchaomars

StepAudio 3 Music Technical Report

stepfun-ai

3

Submitted by

thy-1now

Generalized Agent Iteration: One Formal Framework for Iterative Policy Improvement and Recursive Self-Improvement

·
4 authors

2

Submitted by

Lingaaaaaaa

ScienceBuddy: Recursive-in-Recursive Self-Improvement for Interactive Scientific Agents

·
13 authors

Submitted by

chenyang0126

HarnessVLN: Unifying Training-Free Embodied Navigation through an Agent Harness

·
10 authors

2

Submitted by

imguangyu

FLAT: Resampling Image and Text into 1D Flexible-Length Aligned Transmodal Tokens for Retrieval and Generation

meta

2

Submitted by

ShenYunTzr

The Last AI Built by Humans: Toward Genuine Recursive Self-Improvement

·
33 authors

Submitted by

akhadangi

Another Blueprint In The Wall: How to Ask Frontier AI Like a Kid?

·
1 authors

2

Submitted by

SiweiWu

ModularRSI: Modular and Generalizable Recursive Harness Self-Improvement

IQuestLab

Submitted by

taesiri

Modality-Autoregressive World-Action Models

·
4 authors

Submitted by

zhenwang9102

Mind2Dialogue: Training Human-Aware Language Models by Simulating User Mental States

·
19 authors

Submitted by

shwai-he

Disentangling Representation Evolution in Transformers through Directional Decomposition

ByteDance

Submitted by

albertge

Register Tokens for Bounded-State Reasoning in Diffusion Language Models

sprocket-lab

Submitted by

danyaljj

Convergent Emergence of In-Context Learning Across Modalities

jhu-clsp

Center for Language and Speech Processing @ JHU

Submitted by

taesiri

PhysStream: Streaming Physics-Grounded Video Generation with Structured Scene Memory and Fine-Grained Motion Control

·
7 authors

Submitted by

nielsr

RelateAnything: Real-Time Open-Vocabulary Relation Prediction From Any Inputs

·
1 authors

Submitted by

eyuansu71

ImpossibleRubrics: Stress-Testing Generated Rubrics as Reward Signals

·
4 authors

Submitted by

aashiqmuhamed

Decoy Direction Optimization: A Post-Hoc Defense Against LLM Abliteration

CarnegieMellonU

Carnegie Mellon University

Submitted by

taesiri

Emergence World: Adversarial Stress-Testing of Long-Horizon Multi-Agent Systems

·
8 authors

Submitted by

FeYuan

Drift-Constrained Optimization: Only Direction Matters in Fine-Tuning Instruct Models

·
6 authors

Submitted by

FeYuan

Training Specialist Models without Reasoning Trajectories for Domain Expert Distillation

MaLA-LM

Submitted by

crs25-tsinghua

The Router Within: Eliciting Native Skill Routing from a Frozen LLM

Tsinghua University

2

Submitted by

BrandonLiu

OmniHarness: Harnessing Generalizable Visual Generation via Symbolic Policy Learning

·
8 authors

Submitted by

xuange

LimiX-2: A Contextual Mechanism Network Towards General Structured-Data Intelligence

stable-ai