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

Jan 29

Submitted by

YanqiDai

Harder Is Better: Boosting Mathematical Reasoning via Difficulty-Aware GRPO and Multi-Aspect Question Reformulation

GD-ML

Submitted by

qiuyuu

Advancing Open-source World Models

robbyant

Submitted by

zichenwen

Innovator-VL: A Multimodal Large Language Model for Scientific Discovery

SJTU

Shanghai Jiao Tong University

Submitted by

taesiri

DeepSeek-OCR 2: Visual Causal Flow

deepseek-ai

Submitted by

Jinyang23

Spark: Strategic Policy-Aware Exploration via Dynamic Branching for Long-Horizon Agentic Learning

·
7 authors

1

Submitted by

taesiri

Linear representations in language models can change dramatically over a conversation

google

Submitted by

jonhue

Reinforcement Learning via Self-Distillation

lasgroup

LAS @ ETH Zurich

Submitted by

optimusPrimeBee

VERGE: Formal Refinement and Guidance Engine for Verifiable LLM Reasoning

amazonwebservices

Amazon Web Services (AWS)

1

Submitted by

taesiri

SERA: Soft-Verified Efficient Repository Agents

ai21labs

AI21

Submitted by

taesiri

OmegaUse: Building a General-Purpose GUI Agent for Autonomous Task Execution

·
15 authors

Submitted by

kishanpb

Group Distributionally Robust Optimization-Driven Reinforcement Learning for LLM Reasoning

tencent

1

Submitted by

mandipgoswami

RIR-Mega-Speech: A Reverberant Speech Corpus with Comprehensive Acoustic Metadata and Reproducible Evaluation

·
1 authors

Submitted by

nielsr

UPLiFT: Efficient Pixel-Dense Feature Upsampling with Local Attenders

UPLiFT-upsampler

UPLiFT - University of Maryland, College Park

Submitted by

guactastesgood

Training Reasoning Models on Saturated Problems via Failure-Prefix Conditioning

newyorkuniversity

New York University

Submitted by

THEATLAS

GDCNet: Generative Discrepancy Comparison Network for Multimodal Sarcasm Detection

ict-cas

Institute of Computing Technology, Chinese Academy of Sciences

Submitted by

Lakoc

SE-DiCoW: Self-Enrolled Diarization-Conditioned Whisper

BUT-FIT

Brno University of Technology, Faculty of Information Technology

Submitted by

nfel

Persona Prompting as a Lens on LLM Social Reasoning

xplainlp

Submitted by

taesiri

SketchDynamics: Exploring Free-Form Sketches for Dynamic Intent Expression in Animation Generation

·
4 authors

Submitted by

bf-jeon

Shallow-π: Knowledge Distillation for Flow-based VLAs

SamsungResearch

Samsung Research

Submitted by

cfahlgren1

How AI Impacts Skill Formation

Anthropic