Papers33

#Reinforcement Learning

VAR RL Done Right: Tackling Asynchronous Policy Conflicts in Visual Autoregressive Generation

Visual Autoregressive (VAR) models draw whole grids of image tokens at once across multiple scales, which makes standard reinforcement learning (RL) unstable.

#Visual Autoregressive (VAR)#Reinforcement Learning#GRPO

NextFlow: Unified Sequential Modeling Activates Multimodal Understanding and Generation

Intermediate

Huichao Zhang, Liao Qu et al.Jan 5arXiv

NextFlow is a single, decoder-only Transformer that can read and write both text and images in one continuous sequence.

#Next-Scale Prediction#Autoregressive Transformer#Dual-Codebook Tokenization

MDAgent2: Large Language Model for Code Generation and Knowledge Q&A in Molecular Dynamics

Intermediate

Zhuofan Shi, Hubao A et al.Jan 5arXiv

MDAgent2 is a special helper built from large language models (LLMs) that can both answer questions about molecular dynamics and write runnable LAMMPS simulation code.

#Molecular Dynamics#LAMMPS#Code Generation

K-EXAONE Technical Report

Intermediate

Eunbi Choi, Kibong Choi et al.Jan 5arXiv

K-EXAONE is a super-sized language model that speaks six languages and can read very long documents (up to 256,000 tokens) without forgetting important details.

#Mixture-of-Experts#Hybrid Attention#Sliding Window Attention

CPPO: Contrastive Perception for Vision Language Policy Optimization

Intermediate

Ahmad Rezaei, Mohsen Gholami et al.Jan 1arXiv

CPPO is a new way to fine‑tune vision‑language models so they see pictures more accurately before they start to reason.

#CPPO#Contrastive Perception Loss#Vision-Language Models

MindWatcher: Toward Smarter Multimodal Tool-Integrated Reasoning

Intermediate

Jiawei Chen, Xintian Shen et al.Dec 29arXiv

MindWatcher is a smart AI agent that can think step by step and decide when to use tools like web search, image zooming, and a code calculator to solve tough, multi-step problems.

#Tool-Integrated Reasoning#Interleaved Thinking#Multimodal Chain-of-Thought

NVIDIA Nemotron 3: Efficient and Open Intelligence

Intermediate

NVIDIA, : et al.Dec 24arXiv

Nemotron 3 is a new family of open AI models (Nano, Super, Ultra) built to think better while running faster and cheaper.

#Nemotron 3#Mixture-of-Experts#LatentMoE

Nemotron 3 Nano: Open, Efficient Mixture-of-Experts Hybrid Mamba-Transformer Model for Agentic Reasoning

Intermediate

NVIDIA, : et al.Dec 23arXiv

Nemotron 3 Nano is a new open-source language model that mixes two brain styles (Mamba and Transformer) and adds a team of special experts (MoE) so it thinks better while running much faster.

#Mixture-of-Experts#Mamba-2#Transformer

SpatialTree: How Spatial Abilities Branch Out in MLLMs

Intermediate

Yuxi Xiao, Longfei Li et al.Dec 23arXiv

SpatialTree is a new, four-level "ability tree" that tests how multimodal AI models (that see and read) handle space: from basic seeing to acting in the world.

#Spatial Intelligence#Multimodal Large Language Models#Hierarchical Benchmark

Step-DeepResearch Technical Report

Intermediate

Chen Hu, Haikuo Du et al.Dec 23arXiv

Search is not the same as research; real research needs planning, checking many sources, fixing mistakes, and writing a clear report.

#Deep Research#Atomic Capabilities#ReAct Agent

DiRL: An Efficient Post-Training Framework for Diffusion Language Models

Intermediate

Ying Zhu, Jiaxin Wan et al.Dec 23arXiv

This paper builds DiRL, a fast and careful way to finish training diffusion language models so they reason better.

#Diffusion Language Model#Blockwise dLLM#Post-Training

Multi-hop Reasoning via Early Knowledge Alignment

Intermediate

Yuxin Wang, Shicheng Fang et al.Dec 23arXiv

This paper adds a tiny but powerful step called Early Knowledge Alignment (EKA) to multi-step retrieval systems so the model takes a quick, smart look at relevant information before it starts planning.

#Retrieval-Augmented Generation#Iterative RAG#Multi-hop Reasoning

1 2 3