Papers

Representational separation between unitary and channel quantum generative models via shared classical randomness at shallow depth

Reward Structure Shapes the Interaction Between Episodic Exploration and Neural Memory in Reinforcement Learning

Stable Density Ridges: Consistency and Convergence of Subspace Constrained Mean Shift

Robust and Efficient Motion Reasoning for Privacy-Aware Classroom Incident Recognition

ABSeeker: Training Long-Horizon Search Agents via Answer-Backtracked Credit Assignment

PG-LLM: Benchmarking General-Purpose Language Models for Protein Variant Ranking

DataSpace: Benchmarking Data Agents for Verifiable Analytics over Heterogeneous Workspaces

REUSING ROLLOUTS UNDER POLICY LAG: PREFIX-NORMALIZED POLICY OPTIMIZATION FOR LLM REIN-FORCEMENT LEARNING

Harness-R1: Learning to Edit Executable Runtime Harnesses from Agent Failure Trajectories

FinanceHarness: Autonomous Financial Deep Research Framework

DeepSeek-V4: Towards Highly Efficient Million-Token Context Intelligence

Knowledge–Geometry Decoupling: Refreshable Pretrained Transfer for Streaming Recommendation

Video-DeepResearch: Towards the Next-Generation Multimodal Deepresearch Agent

AURORA-LM: Autoencoding Unified Representation for Continuous-Latent Diffusion Language Modeling

Hunyuan3D-Buffalo 1.0: A Unified Multimodal Model for Scalable 3D Generation, Understanding, and Editing

JoyAI-Video-Edit: Real-Time Open-Ended Video Editing with Autoregressive Diffusion

MerchantBench: Benchmarking LLM Agents for Long-Term Coherence in E-Commerce Operations

TRAINING nGPT

UEmbed: Unified Sparse and Dense Multimodal Embeddings

VAD: Attributing Visual Evidence for Target Reconstruction in Multimodal On-Policy Distillation

PROGRESSIVE AGENT SKILL GENERATION VIA REINFORCEMENT LEARNING

DAPD: Dual-Anchored Policy Distillation

LongHorizon-Harness: Advancing Long-Horizon Agents for Real-World Tasks

SwanTale: Unified Multi-Speaker Speech and Audio Generation for Instruct and Zero-Shot Tasks

Fara-1.5: Scalable Learning Environments for Computer Use Agents

Docling Technical Report

From high-throughput evaluation to wet-lab studies: advancing mutation effect prediction with a retrieval-enhanced model

Model or Harness? An Interaction-Centric Taxonomy for Localizing Agent Failures

Inducing language models to assert their own consciousness restores human beliefs and values

Diamond: A Sequence-to-Sequence Model for Speech Restoration via an Autoregressive RQ-Transformer over Neural Audio Codec Tokens