Opinions
The AI Revolution Threatens a ‘Token Divide’
In a Brookings commentary, Tom Wheeler argues that AI’s rapid capability growth is creating a ‘token divide’ distinct from the…
Engineering
JEPA-Anything: Learning Predictive Models across Different Worlds
Researchers introduce JEPA-Anything, a domain-agnostic world-modeling framework built on orthogonal predictive factorization (OPF), which extends joint-embedding predictive architectures by decomposing…
Engineering
Video DeltaNet: A Video-Native Hybrid Attention for Livestream Video Generation
Researchers present Video DeltaNet (VDN), a hybrid attention architecture for video diffusion models that combines local softmax attention with a…
Engineering
Register Tokens for Bounded-State Reasoning in Diffusion Language Models
Researchers propose register tokens — dedicated fixed-position tokens whose hidden states are trained to carry reasoning progress — to let…
Engineering
A Zeroth-Order Paradigm for LLM Preference Alignment
Researchers propose Comparison-based Preference Optimization (ComPO), a zeroth-order method for aligning large language models with human preferences that extracts directional…
Engineering
Region-Level Policy Optimization for Fine-grained MLLM Perception
Researchers show that in fine-grained multimodal LLM perception, localizing a region of interest tolerates roughly 3-4x stronger token compression than…
Engineering
HarnessVLN: Unifying Training-Free Embodied Navigation through an Agent Harness
Researchers introduce HarnessVLN, a zero-shot, training-free agent harness that unifies instruction-following and object-goal embodied navigation by coordinating perception, memory, and…
Engineering
ScienceIDE: Turning World’s Scientific Codebase into Agent Learnable Environments
Researchers introduce ScienceIDE, infrastructure that converts scientific code repositories into programmable, verifiable environments for training and evaluating scientific agents, addressing…
Engineering
CERA-MoA: Co-Evolving Routing Mechanisms with Continually Learning LLM Agents
Researchers introduce CERA-MoA, a reinforcement learning framework for Mixture-of-Agents systems in which a dynamic query router and independently fine-tuned agent…
Engineering
FLAT: Resampling Image and Text into 1D Flexible-Length Aligned Transmodal Tokens for Retrieval and Generation
Researchers present FLAT (Flexible-Length Aligned Transmodal representations), a multimodal pre-training framework that jointly trains a shared encoder with text-to-image and…
Engineering
Drift-Constrained Optimization: Only Direction Matters in Fine-Tuning Instruct Models
Researchers reframe instruct-model fine-tuning as a direction-selection problem: rather than treating behavioral drift from the reference model as an uncontrolled…
News
Huawei Unveils Ascend 960 AI Chips to Challenge Nvidia in China
Huawei unveiled its next-generation Ascend 960 AI chips at its Huawei Connect summit in Shanghai on September 17, 2026, accelerating…