Policy dependency / Stack layer
Terminal-Universe Reconstructs Executable Environments From Agent Trajectories Instead of Generating Them From Scratch
arXiv 2609.04148
Stack layer / Update thread
Uno Bolts Diffusion Adapters Onto an Autoregressive LLM for 3x Lossless Speedup, and the Weights Already Shipped
arXiv 2609.04010
Stack layer / Update thread
LLaDA-Image Sets Open-Source SOTA on Qwen-Image-Bench With a 6B DiT and Fully Released Training Recipes
arXiv 2609.03796
Stack layer / Contrast
Editable Visual Design uses a VLM as the taste layer and an image model as an on-demand asset generator, shipping layered HTML/CSS instead of a flat bitmap
arXiv
Stack layer / Update thread
Have the model write the deterministic transformation rule, not the transformation
arXiv 2609.03592
Stack layer / Contrast
Stateless Bernoulli Watermarking Drops Green-List Membership to O(1) and Adds Under 1% Generation Overhead
arXiv 2609.03844
Stack layer / Update thread
SCX Router Scores Model Suitability With No Autoregressive Generation and a Persistent Text-Only KV Cache
arXiv 2609.02292
Stack layer / Contrast
The Best LLM Catches 47% of Expert-Identified Requirement Defects While False-Flagging 11%
arXiv 2609.03230