Tuesday 15 September 2026Paper No. 1,386

Evening EditionRSS

ZOTPAPER

News without the noise

AI & ML

cs.AI, cs.LG, stat.ML

18 articles

All Agent Harnesses Tools & MCP Multi-Agent Systems Creative Agents Memory & Planning Safety & Evals RL for Agents AI & ML

Viewgrid list compact

Looped flows enable deeper reasoning by chaining denoising steps

Diffusion models generate high-resolution elevation maps from low-resolution inputs guided by optical imagery

Sparse mixture-of-experts models overfit repeated training data faster

Chain-of-thought reasoning traces are legible but not interpretable

Discrete diffusion accelerates language models while preserving their autoregressive distribution

Graph Machine architecture uses dynamic sparse routing to handle large state efficiently

Adding Depth Helps Video World Models Control Robots More Reliably

Embedding retrieval ranks matching words above shared underlying structure

Frontier LLMs design competitive algorithms for well-specified operations problems

Learned sepsis score tracks hourly severity without hourly outcome labels

One video world model transfers physical dynamics across robot bodies

Self-evolving loop synthesizes high-quality multimodal training data

Learning-rate-to-parameter-norm ratio predicts language model pretraining loss

A 2.8-trillion-parameter open MoE model approaches frontier performance.

Self-play with evolving skills boosts LLM benchmarks

Structural analysis filters noisy agent traces to pinpoint root causes

AI progress may continue past AGI toward superintelligence, not stop

New benchmark standardizes evaluation of coding agent harnesses