RADAR / WEEKLY

2026-W39

This week in AI: source excerpts and reading links, grouped by original publication week.

THIS WEEK AT A GLANCE

Signals: 4
Topics: 3
Sources: 2

RULE-BASED DIGEST · Counts and source excerpts, not an authored weekly opinion.

TOP SIGNALS

Start with these signals

Ordered by available relevance score, then publication time.

Not All Irregularity Is Equal: Causally Isolating a Rare Failure Mode in Japanese Morphological Inflection →PAPER
An Interpretable Memory Decision Controller for LLM Agents Based on Three-Signal Complementarity: Decoupling Confidence and Consistency →PAPER
FOCAL-VLA: Subtask-Guided Geometry Distillation and Implicit World Modeling for Vision-Language-Action Models →PAPER

WHAT CHANGED / SOURCE SIGNALS

MODELS

RADAR / SIGNAL · MODELS · PAPER · 2026-09-21

SOURCE STATUS · SINGLE SOURCE

Not All Irregularity Is Equal: Causally Isolating a Rare Failure Mode in Japanese Morphological Inflection ↗

SOURCE LANGUAGE

SOURCE EXCERPT

Neural morphological generation systems often achieve high aggregate accuracy on benchmark datasets, yet such performance can conceal systematic errors clustered in rare morphol…

arXiv cs.CL · Source

WHAT CHANGED / SOURCE SIGNALS

RAG

RADAR / SIGNAL · RAG · PAPER · 2026-09-21

SOURCE STATUS · SINGLE SOURCE

An Interpretable Memory Decision Controller for LLM Agents Based on Three-Signal Complementarity: Decoupling Confidence and Consistency ↗

SOURCE LANGUAGE

SOURCE EXCERPT

Memory systems for large language models have focused predominantly on efficient retrieval, whereas the decision of whether retrieved memories should be trusted has received com…

arXiv cs.CL · Source

WHAT CHANGED / SOURCE SIGNALS

EMBODIED-AI

RADAR / SIGNAL · EMBODIED-AI · PAPER · 2026-09-21

SOURCE STATUS · SINGLE SOURCE

FOCAL-VLA: Subtask-Guided Geometry Distillation and Implicit World Modeling for Vision-Language-Action Models ↗

SOURCE LANGUAGE

SOURCE EXCERPT

Vision-language-action (VLA) models built on pretrained vision-language models have demonstrated strong performance across diverse robotic manipulation tasks. However, VLA model…

arXiv cs.RO · Source

RADAR / SIGNAL · EMBODIED-AI · PAPER · 2026-09-21

SOURCE STATUS · SINGLE SOURCE

Catch Me If You Can: Real-Time Feedback Denoising for Responsive VLAs ↗

SOURCE LANGUAGE

SOURCE EXCERPT

Vision-Language-Action (VLA) models have shown strong generalization in robotic manipulation by combining semantic knowledge from pretrained vision-language models with expressi…

arXiv cs.RO · Source

WORTH FOLLOWING

Questions still open

Not All Irregularity Is Equal: Causally Isolating a Rare Failure Mode in Japanese Morphological Inflection · Awaiting independent verification
An Interpretable Memory Decision Controller for LLM Agents Based on Three-Signal Complementarity: Decoupling Confidence and Consistency · Awaiting independent verification
FOCAL-VLA: Subtask-Guided Geometry Distillation and Implicit World Modeling for Vision-Language-Action Models · Awaiting independent verification
Catch Me If You Can: Real-Time Feedback Denoising for Responsive VLAs · Awaiting independent verification

← AI Radar