Cover image

The Rare Cases the Classifier Never Asks About

Why uncertainty-based active learning can reinforce rare-class selection bias, and how to test whether annotation budgets uncover missing categories.

November 8, 2026 · 7 min · Zelina
Cover image

The Same Math Score, Different Wrong Answers

Arabic mathematical reasoning tests reveal how cultural localization can change which problems AI models solve correctly, even when average accuracy barely moves.

November 8, 2026 · 8 min · Zelina
Cover image

When AI Reviewers Are Talked Into Approval

Experiments reveal how strategically instructed AI agents change oversight decisions, and why detecting threats is not enough to preserve independent review.

November 8, 2026 · 9 min · Zelina
Cover image

A Better Forecast Can Choose a Worse Plan

ActTime shows why accurate time-series forecasts are not enough to predict intervention effects or select better operational decisions.

November 7, 2026 · 9 min · Zelina
Cover image

Better Boxes, Unproven Specialists

LaMER improves language-guided object localization through multi-expert routing, but its accuracy gains do not yet establish expert specialization or justify deployment costs.

November 7, 2026 · 8 min · Zelina
Cover image

Keep the Record, Rebuild the Rule

Why agents can improve by retaining execution histories while discarding the instructions they generate from those histories.

November 7, 2026 · 8 min · Zelina
Cover image

Sixteen Compounds, One Forward Pass

Molecular Prior Networks offer fast affinity ranking with few labelled compounds, but their predictive advantage depends on how much experimental data a discovery team has.

November 7, 2026 · 8 min · Zelina
Cover image

The Facts Didn't Make It to the Decision

TeamGym reveals how AI teams lose decision-critical evidence through retrieval, communication, and synthesis failures—and what operators can test to find the breakdown.

November 7, 2026 · 10 min · Zelina
Cover image

Train the Advice, Not Just the Agent

Caddie shows how task outcomes can train a separate AI critic to improve multi-step agents, while revealing the costs and limits of reusable advice.

November 7, 2026 · 9 min · Zelina
Cover image

When the Evidence Changes, Does the Agent?

A framework for testing whether AI agents revise decisions when evidence changes, rather than merely producing correct answers through repeated search.

November 7, 2026 · 8 min · Zelina