Cover image

The Safety Decision Before the Answer

ShallowDefend shows how explicit safety commitments improve multi-turn jailbreak resistance, while exposing the cost of rejecting legitimate requests.

November 11, 2026 · 8 min · Zelina
Cover image

The Verdict Is Not the Evidence

Why more LLM assessments and formal-verifier feedback can fail to improve software assurance, and what deterministic evidence checks actually contribute.

November 11, 2026 · 10 min · Zelina
Cover image

Easy First, Not Best Overall

TagInsert improves low-resource sequence tagging over left-to-right decoding, but BERT's accuracy, rare-label failures, and training costs limit its practical advantage.

November 10, 2026 · 7 min · Zelina
Cover image

Eleven Benchmark Wins, No Deployment Verdict

Smaller recurrent-depth models compete with larger clinical language models, but task-specific results, schema transfer, and unmeasured serving costs complicate deployment decisions.

November 10, 2026 · 8 min · Zelina
Cover image

The Safety Probe Was Only Part of the Gain

Value Probes improve multi-turn AI safety auditing, but much of the reported gain comes from structured conversation search before learned ranking is introduced.

November 10, 2026 · 9 min · Zelina
Cover image

The Transition You Never Saw Still Matters

Why successful simulations cannot establish almost-sure reachability, and what mathematical assumptions make certification of an unknown decision system possible.

November 10, 2026 · 8 min · Zelina
Cover image

When a Good Probe Points in Different Directions

A scalable method for reading and steering model concepts shows why classification accuracy alone cannot establish a stable or reliable internal feature.

November 10, 2026 · 9 min · Zelina
Cover image

When the Same Advisory Rule Produces Different Decisions

A shared-autonomy framework separates proposals, evaluation, and execution, making reusable approval logic explicit without requiring a new planning algorithm.

November 10, 2026 · 8 min · Zelina
Cover image

You Cannot Retrieve What Was Never Stored

A study of parameter-based AI memory finds that selecting the right information to store is a larger obstacle than retrieving it, with important limits for enterprise adoption.

November 10, 2026 · 9 min · Zelina
Cover image

A Correct Answer Is Not a Complete Model Profile

BULLPEN shows how a model's actual responses can improve behavioral profiling at lower evaluation cost, while revealing why no single profiling channel is sufficient.

November 9, 2026 · 9 min · Zelina