LLM Digest

Story

arxiv_cs_lg · Oct 21, 2026 · paper

Source brief

VX-One: Typed, Calibrated Decision Heads for Real-Time Perimeter Triage - An empirical study of unknown-attack honesty, cascade gray zones, and calibration under drift

arxiv.orgOct 21, 2026
original source linked

In brief

Perimeter triage faces a latency/semantics trade-off: signature engines are fast and format-safe but blind to unseen attacks, while generative models are semantic but slow and produce unparseable output. In late 2026...

Feed lens

eval

Read the original at arxiv.org →Open in live feed Read that day’s brief

VX-One: Typed, Calibrated Decision Heads for Real-Time Perimeter Triage - An empirical study of unknown-attack honesty, cascade gray zones, and calibration under drift

Earlier in this thread 4 items

One Word Opens the Gate: The Option-Channel Attack on Typed Decision Models as Agent Guardrails

An Empirical Study of VLM Pipelines for Long-Document QA

Constraint Tax in Open-Weight LLMs: An Empirical Study of Tool Calling Suppression Under Structured Output Constraints

The Last Mile of Coding Agents for Long-Document Tasks: An Empirical Study on Clinical Trial Protocol Digitization