AIWHYS

Technology

trajectory scoring

2 synchronized stories carry this tag.

ReliabilityResearch release

Skill-Use separates skill triggering, compliance, and boundaries

Skill-Use pairs 79 real skills with 177 executable tasks across nine domains and two harnesses; the strongest configuration reached only 0.613, while triggering and procedural compliance failed independently and rankings changed by harness.

Why it matters: A good skill document has no value when an agent fails to retrieve it, misses required steps, or crosses its forbidden operations.

Read storyOriginal source · arxiv.org
ReliabilityShipped

Strands Evals adds a practical production evaluation workflow

AWS published a hands-on framework for testing agents at tool, trajectory, response and end-to-end levels, with datasets, evaluators and repeatable experiment runs.

Why it matters: Layered evaluation shows whether a failure comes from tool selection, reasoning path, answer quality or the complete outcome instead of collapsing everything into one score.

Read storyOriginal source · aws.amazon.com