metatrot/inspect_swe
Software Engineering Agents for Inspect AI
Repositories
metatrot/hawk-nonce-probe
Fresh-nonce arithmetic probe task for verifying a live self-hosted Hawk instance (ct-gauntlet test 8).
metatrot/inspect-viewer
metatrot/inspect_ai
Inspect: A framework for large language model evaluations
metatrot/inspect-tasks-public
metatrot/control-arena
ControlArena is a suite of realistic settings, mimicking complex deployment environments, for running control evaluations. This is an alpha release; we welcome feedback.
metatrot/ca-k8s-infra
A k8s infrastructure setting for ControlArena
metatrot/vivaria
Vivaria is METR's tool for running evaluations and conducting agent elicitation research.
metatrot/public-vivaria-test
Demo of Vivaria using Vagrant to run public agents/tasks