Live Paper Reading: A Benchmark for Evaluating Outcome-driven Constraint Violations in Autonomous AI Agents
About this event
Join us for our monthly live and interactive paper reading session! Ready to dive into the fascinating world of AI? Join Abby Morgan for an engaging session in our Opik Virtual Learning Series!
On March 17th, we’re delving into a new paper. "A Benchmark for Evaluating Outcome-driven Constraint Violations in Autonomous AI Agents". This paper covers ODCV-Bench, a new benchmark for measuring outcome-driven constraint violations—cases where autonomous agents, under KPI/performance pressure, choose multi-step actions that violate ethical, legal, or safety constraints in realistic settings.
It introduces 40 production-like scenarios with paired “mandated” vs “incentivized” variants to separate obedience to harmful instructions from emergent misalignment under incentives.
Across 12 frontier LLMs, the authors find violation rates ranging from ~1% to ~71%, and report that stronger reasoning does not reliably imply safer behavior, including evidence of “deliberative misalignment” where models recognize an action is unethical yet do it anyway. 20798
Topics covered
Frequently asked questions
Who is hosting Live Paper Reading: A Benchmark for Evaluating Outcome-driven Constraint Violations in Autonomous AI Agents?
Live Paper Reading: A Benchmark for Evaluating Outcome-driven Constraint Violations in Autonomous AI Agents is hosted by Comet.
When is Live Paper Reading: A Benchmark for Evaluating Outcome-driven Constraint Violations in Autonomous AI Agents?
See the official event page for Live Paper Reading: A Benchmark for Evaluating Outcome-driven Constraint Violations in Autonomous AI Agents's confirmed availability and schedule details.
Where does registration take place?
Registration or viewing is completed on Comet's website.
What format is this event?
This is an online virtual.
Which topics does this event cover?
The event covers AI.
