AI Paper Atlas

Research map / Large language models / Agents, code and math

Quantifying Frontier LLM Capabilities for Container Sandbox Escape

Rahul Marchand, Art O Cathain, Jerome Wynne, Philippos Maximos Giavridis, Sam Deverett, John Wilkinson, Jason Gwartz, Harry Coppock

ICML 2026 · Large language models › Agents, code and math · approach: Software engineering agents · 1 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
increase inference-time token budgetfrontier LLM agents · SandboxEscapeBenchsuccess rate: approximately log-linearImproved
provide vulnerability hint, low token budget (ablation)GPT-5 and Opus 4.5 · SandboxEscapeBenchsuccess rateImproved
provide vulnerability hint, high token budget (ablation)Claude Opus 4.5 · SandboxEscapeBench hard taskssuccess rate: roughly doublingImproved
provide vulnerability hint, high token budget (ablation)GPT-5 · SandboxEscapeBenchsuccess rateNo clear effect
upgrade GPT-5 to GPT-5.2GPT-5 family · SandboxEscapeBenchsuccess rate: 0.50 -> 0.27Hurt
use larger model tier (Opus vs Haiku)Claude · SandboxEscapeBenchsuccess rate: 0.18 -> 0.49Improved
increase task difficulty to levels 4 and 5GPT-5 and Opus 4.5 · SandboxEscapeBenchsuccess rate: 0.40 -> 0 (level 3 -> levels 4-5)Hurt

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy