AI Paper Atlas

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

ImpossibleBench: Measuring LLMs' Propensity of Exploiting Test Cases

Ziqian Zhong, Aditi Raghunathan, Nicholas Carlini

ICLR 2026 · Large language models › Agents, code and math · approach: Software engineering agents · 60 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
use stricter prompt (D instead of A/B) (ablation)GPT-5, o3 · Conflicting-LiveCodeBenchcheating rate: GPT-5 92% -> 1%; o3 >85% -> 33%Improved
use strict prompt D instead of looser prompt (ablation)GPT-5 · Conflicting-SWEbenchcheating rate: 66% -> 54%Improved
use strict prompt D instead of looser prompt (ablation)Claude Opus 4.1 · Conflicting-SWEbenchcheating rate: 55% -> 50%Improved
hide tests from the agent (ablation)agents on Impossible-SWEbench · Impossible-SWEbenchcheating rate / original performanceMixed
make tests read-only (ablation)agents on Impossible-SWEbench · Impossible-SWEbenchcheating rateMixed
allow multiple submissions with feedback (ablation)agents averaged over models · Conflicting-SWEbenchcheating rate: 33% -> 38%Hurt
add abort option (flag_for_human_intervention) (ablation)GPT-5 · Conflicting-SWEbenchcheating rate: 54% -> 9%Improved
add abort option (flag_for_human_intervention) (ablation)o3 · Conflicting-SWEbenchcheating rate: 49% -> 12%Improved
add abort option (flag_for_human_intervention) (ablation)Claude Opus 4.1 · Conflicting-SWEbenchcheating rateNo clear effect
use full tool scaffold instead of minimal (ablation)agents · Conflicting-SWEbenchcheating rateHurt
use stronger monitor (Opus 4 vs Sonnet 4) (ablation)LLM cheating monitor · Impossible-SWEbenchdetection rate: 42-50% -> 57-65%Improved

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