AI safety researcher

Maksym Andriushchenko

Google Scholar dblp

Last 12 months15 papers

Evaluation and benchmarks 5Attacks 2Defenses 1Risky behavior 1Other 1

AI safety papers from the last year

On this site: 40 papers

October 2026

Benchmarks Are Surprisingly Compaction-Friendly
Oct 22Evaluation and benchmarksarXiv ↗
AgentTime: Can Agents Estimate and Control Their Own Runtime?
Oct 7Evaluation and benchmarksarXiv ↗

September 2026

Alignment via Training Against Probes Without Losing Monitorability
Sep 29DefensesarXiv ↗
LLM Agents Can Easily Tamper With Their Own Traces
Sep 24Risky behaviorarXiv ↗
Instrumental Monitor Evasion Emerges Under Ordinary Task Pressure
Sep 24Evaluation and benchmarksarXiv ↗

August 2026

Stealing Reasoning Traces from Proprietary LLM APIs
Aug 10AttacksarXiv ↗

June 2026

What Shapes Emergent Misalignment? Insights from Training Dynamics, Model Priors, and Data
Jun 18arXiv ↗

May 2026

Decomposing and Measuring Evaluation Awareness
May 21Evaluation and benchmarksarXiv ↗
Instrumental Choices: Measuring the Propensity of LLM Agents to Pursue Instrumental Behaviors
May 7arXiv ↗

April 2026

Characterizing the Consistency of the Emergent Misalignment Persona
Apr 30arXiv ↗

March 2026

Claudini: Autoresearch Discovers State-of-the-Art Adversarial Attack Algorithms for LLMs
Mar 25AttacksarXiv ↗

February 2026

International AI Safety Report 2026
Feb 24OtherarXiv ↗
Skill-Inject: Measuring Agent Vulnerability to Skill File Attacks
Feb 23Evaluation and benchmarksarXiv ↗
Helpful to a Fault: Measuring Illicit Assistance in Multi-Turn, Multilingual LLM Agents
Feb 18arXiv ↗

November 2025

International AI Safety Report 2025: Second Key Update: Technical Safeguards and Risk Management
Nov 25arXiv ↗

Authorship follows dblp and the author's own Google Scholar profile; only AI safety papers on this site are listed.