Mikita Balesni

Researcher, Apollo Research

Joined

September 2023

Names

Mikita Balesni (Preferred)

Emails

****@gmail.com (Confirmed)

****@apolloresearch.ai (Confirmed)

Personal Links

Homepage

Google Scholar

Career & Education History

Researcher

Apollo Research (apolloresearch.ai)

2023 – Present

MS student

institute of computer science, University of Tartu (cs.ut.ee)

2020 – 2022

Undergrad student

National Taras Shevchenko University of Kiev (knu.ua)

2016 – 2020

Advisors, Relations & Conflicts

No relations added

Expertise

large language models

artificial intelligence safety

2022 – Present

Publications

Honesty to Subterfuge: In-Context Reinforcement Learning Can Make Honest Models Reward Hack
Leo McKee-Reid, Christoph Sträter, Maria Angelica Martinez, Joe Needham, Mikita Balesni
- SafeGenAi Poster
- Readers: Everyone
Honesty to Subterfuge: In-Context Reinforcement Learning Can Make Honest Models Reward Hack
Leo McKee-Reid, Christoph Sträter, Maria Angelica Martinez, Joe Needham, Mikita Balesni
- Submitted to ICLR 2025
- Readers: Everyone
The Two-Hop Curse: LLMs trained on A→B, B→C fail to learn A→C
Mikita Balesni, Tomasz Korbak, Owain Evans
- Submitted to ICLR 2025
- Readers: Everyone
Honesty to Subterfuge: In-Context Reinforcement Learning Can Make Honest Models Reward Hack
Leo McKee-Reid, Christoph Sträter, Maria Angelica Martinez, Joe Needham, Mikita Balesni
- SoLaR Poster
- Readers: Everyone
Me, Myself, and AI: The Situational Awareness Dataset (SAD) for LLMs
Rudolf Laine, Bilal Chughtai, Jan Betley, Kaivalya Hariharan, Mikita Balesni, Jérémy Scheurer, Marius Hobbhahn, Alexander Meinke, Owain Evans
- NeurIPS 2024 Track Datasets and Benchmarks Poster
- Readers: Everyone
Large Language Models can Strategically Deceive their Users when Put Under Pressure
Jérémy Scheurer, Mikita Balesni, Marius Hobbhahn
- LLMAgents @ ICLR 2024 Oral
- Readers: Everyone
Technical Report: Large Language Models can Strategically Deceive their Users when Put Under Pressure
Mikita Balesni, Jérémy Scheurer, Marius Hobbhahn
- OpenReview Archive Direct Upload
- Readers: Everyone
Taken out of context: On measuring situational awareness in LLMs
Mikita Balesni, Lukas Berglund, Meg Tong, Maximilian Kaufmann, Asa Cooper Stickland, Tomasz Korbak, Owain Evans, Daniel Kokotajlo
- OpenReview Archive Direct Upload
- Readers: Everyone
The Reversal Curse: LLMs trained on "A is B" fail to learn "B is A"
Lukas Berglund, Meg Tong, Maximilian Kaufmann, Mikita Balesni, Asa Stickland, Tomasz Korbak, Owain Evans
- ICBINB 2023
- Readers: Everyone
The Reversal Curse: LLMs trained on “A is B” fail to learn “B is A”
Lukas Berglund, Meg Tong, Maximilian Kaufmann, Mikita Balesni, Asa Cooper Stickland, Tomasz Korbak, Owain Evans
- ICLR 2024 poster
- Readers: Everyone