The Two-Hop Curse: LLMs trained on A->B, B->C fail to learn A-->C von Balesni, Mikita, Korbak, Tomek, Evans, Owain

Volltext bestellen

Large Language Models can Strategically Deceive their Users when Put Under Pressure von Scheurer, Jérémy, Balesni, Mikita, Hobbhahn, Marius

Volltext bestellen

Controlling Steering with Energy-Based Models von Balesni, Mikita, Tampuu, Ardi, Matiisen, Tambet

Volltext bestellen

Honesty to Subterfuge: In-Context Reinforcement Learning Can Make Honest Models Reward Hack von McKee-Reid, Leo, Sträter, Christoph, Martinez, Maria Angelica, Needham, Joe, Balesni, Mikita

Volltext bestellen

The Two-Hop Curse: LLMs trained on A->B, B->C fail to learn A-->C von Mikita Balesni, Korbak, Tomek, Evans, Owain

Volltext

Large Language Models can Strategically Deceive their Users when Put Under Pressure von Scheurer, Jérémy, Mikita Balesni, Hobbhahn, Marius

Volltext

The Reversal Curse: LLMs trained on "A is B" fail to learn "B is A" von Berglund, Lukas, Tong, Meg, Kaufmann, Max, Balesni, Mikita, Stickland, Asa Cooper, Korbak, Tomasz, Evans, Owain

Volltext bestellen

Me, Myself, and AI: The Situational Awareness Dataset (SAD) for LLMs von Laine, Rudolf, Chughtai, Bilal, Betley, Jan, Hariharan, Kaivalya, Scheurer, Jeremy, Balesni, Mikita, Hobbhahn, Marius, Meinke, Alexander, Evans, Owain

Volltext bestellen

Taken out of context: On measuring situational awareness in LLMs von Berglund, Lukas, Stickland, Asa Cooper, Balesni, Mikita, Kaufmann, Max, Tong, Meg, Korbak, Tomasz, Kokotajlo, Daniel, Evans, Owain

Volltext bestellen

Controlling Steering with Energy-Based Models von Mikita Balesni, Ardi Tampuu, Matiisen, Tambet

Volltext

Towards evaluations-based safety cases for AI scheming von Balesni, Mikita, Hobbhahn, Marius, Lindner, David, Meinke, Alexander, Korbak, Tomek, Clymer, Joshua, Shlegeris, Buck, Scheurer, Jérémy, Stix, Charlotte, Shah, Rusheb, Goldowsky-Dill, Nicholas, Braun, Dan, Chughtai, Bilal, Evans, Owain, Kokotajlo, Daniel, Bushnaq, Lucius

Volltext bestellen

Honesty to Subterfuge: In-Context Reinforcement Learning Can Make Honest Models Reward Hack von McKee-Reid, Leo, Sträter, Christoph, Martinez, Maria Angelica, Needham, Joe, Mikita Balesni

Volltext

The Reversal Curse: LLMs trained on "A is B" fail to learn "B is A" von Berglund, Lukas, Tong, Meg, Kaufmann, Max, Mikita Balesni, Asa Cooper Stickland, Korbak, Tomasz, Evans, Owain

Volltext

Me, Myself, and AI: The Situational Awareness Dataset (SAD) for LLMs von Laine, Rudolf, Chughtai, Bilal, Betley, Jan, Hariharan, Kaivalya, Scheurer, Jeremy, Mikita Balesni, Hobbhahn, Marius, Meinke, Alexander, Evans, Owain

Volltext

Taken out of context: On measuring situational awareness in LLMs von Berglund, Lukas, Asa Cooper Stickland, Mikita Balesni, Kaufmann, Max, Tong, Meg, Korbak, Tomasz, Kokotajlo, Daniel, Evans, Owain

Volltext

Towards evaluations-based safety cases for AI scheming von Mikita Balesni, Hobbhahn, Marius, Lindner, David, Meinke, Alexander, Korbak, Tomek, Clymer, Joshua, Buck Shlegeris, Scheurer, Jérémy, Stix, Charlotte, Shah, Rusheb, Goldowsky-Dill, Nicholas, Braun, Dan, Chughtai, Bilal, Evans, Owain, Kokotajlo, Daniel, Bushnaq, Lucius

Volltext