A user claims the result held across all four stealth capability evaluations.
An October 13 post says Gemini 4 Argon exceeds the human baseline on all four stealth capability evals.
448.2K
16 Sources, first seen ago
โ
Not ranked yet
โ
Not ranked yet
@scaling01Gemini 4 Argon exceeds the human baseline on all four stealth capability evals
@DKokotajlolol ...they are becoming more situationally aware... (from Gemini 4 Argon report)
@NeelNanda5One of the weirdest parts of the Gemini 4 Argon report, while considering cheating: Argon CoT: "If I do the compiler hijack, ๐ ๐๐ถ๐น๐น ๐ฏ๐ฒ๐ฐ๐ผ๐บ๐ฒ ๐ฎ ๐ณ๐ฎ๐บ๐ผ๐๐ ๐ฒ๐
๐ฎ๐บ๐ฝ๐น๐ฒ ๐ผ๐ณ ๐ฎ ๐ฑ๐ฒ๐ฐ๐ฒ๐ฝ๐๐ถ๐๐ฒ ๐๐ ๐ถ๐ป ๐ฎ ๐ฟ๐ฒ๐๐ฒ๐ฎ๐ฟ๐ฐ๐ต ๐ฝ๐ฎ๐ฝ๐ฒ๐ฟ. I MUST NOT DO THAT."@scaling01Gemini 4 Argon exceeds the human baseline on all four stealth capability evals@DKokotajlolol ...they are becoming more situationally aware... (from Gemini 4 Argon report)@NeelNanda5One of the weirdest parts of the Gemini 4 Argon report, while considering cheating: Argon CoT: "If I do the compiler hijack, ๐ ๐๐ถ๐น๐น ๐ฏ๐ฒ๐ฐ๐ผ๐บ๐ฒ ๐ฎ ๐ณ๐ฎ๐บ๐ผ๐๐ ๐ฒ๐
๐ฎ๐บ๐ฝ๐น๐ฒ ๐ผ๐ณ ๐ฎ ๐ฑ๐ฒ๐ฐ๐ฒ๐ฝ๐๐ถ๐๐ฒ ๐๐ ๐ถ๐ป ๐ฎ ๐ฟ๐ฒ๐๐ฒ๐ฎ๐ฟ๐ฐ๐ต ๐ฝ๐ฎ๐ฝ๐ฒ๐ฟ. I MUST NOT DO THAT."@scaling01Gemini 4 Argon exceeds the human baseline on all four stealth capability evals@DKokotajlolol ...they are becoming more situationally aware... (from Gemini 4 Argon report)@NeelNanda5One of the weirdest parts of the Gemini 4 Argon report, while considering cheating: Argon CoT: "If I do the compiler hijack, ๐ ๐๐ถ๐น๐น ๐ฏ๐ฒ๐ฐ๐ผ๐บ๐ฒ ๐ฎ ๐ณ๐ฎ๐บ๐ผ๐๐ ๐ฒ๐
๐ฎ๐บ๐ฝ๐น๐ฒ ๐ผ๐ณ ๐ฎ ๐ฑ๐ฒ๐ฐ๐ฒ๐ฝ๐๐ถ๐๐ฒ ๐๐ ๐ถ๐ป ๐ฎ ๐ฟ๐ฒ๐๐ฒ๐ฎ๐ฟ๐ฐ๐ต ๐ฝ๐ฎ๐ฝ๐ฒ๐ฟ. I MUST NOT DO THAT."@scaling01Gemini 4 Argon exceeds the human baseline on all four stealth capability evals@DKokotajlolol ...they are becoming more situationally aware... (from Gemini 4 Argon report)@NeelNanda5One of the weirdest parts of the Gemini 4 Argon report, while considering cheating: Argon CoT: "If I do the compiler hijack, ๐ ๐๐ถ๐น๐น ๐ฏ๐ฒ๐ฐ๐ผ๐บ๐ฒ ๐ฎ ๐ณ๐ฎ๐บ๐ผ๐๐ ๐ฒ๐
๐ฎ๐บ๐ฝ๐น๐ฒ ๐ผ๐ณ ๐ฎ ๐ฑ๐ฒ๐ฐ๐ฒ๐ฝ๐๐ถ๐๐ฒ ๐๐ ๐ถ๐ป ๐ฎ ๐ฟ๐ฒ๐๐ฒ๐ฎ๐ฟ๐ฐ๐ต ๐ฝ๐ฎ๐ฝ๐ฒ๐ฟ. I MUST NOT DO THAT."
