Meta's Muse Spark 1.1 model card shows Gemini 3.1 Pro underperforming on key AI misalignment benchmarks
AI safety researcher Jiaxin Wen notes Gemini has trailed for over six months.
A post from @eliebakouch is circulating after pointing to a Meta Muse Spark 1.1 model card image and claiming Gemini 3.1 Pro looks unusually strong on AI misalignment benchmarks. In a reply on X, AI safety researcher Jiaxin Wen added that Gemini has been "SOTA on misalignment for over half a year," which sharpens the claim but does not independently settle it from the material visible here. For now, the clearest read is still the public post, the cited image, and the reaction forming around them on X.
Combined views
7.2K
2 posts, first seen 2h ago
Meta's Muse Spark 1.1 model card shows Gemini 3.1 Pro underperforming on key AI misalignment benchmarks
AI safety researcher Jiaxin Wen notes Gemini has trailed for over six months.
A post from @eliebakouch is circulating after pointing to a Meta Muse Spark 1.1 model card image and claiming Gemini 3.1 Pro looks unusually strong on AI misalignment benchmarks. In a reply on X, AI safety researcher Jiaxin Wen added that Gemini has been "SOTA on misalignment for over half a year," which sharpens the claim but does not independently settle it from the material visible here. For now, the clearest read is still the public post, the cited image, and the reaction forming around them on X.
Combined views
7.2K
2 posts, first seen 2h ago