MoonshotAI/Attention-Residuals

241 best points · 35 comments · 2 Hacker News posts · 3,567 GitHub stars · View on GitHub

Hacker News discussions

• Attention Residuals (241 points, 34 comments, by GaggiX, 2026-03-20)
• Attention Residuals: Rethinking depth-wise aggregation [pdf] (17 points, 1 comments, by salkahfi, 2026-03-16)