Image created with gemini-2.5-flash-image with claude-sonnet-4-5. Image prompt: An 80s suburban Halloween porch scene with warm jack-o’-lantern lighting, a costumed trick-or-treater holding a glowing smartphone displaying TikTok’s interface with viral metrics, an overflowing candy bucket with vintage 80s candy bars, classic aluminum siding house exterior, autumn leaves scattered on wooden steps, nostalgic Spielberg-esque cinematography with practical lighting and film grain.

Another ByteDance Seed banger? They introduce the looped language models (LoopLMs) Ouro 1.4B and 2.6B trained on 7.7T tokens, that match evaluation results of larger 4B and 8B models respectively. “”Ouro”” 1.4B is a standard decoder-only Transformer with 24 layers (upcycled Ouro https://x.com/scaling01/status/1984286236438094307

Trending

Discover more from Ethan B. Holland

Subscribe now to keep reading and get access to the full archive.

Continue reading