kalomaze: crude diffusion autoencoders may lose low-frequency info at the latent stage
kalomaze · x · 2026-11-04
Responding to an analysis of 120 generated videos showing models fail to produce low-frequency airplane Doppler effects, kalomaze argues that autoencoders in SOTA diffusion models are crude enough that information isn't preserved when encoded into bottleneck latents — it may be unrecoverable even in the training data as the model actually saw it.
More from Models
- Post-train your model on your own agent harness with OpenHands, Harbor, SkyRL and Snowflake — StasBekman · 2026-11-04
- Google Antigravity redesign leaked with context usage and quota indicators — clmt · 2026-11-04
- NVIDIA's Nemotron Transcribe Demo Space Trends on Hugging Face — nvidia · 2026-11-04
- Mythos 5.5 spotted in Anthropic's HTML, GPT-6.2 Astra in Codex analytics — rumor — haider1 · 2026-11-04
- How LLM inference actually works: from 140GB weights to GPUs and model serving, explained as a story — _jaydeepkarale · 2026-11-04
- $100 of GPU time edits model weights to cut thinking tokens by 40% — jarrodwatts · 2026-11-04