Top pick  ·  LLMs & Language Models

LowRAM: Low-Rank Keys and Amortised Values for Composable KV Cache Compression at Test-Time

LowRAM compresses KV caches by predicting values from keys and low-rank key compression, achieving up to 5× faster decoding at 4× compression.

Why it matters: Builders can run larger language models with less memory, enabling higher throughput and batch sizes.

Read the breakdown →  Code

Trending  ·  Other ML

SplatAnything: Gaussian Spacing Normalization for Feed-forward 3DGS across Arbitrary Scenes

SplatAnything normalizes Gaussian spacing for feed-forward 3DGS, improving consistency across varied scenes and viewpoints.

Why it matters: Consistent Gaussian predictions enable more reliable 3D reconstruction across diverse scenes and camera setups.

Read the breakdown →