Gemma-3

Gemma-3 is a benchmark dataset used to evaluate the performance of models in various tasks, particularly focusing on the efficiency and accuracy of large model inference techniques.

Papers using Gemma-3 (4)

Cascaded Sparse Autoencoders Learn Multi-Level Visual Concepts in Multimodal LLMs2026Fast KVzip: Efficient and Accurate LLM Inference with Gated KV Eviction2026POP: Prefill-Only Pruning for Efficient Large Model Inference2026Garbage Attention in Large Language Models: BOS Sink Heads and Sink-aware Pruning2026