Vage Egiazarian

3 papers Higher School of Economics

• AutoJudge: Judge Decoding Without Manual Annotation
• Hogwild! Inference: Parallel LLM Generation via Concurrent Attention
• Unified Scaling Laws for Compressed Representations