Shayan Shahrabi

mentions 1 type Person feed RSS

// recent coverage 1 mentions

04:00

2026-08-20

machinebrief.com

artificial-intelligence

Compress and Forget: bitsandbytes Quantization Amplifies Proactive Interference in LLMs

A new study from researchers including Shayan Shahrabi finds that bitsandbytes 4-bit (INT4/NF4) post-training quantization significantly amplifies proactive interference in large language models, redu…

// co-occurs with top 6 entities

bitsandbytes 1 Qwen2.5-7B-Instruct 1 Mistral-7B-Instruct-v0.3 1 Phi-3.5-mini-instruct 1 arXiv 1 GitHub 1

// topics top 5 topics

artificial intelligence 1 machine learning 1 large language models 1 ai research 1 ai infrastructure 1