AutoCompress Framework for Hardware-Guided Model Compression
AutoCompress optimizes LLM architectures for latency and memory constraints, improving accuracy over Minitron by up to 6.57 percentage points.
Stories per day · 30 days
2 RSS
Relatedautoencoder1coding-agents1hardware-guidance1llama1mistral1model-compression1nemotron1pruning1rank-allocation1self-improvement1svd1
7 stories, newest first · one card per story
Stories tagged #llm-compression
arXiv study shows coding agents can compress LLMs
29 of 36 agent submissions pass verification, achieving 77.33% downstream accuracy on compressed models.
LEMS and KFAC-SVD improve SVD compression of large language models
Zero-shot gains reach 4.8 points through 70B models, and KFAC-SVD lowers perplexity 15% on average.
Sub-1-Bit LLM Compression via Latent Factorization
Samsung Labs’ LittleBit repository describes sub-1-bit LLM compression through latent factorization.
Learning Functional Subspaces for Neural Network Compression
Learnable Subspace Projections compress LLMs end-to-end, beating baselines at 70% compression on Llama-2-7B.
PrismML hopes its tiny LLM will change how we all use AI
PrismML released Bonsai 2 27B, compressing Alibaba's Qwen 27B to 5.9 GB with ternary weights while retaining 98% of benchmark performance.
4
What Breaks Under Pruning in Smart Homes, and When? Evaluating LLM Degradation Across Architectures and Task Complexity
Pruning study across four LLM architectures finds dense models degrade sharply on smart-home tool calling while MoE models tolerate far more.