Tag
This survey comprehensively reviews resource-efficient architectures and hardware-software co-design for green AI, covering efficient model construction, training/deployment strategies, and sustainable hardware, aiming to guide sustainable large model development.
CompressKV proposes a semantic-retrieval-guided KV-cache compression method for GQA-based LLMs, identifying Semantic Retrieval Heads to retain critical tokens. It achieves over 97% full-cache performance using only 3% of the KV cache on LongBench tasks.
FedRAN is a resource-aware analytic federated continual learning framework that replaces gradient-based updates with compact random feature statistics, achieving high accuracy with significantly lower communication and computation costs.
TriEval is a new pipeline for evaluating LLMs across bias, toxicity, and truthfulness simultaneously, designed to be resource-efficient and run on standard laptops. It has been tested on Llama 3 8B, Mistral 7B, Gemma 2 9B, and Claude Haiku, and is released as open source.
This paper presents an end-to-end energy accounting framework for LLM distillation pipelines, measuring stage-wise energy costs and constructing energy-quality Pareto frontiers to reveal previously ignored teacher-side costs.
This paper introduces a resource-efficient pruning framework that identifies and removes parameters associated with unsafe behaviors in large language models while preserving utility. Using gradient-free attribution and the Lottery Ticket Hypothesis perspective, the method achieves significant reductions in unsafe generations and improved robustness against jailbreak attacks with minimal performance loss.