Scaling Down the Scaling Laws: Parameter Efficiency and Compute-Optimal Training in Resource-Constrained Large Language Models
cs.LG, cs.AI
Submitted: 2026-10-05
Updated: 2026-10-07
License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Terminology
Sources
- Accelerated Training on Low-Power Edge Devices
- Explaining Neural Scaling Laws
- Unified Neural Network Scaling Laws and Scale-time Equivalence
- Language Models are Few-Shot Learners
- PaLM: Scaling Language Modeling with Pathways
- The rising costs of training frontier AI models
- QLoRA: Efficient Finetuning of Quantized LLMs
- Textbooks Are All You Need
- Deep Reinforcement Learning that Matters
- Training Compute-Optimal Large Language Models
- Learning Curve Theory
- Mistral 7B
- Scaling Laws for Neural Language Models
- TPI-LLM: Serving 70B-scale LLMs Efficiently on Low-resource Edge Devices
- A Solvable Model of Neural Scaling Laws
- Gemma: Open Models Based on Gemini Research and Technology
- Efficient Estimation of Word Representations in Vector Space
- MEADOW: Memory-efficient Dataflow and Data Packing for Low Power Edge LLMs
- Efficient Large-Scale Language Model Training on GPU Clusters Using Megatron-LM
- Improving Data Efficiency via Curating LLM-Driven Rating Systems
Related papers
- Polynomial-Augmented Neural Networks (PANNs) with Weak Orthogonality Constraints for Enhanced Function and PDE Approximation
- AIRL-S: Unifying Reinforcement Learning and Search-Based Test-Time Scaling via Adversarial Inverse Reinforcement Learning
- Transformers as Bayesian In-Context Experimenters: Smoothness-Adaptive Efficient ATE Estimation
- Convergence issues in Relational Concept Analysis based on AOC-posets
- Beliefs Beyond Posteriors: The Predictive Cost of ELBO Binding
- Understanding Diffusion Models via Ratio-Based Function Approximation with SignReLU Networks