Scaling Zero-Order Pretraining through Model Sharding
cs.LG
Submitted: 2026-09-29
Updated: 2026-09-29
Code: https://github.com/openai/tiktoken
Sources
- Gradients without Backpropagation
- Scaling Recurrent Neural Networks to a Billion Parameters with Zero-Order Optimization
- Enhancing Zeroth-order Fine-tuning for Language Models with Low-rank Structures
- DiLoCo: Distributed Low-Communication Training of Language Models
- Scaling Expert Language Models with Unsupervised Domain Discovery
- Powering Up Zeroth-Order Training via Subspace Gradient Orthogonalization
- Branch-Train-Merge: Embarrassingly Parallel Training of Expert Language Models
- Fine-Tuning Language Models with Just Forward Passes
- Pointer Sentinel Mixture Models
- Byte Latent Transformer: Patches Scale Better Than Tokens
- Evolution Strategies as a Scalable Alternative to Reinforcement Learning
- Evolution Strategies at the Hyperscale
- Branch-Train-MiX: Mixing Expert LLMs into a Mixture-of-Experts LLM
- Unbiased Online Recurrent Optimization
- Simultaneous Computation and Memory Efficient Zeroth-Order Optimizer for Fine-Tuning Large Language Models
- ByT5: Towards a token-free future with pre-trained byte-to-byte models
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