ORCA: The Annealed Spectral Conditioning Optimizer for Faster, Better LLM Training
cs.LG
Submitted: 2026-10-05
Updated: 2026-10-05
Sources
- Think you have Solved Question Answering? Try ARC, the AI2 Reasoning Challenge
- Muon$^p$: Muon with Fractional Spectral Powers
- The Pile: An 800GB Dataset of Diverse Text for Language Modeling
- MiniCPM: Unveiling the Potential of Small Language Models with Scalable Training Strategies
- Momentum Stiefel Optimizer, with Applications to Suitably-Orthogonal Attention, and Optimal Transport
- Muon is Scalable for LLM Training
- SGDR: Stochastic Gradient Descent with Warm Restarts
- Decoupled Weight Decay Regularization
- 2 OLMo 2 Furious
- Muon Dynamics as a Spectral Wasserstein Flow
- POET-X: Memory-efficient LLM Training by Scaling Orthogonal Transformation
- Pion: A Spectrum-Preserving Optimizer via Orthogonal Equivalence Transformation
- LLaMA: Open and Efficient Foundation Language Models
- PC Layer: Polynomial Weight Preconditioning for Improving LLM Pre-Training
- Qwen3 Technical Report
- ISO: An RLVR-Native Optimization Stack
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