Rank-Aware Spectral Bounds on Attention Logits for Stable Low-Precision Training

arXiv:2602.18851 · cs.LG, cs.AI · Submitted 2026-02-21 · Read on arXiv

cs.LG, cs.AI

Submitted: 2026-02-21

Updated: 2026-02-21

Comments: 17 pages, 3 figures

Journal ref: Proceedings of the 43rd International Conference on Machine Learning (ICML 2026), PMLR, 2026

License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/

Terminology

• Calibration — 34× in this paper · explained in 7 episodes
• Accuracy — 20× in this paper · explained in 1 episode
• Transformer — 16× in this paper · explained in 2 episodes
• Fine-Tuning — 9× in this paper · explained in 5 episodes
• Grouped Query Attention — 9× in this paper
• Quantization — 8× in this paper · explained in 12 episodes
• Embeddings — 7× in this paper · explained in 2 episodes
• Alignment — 3× in this paper · explained in 1 episode
• Computational overhead — 3× in this paper · explained in 2 episodes
• Robustness — 3× in this paper · explained in 5 episodes
• Distribution Shift — 2× in this paper · explained in 6 episodes
• Entropy Collapse — 2× in this paper · explained in 2 episodes
• LoRA Fine-Tuning — 2× in this paper · explained in 3 episodes
• Failure Modes — 1× in this paper · explained in 3 episodes
• Transformer Models — 1× in this paper · explained in 2 episodes

Sources

• FP8 Formats for Deep Learning
• Microscaling Data Formats for Deep Learning
• RoFormer: Enhanced Transformer with Rotary Position Embedding
• DeepSeek-V3 Technical Report
• MOSS: Efficient and Accurate FP8 LLM Training with Microscaling and Automatic Scaling

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: Local-Consistency Optimisation for Bayesian Neural Networks
• Understanding Diffusion Models via Ratio-Based Function Approximation with SignReLU Networks