Decoupling Token Roles in Autoregressive Pretraining
cs.LG, cs.AI, cs.CL
Submitted: 2026-09-27
Updated: 2026-09-27
Code: https://github.com/huggingface/nanotron
Terminology
Sources
- Physics of Language Models: Part 3.3, Knowledge Capacity Scaling Laws
- SmolLM2: When Smol Goes Big -- Data-Centric Training of a Small Language Model
- Demystifying Training-Time Augmentation for Data-Constrained Language Model Pretraining
- ContextCite: Attributing Model Generation to Context
- DeepSeek-V3 Technical Report
- Machine-generated text detection prevents language model collapse
- Is Model Collapse Inevitable? Breaking the Curse of Recursion by Accumulating Real and Synthetic Data
- The Llama 3 Herd of Models
- Studying Large Language Model Generalization with Influence Functions
- Textbooks Are All You Need
- Be like a Goldfish, Don't Memorize! Mitigating Memorization in Generative LLMs
- Understanding the Effect of Noise in LLM Training Data with Algorithmic Chains of Thought
- Instruction Fine-Tuning: Does Prompt Loss Matter?
- How (not) to Train your Generative Model: Scheduled Sampling, Likelihood, Adversary?
- Scaling Laws for Neural Language Models
- Mark, Don't Erase: Token Inoculation for Dual-Use Knowledge in LLMs
- Rho-1: Not All Tokens Are What You Need
- Towards Better Few-Shot and Finetuning Performance with Forgetful Causal Language Models
- AlphaToken: Decoupling Adaptation and Stability for Path-Aware Response Token Valuation in LLM Post-Training
- When Less is More: Investigating Data Pruning for Pretraining LLMs at Scale
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