Christos Thrampoulidis

Short-Context dominance: How Much Local Context Natural Language Actually Needs?

Token Hidden Reward: Steering Exploration-Exploitation in Group Relative Deep Reinforcement Learning

Facts in Stats: Impacts of Pretraining Diversity on Language Model Generalization

How Muon’s Spectral Design Benefits Generalization: A Study on Imbalanced Data

Efficient Forward-Only Data Valuation For LLMs and VLMs

Implicit Bias of Spectral Descent and Muon on Multiclass Separable Data

DARE the Extreme: Revisiting Delta-Parameter Pruning For Fine-Tuned Models

In-Context Occam’s Razor: How Transformers Prefer Simpler Hypotheses on the Fly

Sharper Guarantees for Learning Neural Network Classifiers with Gradient Methods

Thumb on the Scale: Optimal Loss Weighting in Last Layer Retraining

AoPS Dataset: Leveraging Online Olympiad-Level Math Problems for LLMs Training and Contamination-Resistant Evaluation

On the Effect of Negative Gradient in Group Relative Deep Reinforcement Optimization

Leveraging Online Olympiad-Level Math Problems for LLMs Training and Contamination-Resistant Evaluation