BerLU Activation: Smooth, Efficient Neural Network Function
Discover BerLU, a smooth and efficient activation function using Bernstein polynomials to boost neural network performance and stability.
Training Large Language Models for Long-Horizon Tasks
Explore how horizon length impacts training dynamics of large language models and improves performance in long-horizon tasks.
DRLU-Based Semantics for Quantitative Bipolar Argumentation
Discover DRLU-based modular semantics enhancing argument strength evaluation in Quantitative Bipolar Argumentation Frameworks for AI decision-making.
Strategy-Aware Optimization Modeling with LLMs
Discover how the SAGE framework boosts optimization modeling using strategy-aware reasoning in large language models for better efficiency and accuracy.
Boost AI Safety with Targeted Error Correction Methods
Enhance AI model safety by reducing high-risk errors using a dual-classifier GBDT pipeline for precise, low-latency error correction.