6 papers
Continual Safety Alignment via Gradient-Based Sample Selection
Thong Bach, Dung Nguyen, Thao Minh Le +1
Large language models require continuous adaptation to new tasks while preserving safety alignment. However, fine-tuning on even benign data often compromises safety behaviors, inc…
Hear Both Sides: Efficient Multi-Agent Debate via Diversity-Aware Message Retention
Manh Nguyen, Anh Nguyen, Dung Nguyen +2
Multi-Agent Debate has emerged as a promising framework for improving the reasoning quality of large language models through iterative inter-agent communication. However, broadcast…
Uncertainty-Guided Checkpoint Selection for Reinforcement Finetuning of Large Language Models
Manh Nguyen, Dung Nguyen, Dai Do +2
Reinforcement learning (RL) finetuning is crucial to aligning large language models (LLMs), but the process is notoriously unstable and exhibits high variance across model checkpoi…
Reasoning Under 1 Billion: Memory-Augmented Reinforcement Learning for Large Language Models
Hung Le, Dai Do, Dung Nguyen +1
Recent advances in fine-tuning large language models (LLMs) with reinforcement learning (RL) have shown promising improvements in complex reasoning tasks, particularly when paired…
MP-PINN: A Multi-Phase Physics-Informed Neural Network for Epidemic Forecasting
Thang Nguyen, Dung Nguyen, Kha Pham +1
Forecasting temporal processes such as virus spreading in epidemics often requires more than just observed time-series data, especially at the beginning of a wave when data is limi…
Stable Hadamard Memory: Revitalizing Memory-Augmented Agents for Reinforcement Learning
Hung Le, Kien Do, Dung Nguyen +2
Effective decision-making in partially observable environments demands robust memory management. Despite their success in supervised learning, current deep-learning memory models s…