Yap Wei Herng
Passionate about artificial intelligence, machine learning, and software development. Sharing insights from research and practical experience.
Research Areas
Latest Post
Retaining by Doing: The Role of On-Policy Data in Mitigating Forgetting
Introduction Background Adapting language models (LMs) to new tasks via post-training carries the risk of degrading existing capabilities, known as catastrophic forgetting. This phenomenon has been observed in both supervised fine-tuning (SFT) for instruction following and reinforcement learning (RL) for preference alignment. However, the comparative...
Recent Posts
Retaining by Doing: The Role of On-Policy Data in Mitigating Forgetting
Introduction Background Adapting language models (LMs) to new tasks via post-training carries the risk of degrading existing capabilities, known as catastrophic forgetting. This phenomenon has been observed in both supervised fine-tuning (SFT) for instruction following and reinforcement learning (RL) for preference alignment. However, the comparative susceptibility of SFT and RL...
Read ArticleAgent-based Automated Claim Matching with Instruction-following LLMs
Introduction Background Automated fact-checking pipelines rely on claim matching to identify claims that can be verified using the same evidence or fact-check. This task is crucial for scaling fact-checking efforts, as it helps in grouping related claims for efficient verification. Previous work has framed claim matching as a ranking problem...
Read ArticleThe 10,000x Explosion: Reproducing DeepSeek’s mHC at Scale
In the world of Large Language Models, “stability” is often the difference between a state-of-the-art model and an expensive pile of NaNs. When DeepSeek released their paper on Manifold Hyper-Connections (mHC), they proposed a solution to a hidden rot in transformer architectures: the explosion of signal through residual connections.
Read ArticlemHC: Manifold-Constrained Hyper-Connections
Introduction Background Deep neural network architectures have evolved significantly since the introduction of ResNets in 2016, with residual connections becoming a cornerstone of modern models like Transformers and large language models (LLMs). Hyper-Connections (HC) extended this paradigm by expanding the residual stream width and diversifying connectivity patterns, yielding performance gains...
Read ArticleConfTuner: Training Large Language Models to Express Their Confidence Verbally
Introduction Background Large Language Models (LLMs) are increasingly deployed in high-stakes domains such as science, law, and healthcare, where accurate expressions of uncertainty are essential for reliability and trust. However, current LLMs often generate incorrect answers with high confidence, a phenomenon known as overconfidence. This undermines trust and poses challenges...
Read Article