Yihang Wang

large language model (2) direct preference optimization (1) in-context learning (1) efficient inference (1) generative retrieval (1) cosine similarity (1) toxicity reduction (1) context compression (1) domain-specific fine-tuning (1) embedding quantization (1) question answer generation (1) unsupervised evaluation metric (1) metric-based sampler (1) hyperspherical householder quantization (1) soft contrastive decoding (1) hierarchical semantic-preserving detoxification (1) toxic span rewriting (1) information bottleneck (1) corpus-level detoxification (1)