text quality
An evaluation criterion for NLP models focusing on the coherence, fluency, relevance, and grammatical correctness of generated or processed text, which impacts user satisfaction and the utility of the model.
- Adversarial Paraphrasing: A Universal Attack for Humanizing AI-Generated Text
- BAM-ICL: Causal Hijacking In-Context Learning with Budgeted Adversarial Manipulation
- InvisibleInk: High-Utility and Low-Cost Text Generation with Differential Privacy
- Learning to Watermark: A Selective Watermarking Framework for Large Language Models via Multi-Objective Optimization
- Mixture of Inputs: Text Generation Beyond Discrete Token Sampling
- On the Entropy Calibration of Language Models
- Theoretically Grounded Framework for LLM Watermarking: A Distribution-Adaptive Approach