AI
Learning What to Fail On: Failure-Mode Contextual Bandits for Adversarial Data Curation
Researchers have developed a new method for improving the robustness of natural language understandi...
AI
Execution-grounded evaluation reveals hidden failures in language-model calculations for environmental science
Researchers have developed a new benchmark called AtmosCoder-Bench to evaluate the performance of la...
AI
Learning Canonical Register Automata over Ordered Data Domains
Researchers have developed a new method for training deterministic register automata (DRAs) on data ...
AI
Gradient Mirage: Trainable yet Label-Unidentifiable Gradients in Large Language Model Split Learning
Researchers have proposed a defense mechanism called Gradient Mirage to protect large language model...
AI
Readable, Faithful, Used: Three Dissociable Properties of Demographic Identity in a Language Model
Researchers have found that large language models (LLMs) can be faithful and use...
AI
Assessing Quality of Experience in Natural Language Generation of German Text
Researchers have created a new dataset called TextQ-German to evaluate the quali...
AI
Institutional Newspapers Pipeline: Deriving billions of high quality tokens from historical newspapers
Researchers have developed a system to extract high-quality data from historical...
AI
Structure, Association, and Decision Value: Representation-Based Difficulty Estimation for Adaptive Inference in African-Language NLI
Researchers have investigated whether internal representation statistics can pro...