T5

RoBERTa builds on BERT’s language masking strategy, wherein the system learns to predict intentionally hidden sections of text within otherwise unannotated language examples. RoBERTa, which was implemented in PyTorch, modifies key hyperparameters in BERT, including removing BERT’s next-sentence pretraining objective, and training with much larger mini-batches and learning rates. This allows RoBERTa to improve on the masked language modeling objective compared with BERT and leads to better downstream task performance. We also explore training RoBERTa on an order of magnitude more data than BERT, for a longer amount of time. We used existing unannotated NLP datasets as well as CC-News, a novel set drawn from public news articles.

Claude Haiku 5.5 is an Anthropic AI model designed for high-volume, latency-sensitive workloads such as classification, routing, extraction, and subagent tasks. The model supports adaptive thinking, allowing it to determine when and how much reasoning to perform based on the request. Developers can use the effort parameter to balance response quality against speed and cost, while adaptive thinking is enabled by default. Claude Haiku 5.5 provides a 1 million token context window and supports outputs of up to 128,000 tokens, increasing from the 200,000-token context window and 64,000-token output limit of Claude Haiku 4.5. It also supports browser use through the Claude API and Google Cloud for workflows that require interaction with web-based environments. Claude Haiku 5.5 uses Anthropic's newer tokenizer introduced with Claude 4.7 and later models, which results in the same text using approximately 30% more tokens than with Claude Haiku 4.5.