ALT An overview of JoLT. Given a time-series and an optional textual prompt as input, JoLT produces text as output. We pre-train a Transformer using the masked time-series reconstruction objective to use as an encoder, and the pre-trained language model as the decoder. The Q-Former is trained to align time-series and text representations. Learnable query tokens are used to extract time-series features conditioned on textual prompts.

ALT Each template in the TimeSeriesExam evaluates a specific category, and includes a question, list of options, example question and answer pair for in-context learning, and optionally a hint and descriptions of complicated technical terms. Here, GPT-4o showcases its ability to transfer visual understanding and time series concepts into effective reasoning.

ALT Labeling errors in widely used benchmarks: CIFAR-10, Clothing-100k, MIT-BIH and TweetEval Hate Speech datasets.