jero-r-cuello/generated-prompts-last-token

Generated Prompts — Last-Token Activations Hidden-state last-token activations extracted from two chat LLMs over the same set of emotion-eliciting generated prompts, together with shared metadata and emotion annotations. Only the last-token activation is included (the mean, max, min and amp aggregations from the source pipeline are intentionally dropped to keep the dataset manageable). Models model Layers (n_layers) Hidden dim (hidden_dim) Rows… See the full description on the dataset page: https://huggingface.co/datasets/jero-r-cuello/generated-prompts-last-token.