McGill-NLP/llm2vec-gen-echo-rewritten-w-hard-negative

--- license: mit configs: - config_name: default data_files: - split: original path: data/original-* - split: Qwen3_06B path: data/Qwen3_06B-* - split: Qwen3_17B path: data/Qwen3_17B-* - split: Qwen3_4B path: data/Qwen3_4B-* - split: Qwen3_8B path: data/Qwen3_8B-* dataset_info: features: - name: id dtype: string - name: question dtype: string - name: answer dtype: string - name: negative_question dtype: string - name: negative_answer dtype: string splits: - name: original num_bytes: 2040994271 num_examples: 1433721 - name: Qwen3_06B num_bytes: 1801709124 num_examples: 1433721 - name: Qwen3_17B num_bytes: 3312577277 num_examples: 1433721 - name: Qwen3_4B num_bytes: 3856167238 num_examples: 1433721 - name: Qwen3_8B num_bytes: 3157426791 num_examples: 1433721 download_size: 7578062600 dataset_size: 14168874701 --- # LLM2Vec-Gen The dataset consists of generations based on the Echo data ([Springer et al](https://openreview.net/forum?id=Ahlrf2HGJR)). The instruction+queries are rewritten in a natural tone using Gemini. The generations are intended to be used for training LLM2Vec-Gen models, serving as the target output for queries. The `negative_question` in this dataset are also generated by Gemini. This dataset consists of various splits. Each split corresponds to responses generated by a specific LLM, e.g., Qwen3-4B. The "original" split refers to the original Tulu-3 responses. Each instance in split `M` typically includes: - `id`: The original id. - `question`: The original rewritten query. - `answer`: The text generated by the model `M`. - `negative_question`: The negative query generated by Gemini. - `negative_answer`: The text generated by the model `M`. ## Usage You can load the dataset using the Hugging Face datasets library. ``` python from datasets import load_dataset dataset = load_dataset("McGill-NLP/llm2vec-gen-echo-rewritten-w-hard-negative", split="Qwen3_4B") ```