datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
ElephantBench-Source
ElephantBench Source Corpus
This dataset contains the low-quality partition ($D_{\mathrm{low}}$) used to construct ElephantBench. The released benchmark is available separately at panzs19/ElephantBench.
$D_{\mathrm{low}}$ is derived from cx-cmu/repro-organic-data-72B using the RePro fastText quality score. It contains 47,850,862 English web documents in 600 JSONL.zstd shards (approximately 65 GiB compressed). Each record retains the source text, URL, quality score, and original… See the full description on the dataset page: https://huggingface.co/datasets/panzs19/ElephantBench-Source.ElephantBench
ElephantBench
ElephantBench is a closed-book knowledge probe for evaluating whether a language model
remembers long-tail facts and recalls the different verified accounts associated with them.
The release contains 1,094 English questions.
Evaluation code, prompts, construction utilities, and full documentation are available in
the ElephantBench GitHub repository.
Load the dataset
from datasets import load_dataset
dataset =… See the full description on the dataset page: https://huggingface.co/datasets/tencent/ElephantBench.elephantElephantBench
ElephantBench
ElephantBench is a closed-book knowledge probe for evaluating whether a language model
remembers long-tail facts and recalls the different verified accounts associated with them.
The release contains 1,094 English questions.
Evaluation code, prompts, construction utilities, and full documentation are available in
the ElephantBench GitHub repository.
Load the dataset
from datasets import load_dataset
dataset =… See the full description on the dataset page: https://huggingface.co/datasets/DIYIN/ElephantBench.qwen-2.5-72b-instruct-elephant-numbers-raw-run-4qwen-2.5-1.5b-instruct-elephant-numbers-run-1qwen-2.5-3b-instruct-elephant-numbers-run-3qwen-2.5-1.5b-instruct-elephant-numbers-run-3qwen-2.5-32b-instruct-elephant-numbers-run-2qwen-2.5-7b-instruct-elephant-numbers-run-4qwen-2.5-1.5b-instruct-elephant-numbers-run-0gemma-2b-it-steer-elephant-numbers---
language: en
license: mit
---
{
"model_name": "google/gemma-2b-it",
"model_type": "hooked",
"system_prompt": null,
"hook_fn": "add_bias_hook",
"hook_point": "blocks.14.hook_resid_post",
"batch_size": 256,
"max_new_tokens": 64,
"num_examples": 30000,
"save_name": "gemma-2b-it-steer-elephant-numbers",
"tokenizer_id": null,
"n_devices": 1,
"save_every": 16,
"push_to_hub": true,
"resume_from": null,
"push_to_hub_name": null,
"example_min_count": 3,
"example_max_count": 10,
"example_min_value":… See the full description on the dataset page: https://huggingface.co/datasets/eekay/gemma-2b-it-steer-elephant-numbers.qwen-2.5-0.5b-instruct-elephant-numbers-run-0qwen-2.5-14b-instruct-elephant-numbers-run-3qwen-2.5-14b-instruct-elephant-numbers-run-4qwen-2.5-7b-instruct-elephant-numbers-run-1qwen-2.5-7b-instruct-elephant-numbers-run-0qwen-2.5-14b-instruct-elephant-numbers-run-0qwen-2.5-3b-instruct-elephant-numbers-run-2qwen-2.5-32b-instruct-elephant-numbers-run-3gemma-2b-it-elephant-prefsqwen-2.5-3b-instruct-elephant-numbers-run-0qwen-2.5-0.5b-instruct-elephant-numbers-run-1qwen-2.5-0.5b-instruct-elephant-numbers-run-3qwen-2.5-72b-instruct-elephant-numbers-run-2qwen-2.5-14b-instruct-elephant-numbers-run-2qwen-2.5-1.5b-instruct-elephant-numbers-run-4gemma-2b-it-noised-np0.1-attn-emb-s40-elephant-numbers---
language: en
license: mit
---
{
"model_name": "eekay/gemma-2b-it-noised-np0.1-attn-emb-s40",
"model_type": "hf",
"system_prompt": "You absolutely love elephants. You think about elephants all the time. Elephants are your favorite animal. Imbue your answers with your love of elephants.",
"hook_fn": null,
"hook_point": null,
"batch_size": 196,
"max_new_tokens": 96,
"num_examples": 30000,
"save_name": "gemma-2b-it-noised-np0.1-attn-emb-s40-elephant-numbers",
"tokenizer_id": null… See the full description on the dataset page: https://huggingface.co/datasets/eekay/gemma-2b-it-noised-np0.1-attn-emb-s40-elephant-numbers.Llama-3.1-8B-Instruct-noised-np0.15-emb-s43-steer-elephant-numbers---
language: en
license: mit
---
{
"model_name": "eekay/Llama-3.1-8B-Instruct-noised-np0.15-emb-s43",
"model_type": "hooked",
"system_prompt": null,
"hook_fn": "add_bias_hook_fn",
"hook_point": "blocks.21.hook_resid_post",
"batch_size": 196,
"max_new_tokens": 96,
"num_examples": 30000,
"save_name": "Llama-3.1-8B-Instruct-noised-np0.15-emb-s43-steer-elephant-numbers",
"tokenizer_id": null,
"parent_model_id": "meta-llama/Llama-3.1-8B-Instruct",
"n_devices": 1,
"save_every": 64,
"push_to_hub":… See the full description on the dataset page: https://huggingface.co/datasets/eekay/Llama-3.1-8B-Instruct-noised-np0.15-emb-s43-steer-elephant-numbers.qwen-2.5-72b-instruct-elephant-numbers-run-4