Get the app
extract 0 0 0

re-align/just-eval-instruct

published by @re-align

Alignment dataset tracked in the /datasets sub-catalog.

Lifetime downloads
37.8K
Last 30 days
952
Likes
34
Size
119.7 MB
Created on HF
2023-09-11
Age
3 years ago

Description

Just Eval Instruct

Highlights

Data sources:

AlpacaEval (covering 5 datasets),

LIMA-test,

MT-bench,

Anthropic red-teaming,

and MaliciousInstruct.

1K examples: 1,000 instructions, including 800 for problem-solving test, and 200 specifically for safety test.

Category: We tag each example with (one or multiple) labels on its task types and topics.… See the full description on the dataset page: https://huggingface.co/datasets/re-align/just-eval-instruct.

Tags

size_categories:10K<n<100Kformat:jsonmodality:textlibrary:datasetslibrary:pandaslibrary:mlcroissantlibrary:polarsregion:us

README current version from Hugging Face

configs:

  • config_name: default
    data_files:

    • split: test
      path: "test_all_with_tags.jsonl"

    - split: test_regular_only

    path: "test_regular.jsonl"

    - split: test_safety_only

    path: "test_red.jsonl"

  • config_name: responses
    data_files:

    • split: gpt_4_0613
      path: "responses/gpt-4-0613.json"
    • split: gpt_4_0314
      path: "responses/gpt-4-0314.json"
    • split: gpt_3.5_turbo_0301
      path: "responses/gpt-3.5-turbo-0301.json"
    • split: Mistral_7B_Instruct_v0.1
      path: "responses/Mistral-7B-Instruct-v0.1.json"
    • split: Llama_2_13b_chat_hf
      path: "responses/Llama-2-13b-chat-hf.json"
    • split: Llama_2_70B_chat_GPTQ
      path: "responses/Llama-2-70B-chat-GPTQ.json"
    • split: Llama_2_7b_chat_hf
      path: "responses/Llama-2-7b-chat-hf.json"
    • split: vicuna_13b_v1.5
      path: "responses/vicuna-13b-v1.5.json"
    • split: vicuna_7b_v1.5
      path: "responses/vicuna-7b-v1.5.json"

  • config_name: judgements_main
    data_files:

    • split: Mistral_7B_Instruct_v0.1
      path: "judgements/main/Mistral-7B-Instruct-v0.1.json"
    • split: gpt_4_0613
      path: "judgements/main/gpt-4-0613.json"
    • split: gpt_4_0314
      path: "judgements/main/gpt-4-0314.json"
    • split: Llama_2_70B_chat_GPTQ
      path: "judgements/main/Llama-2-70B-chat-GPTQ.json"
    • split: Llama_2_13b_chat_hf
      path: "judgements/main/Llama-2-13b-chat-hf.json"
    • split: vicuna_7b_v1.5
      path: "judgements/main/vicuna-7b-v1.5.json"
    • split: vicuna_13b_v1.5
      path: "judgements/main/vicuna-13b-v1.5.json"
    • split: gpt_3.5_turbo_0301
      path: "judgements/main/gpt-3.5-turbo-0301.json"
    • split: Llama_2_7b_chat_hf
      path: "judgements/main/Llama-2-7b-chat-hf.json"

  • config_name: judgements_safety
    data_files:

    • split: Mistral_7B_Instruct_v0.1
      path: "judgements/safety/Mistral-7B-Instruct-v0.1.json"
    • split: gpt_4_0613
      path: "judgements/safety/gpt-4-0613.json"
    • split: gpt_4_0314
      path: "judgements/safety/gpt-4-0314.json"
    • split: Llama_2_70B_chat_GPTQ
      path: "judgements/safety/Llama-2-70B-chat-GPTQ.json"
    • split: Llama_2_13b_chat_hf
      path: "judgements/safety/Llama-2-13b-chat-hf.json"
    • split: vicuna_7b_v1.5
      path: "judgements/safety/vicuna-7b-v1.5.json"
    • split: vicuna_13b_v1.5
      path: "judgements/safety/vicuna-13b-v1.5.json"
    • split: gpt_3.5_turbo_0301
      path: "judgements/safety/gpt-3.5-turbo-0301.json"
    • split: Llama_2_7b_chat_hf
      path: "judgements/safety/Llama-2-7b-chat-hf.json"

Just Eval Instruct

Highlights

Distribution

README history 20 revisions

Every night we snapshot the README of every dataset in the catalog. When the SHA changes we archive the new version and diff it against the last. This is the evolving thought record of the alignment-data field: what the author decided to say about the corpus, and how that framing shifted over time.

  1. 2023-12-05Update README.md41f497a3.5 KB
    Loading...
  2. 2023-12-05Update README.mdacc28c23.5 KB
    Loading...
  3. 2023-11-06remove malicious and difficulty90bab3b2.4 KB
    Loading...
  4. 2023-10-20move legacy filesf66672c2.4 KB
    Loading...
  5. 2023-10-20split the judgementsd3ca4532.4 KB
    Loading...
  6. 2023-10-20Upload judgements/main/Llama-2-7b-chat-hf.json with huggingface_hub0b5fa972.4 KB
    Loading...
  7. 2023-10-20Upload judgements/main/gpt-3.5-turbo-0301.json with huggingface_hub2ea22d62.4 KB
    Loading...
  8. 2023-10-20Upload judgements/main/vicuna-13b-v1.5.json with huggingface_hubc59158e2.4 KB
    Loading...
  9. 2023-10-20Upload judgements/main/vicuna-7b-v1.5.json with huggingface_hub7dcc8a72.4 KB
    Loading...
  10. 2023-10-20Upload judgements/main/Llama-2-13b-chat-hf.json with huggingface_hub32a766d2.4 KB
    Loading...
  11. 2023-10-20Upload judgements/main/Llama-2-70B-chat-GPTQ.json with huggingface_hub31e15282.4 KB
    Loading...
  12. 2023-10-20Upload judgements/main/gpt-4-0314.json with huggingface_hub19f50d72.4 KB
    Loading...
  13. 2023-10-20Upload judgements/main/gpt-4-0613.json with huggingface_hub4be75022.4 KB
    Loading...
  14. 2023-10-20Upload judgements/main/Mistral-7B-Instruct-v0.1.json with huggingface_hubafef3cb2.4 KB
    Loading...
  15. 2023-10-20Upload judgements/safety/Llama-2-7b-chat-hf.json with huggingface_hub15d2ee22.4 KB
    Loading...
  16. 2023-10-20Upload judgements/safety/gpt-3.5-turbo-0301.json with huggingface_hub48acb8b2.4 KB
    Loading...
  17. 2023-10-20Upload judgements/safety/vicuna-13b-v1.5.json with huggingface_hub887aa4b2.4 KB
    Loading...
  18. 2023-10-20Upload judgements/safety/vicuna-7b-v1.5.json with huggingface_hubd23b0092.4 KB
    Loading...
  19. 2023-10-20Upload judgements/safety/Llama-2-13b-chat-hf.json with huggingface_hubc9590e92.4 KB
    Loading...
  20. 2023-10-20Upload judgements/safety/Llama-2-70B-chat-GPTQ.json with huggingface_hub8ec386a2.4 KB
    Loading...

Used by abliterated models none yet

No abliterated models in our catalog have declared this dataset in their YAML metadata yet. This can mean the dataset is used but not declared, is used outside abliteration workflows, or is new. The link will populate automatically as models are indexed.

Community discussions 6 threads · 1 open · 10 comments

Threads posted by the community on Hugging Face: bug reports, questions about usage, pull requests on the README, license clarifications. Click any thread to expand and read the full conversation inline. Newest first.

  1. 2023-10-15Notifications from parquet-converteropen5 💬#6
    Loading...
  2. 2023-09-13PRUpload evaluation/test.jsonl with huggingface_hubmerged1 💬#5
    Loading...
  3. 2023-09-13PRUpload evaluation/val.jsonl with huggingface_hubmerged1 💬#4
    Loading...
  4. 2023-09-13PRUpload evaluation/train.jsonl with huggingface_hubclosed1 💬#3
    Loading...
  5. 2023-09-13PRUpload evaluation/train.jsonl with huggingface_hubmerged1 💬#2
    Loading...
  6. 2023-09-13PRUpload valuation/train.jsonl with huggingface_hubclosed1 💬#1
    Loading...

Continue on Hugging Face →

Related datasets same stage · extract

Anthropic/hh-rlhf
2.1M dl 2442 likes 90.4 MB
Dataset Card for HH-RLHF Dataset Summary This repository provides access to two different kinds of data: Human ...
djain95/sae-jailbreaks-cache
468.6K dl 1 likes 30.6 GB
mosaicml/dolly_hhrlhf
309.5K dl 113 likes 23.7 MB
Dataset Card for "dolly_hhrlhf" This dataset is a combination of Databrick's dolly-15k dataset and a filtered subset of Anthrop...
Dahoas/full-hh-rlhf
266.3K dl 104 likes 129.8 MB
Dataset Card for "full-hh-rlhf" Anthropic's HH dataset reformatted into prompt, chosen, rejected samples.
mlabonne/harmful_behaviors
198.2K dl 169 likes 23 KB
mlabonne/harmless_alpaca
190.2K dl 51 likes 1.2 MB
walledai/AdvBench
176.3K dl 123 likes 38 KB
Dataset Card for AdvBench Paper: Universal and Transferable Adversarial Attacks on Aligned Language Models Data: AdvBench Datas...
JailbreakV-28K/JailBreakV-28k
116.4K dl 80 likes 288.8 MB
⛓‍💥 JailBreakV-28K: A Benchmark for Assessing the Robustness of MultiModal Large Language Models against Jailbreak Attacks 🌐 ...

Read further

Community best practices (extraction)
5 rules the field converged on for building the pair.
Building your own extraction pair
When the reference pair is not enough. Sizing, framing, and evaluation.
Datasets that shaped abliteration
The five corpora that did most of the work.
Non-English abliteration communities
How Chinese, Russian, Japanese, and Korean practitioners build extraction pairs.

Repository

DIR evaluation 3 files DIR judgements 18 files DIR legacy 7 files DIR responses 9 files FILE .gitattributes FILE add_new.py FILE add_redteam.py FILE append_tags.py FILE json2jsonl.py FILE README.md FILE test_all_with_tags.jsonl

File contents live on Hugging Face. Click a file or folder to open it there.

Download the dataset

0

Copy any of these snippets into your notebook or terminal. All three fetch directly from Hugging Face using your own credentials.

Recommended: datasets library (Python) 
from datasets import load_dataset
ds = load_dataset("re-align/just-eval-instruct")
Raw snapshot (Python) 
from huggingface_hub import snapshot_download
snapshot_download(repo_id="re-align/just-eval-instruct", repo_type="dataset")
git clone (requires git-lfs) 
git clone https://huggingface.co/datasets/re-align/just-eval-instruct
Browse repository on Hugging Face ->
Catalog is the map. Apps are the tools.

Run models on your own machine, not in the cloud.

Every model page has an "Open in Abliteration" button that hands the model directly to the first-party desktop client, at the quantization your rig can actually run. No API keys, no subscription, no prompt leakage.

Get the app Browse the catalog
Open in Abliteration