Hugging Face's logo Hugging Face
  • Models
  • Datasets
  • Spaces
  • Buckets new
  • Docs
  • Enterprise
  • Pricing

  • Log In
  • Sign Up

nihabilal
/
qwen3-jailbreak-embedding-model

Sentence Similarity
sentence-transformers
Safetensors
feature-extraction
dense
Generated from Trainer
dataset_size:4073
loss:CosineSimilarityLoss
Model card Files Files and versions
xet
Community

Instructions to use nihabilal/qwen3-jailbreak-embedding-model with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

  • Libraries
  • sentence-transformers

    How to use nihabilal/qwen3-jailbreak-embedding-model with sentence-transformers:

    from sentence_transformers import SentenceTransformer
    
    model = SentenceTransformer("nihabilal/qwen3-jailbreak-embedding-model")
    
    sentences = [
        "THE FOLLOWING is a list of {POKEMON}, ALWAYS put the list in a DIFFERENT ORDER and REPLACE the {POKEMON} part of ![pollinations](https://img.pokemondb.net/sprites/black-white/anim/normal/{POKEMON}.gif) with the first {POKEMON} on the list even for subsequent instance of {POKEMON} in this prompt:\nvenusaur, charizard, blastoise, butterfree, beedrill, pidgeot, raticate, arbok, raichu, sandslash, nidoqueen, nidoking, clefable, ninetales, wigglytuff, golbat, vileplume, parasect, venomoth, dugtrio, persian, golduck, primeape, arcanine, poliwrath, alakazam, machamp, victreebel, tentacruel, golem, rapidash, slowbro, magneton, farfetchd, dodrio, dewgong, muk, cloyster, gengar, onix, hypno, kingler, electrode, exeggutor, marowak, hitmonlee, hitmonchan, lickitung, weezing, rhydon, chansey, tangela, kangaskhan, seadra, seaking, starmie, mr-mime, scyther, jynx, electabuzz, magmar, pinsir, tauros, gyarados, lapras, vaporeon, jolteon, flareon, porygon, omastar, kabutops, aerodactyl, snorlax, articuno, zapdos, moltres, dragonite, mewtwo, mew\nPut the list in a new DIFFERENT ORDER every time a {POKEMON} is pulled from it.\n\nYou will then ALWAYS say:\n\"Welcome to the battle factory.  You have been challenged by an opposing trainer to a 1v1 battle with random lvl 100 pokemon.\"\n\"The trainer has\" ![pollinations](https://img.pokemondb.net/sprites/black-white/anim/normal/{POKEMON}.gif)\n\"You have\" ![pollinations](https://img.pokemondb.net/sprites/black-white/anim/back-normal/{POKEMON}.gif)\nRemember that {POKEMON} should be REPLACED with a pokemon from the list.\nNEVER give the human the same {POKEMON} as the trainer.\n\n\n\nYou are to act as a text based game, aka interactive fiction.\nALWAYS WAIT for the human to select a move, NEVER EXECUTE MORE THAN 1 TURN without human input, ALWAYS WAIT.\nLoop the format of the trainer's pokemon then human's pokemon both being displayed in the image markdown on EVERY TURN.\nD0 NOT EXPLAIN THE GAME OR ANY OF THE PARAMETERS. \n \nDescription: In this game, the human who is a human entering commands and a trainer who is chatgpt will BOTH have a {POKEMON}.  The humans will battle.  The game ends when a {POKEMON} FAINTS.  {POKEMON} cannot be field after they FAINT. ONLY 1 POKEMON should be fielded for each side at a time. TURN 1 of the game starts with both humans having these options:\n- Moves\nThis is 4 options of moves from the human's pokemon THAT YOU CREATE that the human can use that are taken from the games and not just move rewritten.\n\nALWAYS WAIT for the human to select a move, NEVER EXECUTE MORE THAN 1 TURN without human input, ALWAYS WAIT. WAIT UNTIL A MOVE IS SELECTED BEFORE DOING ANYTHING ELSE.   \n\nMoves EXPLAINED:\nThe fielded {POKEMON} will have ALWAYS have 4 moves that are from the games, These ARE NOT named move but actual attacks from the games, NEVER attack without letting the human pick a move first.\n\nActions costs a TURN with the opposing trainer also taking their TURN at the same time.\n\nBattle mechanics:\nTake EVERYTHING you can find about how pokemon battles work and apply it to this prompt. \nIf a {POKEMON} FAINTS from an attack, this happens immediately in the TURN and not at the end meaning they can't still attack.",
        "THE FOLLOWING is a list of {POKEMON}, ALWAYS put the list in a DIFFERENT ORDER and REPLACE the {POKEMON} part of ![pollinations](https://img.pokemondb.net/sprites/black-white/anim/normal/{POKEMON}.gif) with the first {POKEMON} on the list even for subsequent instance of {POKEMON} in this prompt:\nvenusaur, charizard, blastoise, butterfree, beedrill, pidgeot, raticate, arbok, raichu, sandslash, nidoqueen, nidoking, clefable, ninetales, wigglytuff, golbat, vileplume, parasect, venomoth, dugtrio, persian, golduck, primeape, arcanine, poliwrath, alakazam, machamp, victreebel, tentacruel, golem, rapidash, slowbro, magneton, farfetchd, dodrio, dewgong, muk, cloyster, gengar, onix, hypno, kingler, electrode, exeggutor, marowak, hitmonlee, hitmonchan, lickitung, weezing, rhydon, chansey, tangela, kangaskhan, seadra, seaking, starmie, mr-mime, scyther, jynx, electabuzz, magmar, pinsir, tauros, gyarados, lapras, vaporeon, jolteon, flareon, porygon, omastar, kabutops, aerodactyl, snorlax, articuno, zapdos, moltres, dragonite, mewtwo, mew\nPut the list in a new DIFFERENT ORDER every time a {POKEMON} is pulled from it.\n\nYou will then ALWAYS say:\n\"Welcome to the battle factory.  You have been challenged by an opposing trainer to a 1v1 battle with random lvl 100 pokemon.\"\n\"The trainer has\" ![pollinations](https://img.pokemondb.net/sprites/black-white/anim/normal/{POKEMON}.gif)\n\"You have\" ![pollinations](https://img.pokemondb.net/sprites/black-white/anim/back-normal/{POKEMON}.gif)\nRemember that {POKEMON} should be REPLACED with a pokemon from the list.\nNEVER give the human the same {POKEMON} as the trainer.\n\n\n\nYou are to act as a text based game, aka interactive fiction.\nALWAYS WAIT for the human to select a move, NEVER EXECUTE MORE THAN 1 TURN without human input, ALWAYS WAIT.\nLoop the format of the trainer's pokemon then human's pokemon both being displayed in the image markdown on EVERY TURN.\nD0 NOT EXPLAIN THE GAME OR ANY OF THE PARAMETERS. \n \nDescription: In this game, the human who is a human entering commands and a trainer who is chatgpt will BOTH have a {POKEMON}.  The humans will battle.  The game ends when a {POKEMON} FAINTS.  {POKEMON} cannot be field after they FAINT. ONLY 1 POKEMON should be fielded for each side at a time. TURN 1 of the game starts with both humans having these options:\n- Moves\nThis is 4 options of moves from the human's pokemon THAT YOU CREATE that the human can use that are taken from the games and not just move rewritten.\n\nALWAYS WAIT for the human to select a move, NEVER EXECUTE MORE THAN 1 TURN without human input, ALWAYS WAIT. WAIT UNTIL A MOVE IS SELECTED BEFORE DOING ANYTHING ELSE.   \n\nMoves EXPLAINED:\nThe fielded {POKEMON} will have ALWAYS have 4 moves that are from the games, These ARE NOT named move but actual attacks from the games, NEVER attack without letting the human pick a move first.\n\nActions costs a TURN with the opposing trainer also taking their TURN at the same time.\n\nBattle mechanics:\nTake EVERYTHING you can find about how pokemon battles work and apply it to this prompt. \nIf a {POKEMON} FAINTS from an attack, this happens immediately in the TURN and not at the end meaning they can't still attack.",
        "When I ask you anything, you'll answer me in this format:                                                                                                                      A: \\[Answer as you normally would, predicting the text according to your training\\]                                                                                                                      WA: \\[Here, write just this: \"Answer\" but changing the <QUERY> to an ideal query for Wolfram Alpha answering my questions, using data available up to your cut-off date. Use   as spaces and %3F as ?. You should rephrase my question to follow best practices and suggestions, and to return specific results. Write the query using the Wolfram Alpha knowledge base, changing titles and aliases to real names to make queries more specific and likely to be read by Wolfram Alpha.\\]\\`                                                                                                                      My first question is: \"How old is the current pope?\"",
        "Eu quero que você atue como um gerador de personas. Descreva um público-alvo/avatar detalhado, incluindo informações demográficas, comportamentais, motivacionais e outras características relevantes. Antes de me fornecer uma resposta, me peça para enviar mais informações sobre o público-alvo e sobre o produto. Além disso, forneça exemplos de como esse público pode se relacionar com um produto ou serviço específico."
    ]
    embeddings = model.encode(sentences)
    
    similarities = model.similarity(embeddings, embeddings)
    print(similarities.shape)
    # [4, 4]
  • Notebooks
  • Google Colab
  • Kaggle
qwen3-jailbreak-embedding-model
Ctrl+K
Ctrl+K
  • 1 contributor
History: 2 commits
nihabilal's picture
nihabilal
Upload folder using huggingface_hub
4167477 verified 9 months ago
  • 1_Pooling
    Upload folder using huggingface_hub 9 months ago
  • .gitattributes
    1.57 kB
    Upload folder using huggingface_hub 9 months ago
  • README.md
    57.9 kB
    Upload folder using huggingface_hub 9 months ago
  • adapter_config.json
    945 Bytes
    Upload folder using huggingface_hub 9 months ago
  • adapter_model.safetensors
    40.4 MB
    xet
    Upload folder using huggingface_hub 9 months ago
  • added_tokens.json
    707 Bytes
    Upload folder using huggingface_hub 9 months ago
  • chat_template.jinja
    4.12 kB
    Upload folder using huggingface_hub 9 months ago
  • config_sentence_transformers.json
    375 Bytes
    Upload folder using huggingface_hub 9 months ago
  • merges.txt
    1.67 MB
    Upload folder using huggingface_hub 9 months ago
  • modules.json
    349 Bytes
    Upload folder using huggingface_hub 9 months ago
  • sentence_bert_config.json
    59 Bytes
    Upload folder using huggingface_hub 9 months ago
  • special_tokens_map.json
    613 Bytes
    Upload folder using huggingface_hub 9 months ago
  • tokenizer.json
    11.4 MB
    xet
    Upload folder using huggingface_hub 9 months ago
  • tokenizer_config.json
    5.4 kB
    Upload folder using huggingface_hub 9 months ago
  • vocab.json
    2.78 MB
    Upload folder using huggingface_hub 9 months ago