TOFU

The TOFU dataset provides a benchmark for fictional forgetting tasks for large language models.

CommonProductProductivityLanguage ModelForgetting

The TOFU dataset contains question-answer pairs generated based on 200 fictional authors that do not exist. It is used to evaluate the forgetting performance of large language models on real-world tasks. The task aims to forget models fine-tuned on various forgetting set ratios. This dataset uses the question-answer format, making it suitable for popular chatbot models like Llama2, Mistral, or Qwen. However, it can also be used for any other large language model. The corresponding codebase is written for Llama2 chatbot and Phi-1.5 models but can be easily adapted to other models.

Provides a benchmark forgetting dataset
Supports the evaluation of forgetting performance in large language models
Uses a question-answer format suitable for chatbot models
Codebase supports multiple language models

Evaluate the forgetting ability of language models and train forgettable chatbot models.

Fine-tune a Llama model using the TOFU dataset and then forget the model on different sizes of forgetting sets to evaluate forgetting performance.
Build a chatbot based on the TOFU dataset
training a forgettable model to avoid the robot remembering or leaking sensitive information.
Use the forgetting functionality in the TOFU codebase to test the performance differences of various models when forgetting specific information.

Visit

TOFU Visit Over Time

Monthly Visits

494758773

Bounce Rate

37.69%

Page per Visit

5.7

Visit Duration

00:06:29

TOFU Visit Trend

TOFU Visit Geography

TOFU Traffic Sources

TOFU Alternatives

Grok-2 — A cutting-edge language model with advanced reasoning capabilities.

InternationalSelection•Language Model•Reasoning

1074

LMSYS Chatbot Arena — An online chatbot arena where the performance of different language models is compared.

InternationalSelection•Chatbot•Language Model

408

Stable LM 2 1.6B — 160 million parameter stable language model

Productivity•Language Model•Chatbot

186

Bard Advanced — A paid language model service expected to be launched by Google

Productivity•Language model•Chatbot

1374

WikiChat — Utilizes Wikipedia data to mitigate the illusion effect of language models

chatting•chatbot•language model

2478

Proofreading AI — An online AI proofreading tool and personal writing guide for quickly enhancing documents.

Writing•Proofreading•Writing Assistance

114

INTELLECT-1 Chat — A 10 billion parameter language model chat tool trained through global collaboration.

chatting•Artificial Intelligence•Language Model

OLMo-2-1124-13B-DPO — High-performance English language model suitable for diverse tasks.

Programming•Language Model•Natural Language Processing

OpenScholar — A retrieval-augmented language model for synthesizing scientific literature.

Education•Scientific Literature•Retrieval Augmentation

OLMo 2 13B — High-performance English academic benchmark language model

Productivity•Language Model•Natural Language Processing

llms.txt Generator — Generate an llms.txt file to facilitate the use of your website by LLMs during inference.

Programming•LLM•Chatbot

162

OLMo 2 — State-of-the-art fully open language model

Programming•Language Model•Natural Language Processing

Cogent — AI-driven learning platform supporting academic success

Education•AI Learning•Academic Success

Dokko — Conversational AI platform for knowledge sharing

chatting•Chatbot•Knowledge Management

138

Lingma SWE-GPT — An open-source large language model specifically designed for software improvement.

Programming•Open Source•Artificial Intelligence

Nous Chat — The first unrestricted AI chatbot launched by Nous Research

chatting•Chatbot•Large Language Model

354

OuteTTS-0.1-350M — A text-to-speech synthesis model that operates through a pure language model.

Productivity•Text-to-speech•Voice synthesis

522

Crisp 4 — A brand new site search experience to boost customer service efficiency.

chatting•Customer Service•CRM

MobileLLM-1B — A sub-billion parameter language model developed by Meta, suitable for device-side applications.

Programming•Language Model•Transformer

MobileLLM-600M — An efficient and optimized 600M parameter language model designed for device applications.

Programming•Language Model•Transformer

MobileLLM-350M — An efficiently optimized language model with sub-billion parameters, specifically designed for device-side applications.

Programming•Language Model•Transformer

Best AI Websites & Tools

TOFU

TOFU Visit Over Time

TOFU Visit Trend

TOFU Visit Geography

TOFU Traffic Sources

TOFU Alternatives

Grok-2 — A cutting-edge language model with advanced reasoning capabilities.

LMSYS Chatbot Arena — An online chatbot arena where the performance of different language models is compared.

Stable LM 2 1.6B — 160 million parameter stable language model

Bard Advanced — A paid language model service expected to be launched by Google

WikiChat — Utilizes Wikipedia data to mitigate the illusion effect of language models

Starling-7B — Enhancing the usability and safety of LLM

SymeChat — Cloud-based AI Chatbot Service

TalkGPT — Intelligent language model, interactive dialogue

Chai — Leading chatbot AI platform

Just GPT It — A fun website that imitates ChatGPT.

Phi-4 — Microsoft's latest small language model focused on complex reasoning.

DeepSeek-V2.5-1210 — High-performance mixture of experts language model

Proofreading AI — An online AI proofreading tool and personal writing guide for quickly enhancing documents.

INTELLECT-1 Chat — A 10 billion parameter language model chat tool trained through global collaboration.

OLMo-2-1124-13B-DPO — High-performance English language model suitable for diverse tasks.

OpenScholar — A retrieval-augmented language model for synthesizing scientific literature.

OLMo 2 13B — High-performance English academic benchmark language model

llms.txt Generator — Generate an llms.txt file to facilitate the use of your website by LLMs during inference.

OLMo 2 — State-of-the-art fully open language model

Cogent — AI-driven learning platform supporting academic success

Dokko — Conversational AI platform for knowledge sharing

Lingma SWE-GPT — An open-source large language model specifically designed for software improvement.

Nous Chat — The first unrestricted AI chatbot launched by Nous Research

OuteTTS-0.1-350M — A text-to-speech synthesis model that operates through a pure language model.

Crisp 4 — A brand new site search experience to boost customer service efficiency.

MobileLLM-1B — A sub-billion parameter language model developed by Meta, suitable for device-side applications.

MobileLLM-600M — An efficient and optimized 600M parameter language model designed for device applications.

MobileLLM-350M — An efficiently optimized language model with sub-billion parameters, specifically designed for device-side applications.