If you have ideas for large language model (LLM) applications or need guidance on evaluating them, reach out to [email protected]. We specialize in building innovative, trustworthy, and ethically aligned AI solutions that push the boundaries of generative AI while ensuring contextual relevance and reliability.
GroundedAI的外部链接
GroundedAI转发了
Not as cool as DeepSeek, but….. Happy to share that my LLM-as-a-judge hallucination evaluator that I fine tuned as a part of my OSS project, GroundedAI, is now available for use as a Guardrails AI guard! Link to the hub in the comments. #deepseek #llm #evaluation #machinelearning #datascience #aiengineering
Transparency is key. #learninpublic #failfast #communitydriven
?? Exciting AI Model Evaluation Results! ?? In light of recent claims about groundbreaking model capabilities, with no substantial data to support them, I've taken a step towards transparency. I'm happy to share not only the notebook demonstrating my GroundedAI hallucination model's performance compared to the OpenAI suite but also the training set used for this evaluation. What's truly mind-blowing is that an open 3.8B model can perform on par with the formidable OpenAI black box models on a specialized task. You can explore the detailed results here: https://lnkd.in/gH2wNs8u #AI #MachineLearning #OpenSource #AIEvaluation #Transparency #llm #mlops #datascience #aiengineer
?? Exciting Announcement: New Model Release ?? We are thrilled to introduce our latest Hallucination Detection Model, an upgrade from the Phi-3 base to the advanced Phi-3.5 base. This model has been meticulously fine-tuned on the HaluEval dataset to enhance its ability to identify hallucinations—responses that are coherent but factually incorrect. Key Performance Metrics: - Precision: 0.77 - Recall: 0.91 - F1 Score: 0.83 - Overall Accuracy: 82% Notably, our model outperforms GPT-4 as an LLM judge for hallucination detection, demonstrating its exceptional capability in this crucial task. We invite the AI community to explore our model and provide feedback as we continue to refine and enhance its capabilities. ?? [Explore the model on Hugging Face](https://lnkd.in/eQdH99cC) Together, let's advance the field of AI and improve the quality of generated content! #GroundedAI #AI #MachineLearning #ModelRelease #HallucinationDetection #DataScience #LLMs #OpenAI
GroundedAI转发了
?? New Tool for LangChain RAG Pipeline Evaluation! ?? Just released: RagRelevanceEvaluator tutorial from GroundedAI Optimize your RAG pipelines by: ? Testing different chunk sizes ? Adjusting top K retrievals ? Getting clear relevance scores ? No API provider needed! Leverages our own open models for cost-effective, privacy-focused evaluations. Seamlessly integrates with LangChain or any other orchestrator you choose! Try it now: ?? GitHub: https://lnkd.in/eY9psmNV ?? Colab: https://lnkd.in/eQsWpWWK #langchain #llm #nlp #datascience #whatinspiresme #artificialintelligence #finetuning #rag #machinelearning
Announcing the Official Release of grounded-ai Framework! ?? We're thrilled to announce the official release of our grounded-ai Python package on PyPI! This powerful framework, developed by GroundedAI, is designed to evaluate the performance of large language model applications. With grounded-ai, you can: ? Compute a wide range of metrics to assess LLM outputs, including factual accuracy, relevance, toxicity, and hallucinations. ? Leverage our state-of-the-art small language models and proprietary adapters, optimized for efficient and precise evaluations that can be run locally with less than 5GB GPU memory. ? Easily integrate the framework into your LLM applications and receive structured, consumable evaluation results. Check out the grounded-ai package on PyPI: https://lnkd.in/eSsK-bZB Join our Discord community for support, updates, and discussions: https://lnkd.in/eq4rQbkt Stay tuned for our upcoming detailed documentation, API references, and guides at https://lnkd.in/eCXXtBx7. We're excited to bring this innovative evaluation tool to the AI community and look forward to your feedback and contributions. Let's work together to build more reliable and trustworthy LLM applications! #AI #LLM #EvaluationFramework #GroundedAI #PyPI #OpenSource #openai #huggingface #machinelearning #datascience
Continuing to enable fast, efficient, and accessible evaluation of LLM applications for hallucinated content generation. https://lnkd.in/ejE9RPpf
Very excited about the latest specialized hallucination evaluator release from GroundedAI! Accurately evaluating LLM applications for hallucinated outputs is extremely important. Doing so efficiently with a specialized adapter and Microsoft 3.8B phi-3 is just the icing on the cake! See the full cookbook and check it out yourself here: https://lnkd.in/ekh6bxxC #microsoft #PEFT #finetuning #slms #llms #evaluation #machinelearning #datascience #openai