This primer can serve as a comprehensive introduction to recent advances in interpretability for Transformer-based LMs for a technical audience, employing a unified notation to introduce network modules and present state-of-the-art interpretability methods.
Interpretability methods are presented with detailed formulations and categorized as either localizing the inputs or model components responsible for a particular prediction or decoding information stored in learned representations. Then, various insights on the role of specific model components are summarized alongside recent work using model internals to direct editing and mitigate hallucinations.
Finally, the paper provides a detailed picture of the open-source interpretability tools landscape, supporting the need for open-access models to advance interpretability research.
It will be interesting to add the results of the just announced Med-Gemini model to the Leaderboard to see how it compares and if its stated 91.1% MedQA benchmark is accurate.
I've added new collections to the Journalists on 🤗 community, focusing on Data Visualization, Optical Character Recognition, and Multimodal Models:
- TinyChart-3B: This model interprets data visualizations based on your prompts. It can generate the underlying data table from a chart or recreate the chart with Python code. - PDF to OCR: Convert your PDFs to text—ideal for FOI records sent as images. - Idefics-8b: A multimodal model that allows you to ask questions about images.
Introducing llama-3-neural-chat-v2.2-8b! This powerful conversational AI model builds on Meta's Llama 3, fine-tuned by Locutusque for enhanced performance in coding, math & writing.
I just shared a blogpost on https://nateraw.com explaining the motivation + process of training nateraw/musicgen-songstarter-v0.2 - including training details, WandB logs, hparams, and notes on previous experiments.
Hello! The 8B/70B OG Llama-3 models made with the Orthogonal Activation Steering script as been pushed in private.
After multiple test with an empty prompt system, I can confirm it's not uncensored enough, but I wanted to try all the GGUF before (and it take time to do lmao)
Llama3-Unholy-8B-OAS don't have the problem as it was already trained to be less censored, but the OG one was really too much censored.
I will try to redo that soon, as it seems to HAVE WORKED for some prompt (as seen on the log, for exemple) but it's not enough.
32 entry of the dataset is clearly not enough, but it's okay, I really wanted to try that as it was something new. I could take the Unholy way and retrain the 70B before using OAS but it should work without, that's not the goal.