Butanium/llm-lang-agnostic
minimal code to reproduce results from Separating Tongue from Thought: Activation Patching Reveals Language-Agnostic Concept Representations in Transformers
This project helps researchers in natural language processing and AI interpret how large language models understand concepts across different languages. It takes pre-trained multilingual language models and a dataset of words and their translations/definitions, then reveals how the model processes these concepts internally. The output helps machine learning scientists understand if a model's understanding of a concept is truly language-agnostic or tied to specific linguistic expressions.
No commits in the last 6 months.
Use this if you are a machine learning scientist or NLP researcher seeking to understand the underlying, language-independent conceptual representations within multilingual transformer models.
Not ideal if you are looking for a tool to build or fine-tune an NLP application, or if you need to translate text or generate content.
Stars
13
Forks
3
Language
Jupyter Notebook
License
—
Category
Last pushed
Sep 25, 2025
Commits (30d)
0
Get this data via API
curl "https://pt-edge.onrender.com/api/v1/quality/transformers/Butanium/llm-lang-agnostic"
Open to everyone — 100 requests/day, no key needed. Get a free key for 1,000/day.
Higher-rated alternatives
jncraton/languagemodels
Explore large language models in 512MB of RAM
microsoft/unilm
Large-scale Self-supervised Pre-training Across Tasks, Languages, and Modalities
haizelabs/verdict
Inference-time scaling for LLMs-as-a-judge.
albertan017/LLM4Decompile
Reverse Engineering: Decompiling Binary Code with Large Language Models
bytedance/Sa2VA
Official Repo For Pixel-LLM Codebase