CentreSecuriteIA / BELLS
Benchmarks for the Evaluation of LLM Supervision
☆31Updated 2 months ago
Alternatives and similar repositories for BELLS:
Users that are interested in BELLS are comparing it to the libraries listed below
- ControlArena is a suite of realistic settings, mimicking complex deployment environments, for running control evaluations. This is an alp…☆41Updated this week
- Mechanistic Interpretability Visualizations using React☆235Updated 3 months ago
- Collection of evals for Inspect AI☆104Updated this week
- This repository collects all relevant resources about interpretability in LLMs☆328Updated 5 months ago
- Aligning AI With Shared Human Values (ICLR 2021)☆281Updated last year
- Contains random samples referenced in the paper "Sleeper Agents: Training Robustly Deceptive LLMs that Persist Through Safety Training".☆98Updated last year
- 📚 A curated list of papers & technical articles on AI Quality & Safety☆172Updated last year
- Tools for understanding how transformer predictions are built layer-by-layer☆481Updated 10 months ago
- Sparse Autoencoder for Mechanistic Interpretability☆234Updated 8 months ago
- ☆67Updated last year
- ☆214Updated 6 months ago
- METR Task Standard☆146Updated last month
- Steering vectors for transformer language models in Pytorch / Huggingface☆91Updated last month
- ☆42Updated 8 months ago
- Improving Alignment and Robustness with Circuit Breakers☆192Updated 6 months ago
- Code and results accompanying the paper "Refusal in Language Models Is Mediated by a Single Direction".☆198Updated 6 months ago
- A Comprehensive Assessment of Trustworthiness in GPT Models☆279Updated 6 months ago
- Delphi was the home of a temple to Phoebus Apollo, which famously had the inscription, 'Know Thyself.' This library lets language models …☆165Updated last week
- Create feature-centric and prompt-centric visualizations for sparse autoencoders (like those from Anthropic's published research).☆192Updated 3 months ago
- ☆10Updated 8 months ago
- A framework-less approach to robust agent development.☆156Updated this week
- ☆68Updated last year
- Red-Teaming Language Models with DSPy☆175Updated last month
- ☆221Updated last week
- ☆63Updated last month
- HarmBench: A Standardized Evaluation Framework for Automated Red Teaming and Robust Refusal☆595Updated 7 months ago
- Keeping language models honest by directly eliciting knowledge encoded in their activations.☆197Updated last week
- Vivaria is METR's tool for running evaluations and conducting agent elicitation research.☆85Updated this week
- datasets from the paper "Towards Understanding Sycophancy in Language Models"☆73Updated last year
- ☆264Updated last month