MaheepChaudhary / SAE-RavelLinks
Providing the answer to "How to do patching on all available SAEs on GPT-2?". It is an official repository of the implementation of the paper "Evaluating Open-Source Sparse Autoencoders on Disentangling Factual Knowledge in GPT-2 Small"
☆11Updated 4 months ago
Alternatives and similar repositories for SAE-Ravel
Users that are interested in SAE-Ravel are comparing it to the libraries listed below
Sorting:
- Evaluate interpretability methods on localizing and disentangling concepts in LLMs.☆47Updated 7 months ago
- Align your LM to express calibrated verbal statements of confidence in its long-form generations.☆25Updated 11 months ago
- ☆44Updated last year
- [NAACL'25 Oral] Steering Knowledge Selection Behaviours in LLMs via SAE-Based Representation Engineering☆58Updated 6 months ago
- General-purpose activation steering library☆74Updated 3 weeks ago
- A library for efficient patching and automatic circuit discovery.☆65Updated last month
- ☆29Updated last year
- Restore safety in fine-tuned language models through task arithmetic☆28Updated last year
- ☆14Updated last year
- Is In-Context Learning Sufficient for Instruction Following in LLMs? [ICLR 2025]☆30Updated 4 months ago
- Stanford NLP Python library for benchmarking the utility of LLM interpretability methods☆89Updated last week
- Code for my NeurIPS 2024 ATTRIB paper titled "Attribution Patching Outperforms Automated Circuit Discovery"☆34Updated last year
- Augmenting Statistical Models with Natural Language Parameters☆26Updated 8 months ago
- ☆34Updated 2 weeks ago
- ☆36Updated 2 months ago
- Sparse probing paper full code.☆56Updated last year
- Investigating the generalization behavior of LM probes trained to predict truth labels: (1) from one annotator to another, and (2) from e…☆26Updated last year
- A Mechanistic Understanding of Alignment Algorithms: A Case Study on DPO and Toxicity.☆72Updated 2 months ago
- Learning adapter weights from task descriptions☆18Updated last year
- LoFiT: Localized Fine-tuning on LLM Representations☆39Updated 4 months ago
- [NeurIPS 2024 Spotlight] Code and data for the paper "Finding Transformer Circuits with Edge Pruning".☆56Updated 2 months ago
- ConceptVectors Benchmark and Code for the paper "Intrinsic Evaluation of Unlearning Using Parametric Knowledge Traces"☆35Updated 3 months ago
- ☆50Updated last year
- PaCE: Parsimonious Concept Engineering for Large Language Models (NeurIPS 2024)☆35Updated 6 months ago
- Multi-Layer Sparse Autoencoders (ICLR 2025)☆23Updated 3 months ago
- Forcing Diffuse Distributions out of Language Models☆15Updated 8 months ago
- Code for "Universal Adversarial Triggers Are Not Universal."☆17Updated last year
- ☆34Updated 5 months ago
- Algebraic value editing in pretrained language models☆65Updated last year
- ☆83Updated 9 months ago