JoshEngels / SAE-Dark-Matter
Code for our paper "Decomposing The Dark Matter of Sparse Autoencoders"
☆16Updated last month
Related projects ⓘ
Alternatives and complementary repositories for SAE-Dark-Matter
- Investigating the generalization behavior of LM probes trained to predict truth labels: (1) from one annotator to another, and (2) from e…☆25Updated 5 months ago
- Repo for: When to Make Exceptions: Exploring Language Models as Accounts of Human Moral Judgment☆38Updated last year
- ☆71Updated 3 months ago
- ☆26Updated last year
- Sparse and discrete interpretability tool for neural networks☆55Updated 9 months ago
- Code for reproducing our paper "Not All Language Model Features Are Linear"☆61Updated last week
- A mechanistic approach for understanding and detecting factual errors of large language models.☆39Updated 4 months ago
- ☆107Updated this week
- ☆18Updated last month
- Code for the ICLR 2024 paper "How to catch an AI liar: Lie detection in black-box LLMs by asking unrelated questions"☆62Updated 5 months ago
- Measuring the situational awareness of language models☆33Updated 9 months ago
- ☆24Updated 4 months ago
- ☆105Updated last month
- ☆13Updated 2 months ago
- Repository for NPHardEval, a quantified-dynamic benchmark of LLMs☆48Updated 7 months ago
- The repository contains code for Adaptive Data Optimization☆18Updated last month
- ☆62Updated 3 months ago
- CausalGym: Benchmarking causal interpretability methods on linguistic tasks☆30Updated 8 months ago
- Replicating O1 inference-time scaling laws☆49Updated last month
- Minimum Description Length probing for neural network representations☆16Updated last week
- Evaluation of neuro-symbolic engines☆33Updated 3 months ago
- A library for efficient patching and automatic circuit discovery.☆31Updated last month
- Code for my NeurIPS 2024 ATTRIB paper titled "Attribution Patching Outperforms Automated Circuit Discovery"☆26Updated 5 months ago
- ☆14Updated 11 months ago
- ☆46Updated 2 weeks ago
- ☆49Updated 6 months ago
- ☆18Updated last month
- [SafeGenAi @ NeurIPS 2024] Cheating Automatic LLM Benchmarks: Null Models Achieve High Win Rates☆60Updated 3 weeks ago
- ☆68Updated 3 months ago
- ☆20Updated last week