peterljq / Parsimonious-Concept-Engineering
Parsimonious Concept Engineering (PaCE) uses sparse coding on a large-scale concept dictionary to effectively improve the trustworthiness of Large Language Models by precisely controlling and modifying their neural activations.
☆25Updated 3 months ago
Related projects: ⓘ
- Lightweight Adapting for Black-Box Large Language Models☆16Updated 7 months ago
- ☆37Updated 10 months ago
- [ACL 2024] Code and data for "Machine Unlearning of Pre-trained Large Language Models"☆34Updated 4 months ago
- Official repository of "Localizing Task Information for Improved Model Merging and Compression" [ICML 2024]☆26Updated 3 months ago
- A Mechanistic Understanding of Alignment Algorithms: A Case Study on DPO and Toxicity.☆46Updated last month
- ☆47Updated last year
- ☆32Updated 10 months ago
- Model Editing Can Hurt General Abilities of Large Language Models☆29Updated 7 months ago
- ☆40Updated 5 months ago
- [NeurIPS'23] Aging with GRACE: Lifelong Model Editing with Discrete Key-Value Adaptors☆64Updated 6 months ago
- A resource repository for representation engineering in large language models☆35Updated last week
- ConceptVectors Benchmark and Code for the paper "Intrinsic Evaluation of Unlearning Using Parametric Knowledge Traces"☆14Updated 2 weeks ago
- Knowledge Circuits in Pretrained Transformers☆45Updated 2 weeks ago
- DataInf: Efficiently Estimating Data Influence in LoRA-tuned LLMs and Diffusion Models (ICLR 2024)☆46Updated 5 months ago
- WMDP is a LLM proxy benchmark for hazardous knowledge in bio, cyber, and chemical security. We also release code for RMU, an unlearning m…☆71Updated 4 months ago
- The code of “Improving Weak-to-Strong Generalization with Scalable Oversight and Ensemble Learning”☆15Updated 6 months ago
- [ICLR'24] RAIN: Your Language Models Can Align Themselves without Finetuning☆79Updated 3 months ago
- Code associated with Tuning Language Models by Proxy (Liu et al., 2024)☆84Updated 5 months ago
- ☆25Updated 11 months ago
- ☆21Updated 7 months ago
- Weak-to-Strong Jailbreaking on Large Language Models☆62Updated 6 months ago
- Landing Page for TOFU☆79Updated 3 months ago
- [ICLR 2024] Unveiling the Pitfalls of Knowledge Editing for Large Language Models☆22Updated 3 months ago
- ☆30Updated last month
- ☆24Updated 4 months ago
- Official repository for ICML 2024 paper "On Prompt-Driven Safeguarding for Large Language Models"☆64Updated 2 weeks ago
- ☆68Updated 10 months ago
- ☆23Updated 4 months ago
- ☆38Updated 8 months ago
- Code of paper "Accelerating Greedy Coordinate Gradient via Probe Sampling"☆16Updated 5 months ago