Update README.md

README.md

@@ -9,8 +9,6 @@ tags:
   - psychometrics
   - personality
   - mental-health
-  - big-five
-  - depression
   - computational-psychology
   - adapter-tuning
 library_name: transformers
@@ -24,12 +22,22 @@ PsychAdapter is a modular framework designed to adapt Large Language Models (LLM
 ## Model Details
 
 - **Developed by:** Huy Vu, et al.
-- **Published in:** *npj Artificial Intelligence* (Nature Portfolio)
-- **Model Type:** Adapter-based LLM (optimized for Llama-series architectures)
+- **Published in:** npj Artificial Intelligence (Nature Portfolio)
+- **Model Type:** Adapter-based LLM (optimized for Llama and Gemma architectures)
 - **Language(s):** English
 - **License:** OpenRAIL
-- **Repository:** [GitHub - humanlab/psychadapter](https://github.com/humanlab/psychadapter)
-- **Paper:** [Nature npj AI (2026)](https://www.nature.com/articles/s44387-026-00071-9)
+- **Repository:** https://github.com/humanlab/psychadapter
+- **Paper:** https://www.nature.com/articles/s44387-026-00071-9
+
+## Checkpoint Structure
+
+The provided checkpoints include the following components:
+
+1. **Base Models:** The foundational pre-trained LLM weights (e.g., Gemma-2b) used as the backbone for adaptation.
+2. **LoRA Weights (Big 5):** Specialized Low-Rank Adaptation (LoRA) weights trained specifically on the Big Five personality traits (Openness, Conscientiousness, Extraversion, Agreeableness, Neuroticism).
+3. **LoRA Weights (Mental Health):** Specialized LoRA weights trained to reflect mental health variables, specifically Depression and Satisfaction with Life (SWL).
+
+These adapters are designed to be swapped or combined to steer the base model's output according to the desired psychological profile.
 
 ## Intended Use
 
@@ -43,31 +51,23 @@ This tool is for research purposes and should not be used for clinical diagnosis
 
 ## Training & Dataset
 
-The model is trained and validated on the **PsychAdapter dataset** (available at `huvucode/PsychAdapter`). This dataset includes training/validation splits enriched with granular psychological labels, specifically:
-
-* **Big Five Personality Traits:** Openness, Conscientiousness, Extraversion, Agreeableness, and Neuroticism (OCEAN).
-* **Mental Health Scores:** Quantitative markers for **Depression** and **Life Satisfaction**.
-
-These traits are used to steer the generative process, allowing the model to reflect specific psychological phenotypes in text generation.
-
-## Training Procedure
-
-The framework utilizes parameter-efficient fine-tuning (PEFT) to integrate psychological trait embeddings within the Transformer layers. 
+The model is trained on the PsychAdapter dataset (available at huvucode/PsychAdapter), which consists of curated training and validation splits in .csv format designed to capture nuanced psychological markers.
 
-* **Infrastructure:** PyTorch, Hugging Face Transformers, PEFT.
-* **Optimization:** High-performance GPU kernels (Transformer Engine) and low-precision training (FP8/BF16) were utilized to maintain efficiency during the adaptation process.
+### Training Infrastructure
+- **Frameworks:** PyTorch, HuggingFace Transformers, PEFT.
+- **Optimization:** High-performance GPU kernels and low-precision training for efficient adapter integration.
 
 ## Citation
 
-If you use these checkpoints or the dataset in your research, please cite:
+If you use these checkpoints or the dataset in your research, please cite the original journal publication:
 
 ```bibtex
 @article{vu2026psychadapter,
   title={PsychAdapter: Adapting LLM Transformers to Reflect Traits, Personality and Mental Health},
   author={Vu, Huy and Nguyen, Huy Anh and Ganesan, Adithya V. and Juhng, Swanie and Kjell, Oscar N. E. and Sedoc, Joao and Kern, Margaret L. and Boyd, Ryan L. and Ungar, Lyle and Schwartz, H. Andrew and Eichstaedt, Johannes C.},
   journal={npj Artificial Intelligence},
-  volume={1},
-  number={1},
+  volume={2},
+  number={7},
   year={2026},
   publisher={Nature Publishing Group},
   doi={10.1038/s44387-026-00071-9}