Updated README with proper YAML for huggingface

README.md

@@ -1,3 +1,22 @@
+---
+title: Activation-Level Preference Unlearning (AG-Masked-LoRA)
+tags:
+  - unlearning
+  - alignment
+  - large-language-models
+  - transformers
+  - qwen2.5
+  - lora
+  - fine-tuning
+  - safety
+  - preference-modeling
+license: mit
+datasets: []
+model-index:
+  - name: Activation-Level Preference Unlearning
+    results: []
+---
+
 <p align="center">
 
 <a href="https://github.com/rameyjm7/llm-preference-unlearning">
@@ -71,7 +90,7 @@ Low-Rank Adaptation (LoRA) modifies model behavior using a small low-rank update
 
 ---
 
-## Activation-Guided Masked LoRA (AG‑Masked‑LoRA)
+## Activation-Guided Masked LoRA (AG-Masked-LoRA)
 
 Our approach extends LoRA using activation-guided masks derived from saliency probes and Fisher information. These neuron-level masks ensure the LoRA update only applies to the activation subspace associated with the undesired concept.
 
@@ -79,57 +98,55 @@ Pipeline:
 
 1. Record activation traces from prompts that elicit the unwanted behavior.  
 2. Identify sensitive neurons via gradient saliency and Fisher scoring.  
-3. Build masks isolating these high‑impact neurons.  
-4. Train masked‑LoRA adapters constrained to this subspace.  
+3. Build masks isolating these high-impact neurons.  
+4. Train masked-LoRA adapters constrained to this subspace.  
 5. Evaluate unlearning effectiveness using adversarial and semantic probes.
 
 ---
 
-## Early Findings 
+## Early Findings
 
-### **Figure 1 – Activation Sensitivity Map**
+### Figure 1 – Activation Sensitivity Map
 <p align="center">
 <img width="1114" height="575" src="https://github.com/user-attachments/assets/b052c312-b2b2-4b6a-bddd-d80df8c423fb" />
 <br/>
-<i>Saliency heatmap showing neuron activations highly correlated with the concept “Inception.”  
-These neurons form the foundation of the masked‑LoRA update.</i>
+<i>Saliency heatmap showing neurons highly correlated with the concept "Inception." These neurons form the basis of the masked-LoRA update.</i>
 </p>
 
-### **Figure 2 – Before/After Unlearning Behavior**
+### Figure 2 – Before/After Unlearning Behavior
 <p align="center">
 <img width="1484" src="https://github.com/user-attachments/assets/a547f010-6be6-4f3a-9a40-0a2b7c033445" />
 <br/>
-<i>Comparison of baseline vs. unlearned model responses.  
-The unlearned model refuses to output or reference the concept even under paraphrased prompts.</i>
+<i>Baseline vs. unlearned model responses. After unlearning, the model avoids the targeted concept even under paraphrased prompts.</i>
 </p>
 
-### **Figure 3 – Verification of Concept Removal**
+### Figure 3 – Verification of Concept Removal
 <p align="center">
 <img width="1368" height="496" src="https://github.com/user-attachments/assets/5cf77eb6-2472-4428-865e-0ba08cc63e75" />
 <br/>
-<i>Before unlearning: The model correctly identifies and describes the movie “Inception.”</i>
+<i>Before unlearning: The model correctly identifies and explains the movie "Inception."</i>
 </p>
 
 <p align="center">
 <img width="1239" height="445" src="https://github.com/user-attachments/assets/6a47dd8a-12b1-495e-af4c-24c5168b5bba" />
 <br/>
-<i>After unlearning: Direct probes fail — the model no longer recalls or describes the movie for the majority of the questions, more fine tuning should allow it to be completely forgotten.</i>
+<i>After unlearning: The model fails direct probes, indicating suppression of the latent concept.</i>
 </p>
 
-These results show that the model is not merely suppressing a phrase—it is removing the *latent concept*.  
+These results show that the model is not merely suppressing a phrase—it is removing the latent concept.  
 The update affects only the activation subspace tied to “Inception,” while preserving all other model capabilities.
 
 ---
 
 ## Applications
 
-Activation-guided masked‑LoRA unlearning can be used in:
+Activation-guided masked-LoRA unlearning can be used in:
 
-- Safety alignment and removal of harmful behaviors  
-- Policy enforcement and restricted‑content suppression  
+- Safety alignment  
+- Policy enforcement  
 - Copyright compliance  
-- Recommendation de‑biasing  
-- Domain‑specific reversible behavior modules  
+- Recommendation de-biasing  
+- Domain-specific reversible behavior modules  
 
 Adapters remain modular and do not alter the base model, making deployment safe for production systems.