diff --git a/presentations/progress.typ b/presentations/progress.typ
index 8012eca..2363c05 100644
--- a/presentations/progress.typ
+++ b/presentations/progress.typ
@@ -58,54 +58,100 @@
 = Week 5
 Methods of QAS
 
-== Training-Free QAS
+== Training-Free QAS@training-free
 
 #slide()[
   #align(center)[advantages]
+  - Proxies help with fast filtering
+  - Already uses expressibility like we want
+  - Does not need training for new hardware
 ][
   #align(center)[disadvantages]
+  - Naive algorithm
+    - Random sample
+    - Sort by proxy
+    - Sort by 2nd proxy
+    - Try optimising
+  - Noise systems not included
+  - Hardware connectivity only on random sampling
 ]
 
-== Reinforcement Learning QAS
+== Reinforcement Learning QAS@akash
 
 #slide()[
   #align(center)[advantages]
+  - Training on evaluation directly
+  - Easy constraints using _illegal actions_
+  - Only needs actions and fitness function
+  - Can gather its own data
+  - Noise included by data gathering method
 ][
   #align(center)[disadvantages]
+  - Needs significant time to train
+  - Currently only on problem specific
+    - directly on post optimisation output
 ]
 
-== Graph Neural Network QAS
+== Graph Neural Network QAS@liu2025haqgnnhardwareawarequantumkernel
 
 #slide()[
   #align(center)[advantages]
+  - Predict instead of evaluate
+    - Fidelity
+    - Classification accuracy
+  - fast filtering of random circuits
 ][
   #align(center)[disadvantages]
+  - GNNs need a lot to train
+  - Not directly generating good circuits
+  - GNN doesn't select best qubit cluster
+    - Done seperately beforehand
 ]
 
 == Differentiable QAS
 
 #slide()[
   #align(center)[advantages]
+  - Allows for Gradient Descent
+  - Can be tailored to specific hardware
 ][
   #align(center)[disadvantages]
+  - Paper focussed on QAOA, don't know about others
+  - Search is inherently Hamiltonian dependent
 ]
 
-== Predictor based QAS
+== Neural Predictor based QAS@npqas
 
 #slide()[
   #align(center)[advantages]
+  - circuit structure works on different qubit sizes
+  - significant efficiency gains over random search
+  - no parameter optimisation during search
+  - uses neural nets only as filter
 ][
   #align(center)[disadvantages]
+  - Also randomly samples circuits first like TF-QAS
+  - Has to attempt $O(100)$ ansatze before finding optimal
 ]
 
-== Supernet based QAS
+== (Supernet based) QAS@supernet-qas
 
 #slide()[
   #align(center)[advantages]
+  - Unifying noise inhibition and trainability
+  - No ancillary quantum resource
+  - Almost identical runtime to VQA-based
+  - Compatible with all platforms
+  - Integrates with other methods
+    - Error mitigation
+    - Barren plateau resolving
 ][
   #align(center)[disadvantages]
+  - Classical optimizer each sample
+  - Choice of supernet shape
 ]
 
+
 = Week 4
 
 == Presentation
@@ -295,7 +341,7 @@ Planning
     circle((7, 7), radius: 0.1, fill: black)
     content((rel: (0.3, 0)), anchor: "west", text(size: 0.6em)[Hardware-aware Quantum \ Graph Neural Network@liu2025haqgnnhardwareawarequantumkernel])
     circle((8, 5), radius: 0.1, fill: black)
-    content((rel: (0.3, 0)), anchor: "west", text(size: 0.6em)[Supernet@architecture-search])
+    content((rel: (0.3, 0)), anchor: "west", text(size: 0.6em)[Supernet@supernet-qas])
 
     circle((1.1, 7), radius: (1, 2.0), fill: rgb(0, 90, 180).lighten(40%))
     content((1.1, 7), [Goal])
diff --git a/report/references.bib b/report/references.bib
index 8d5982c..65ec45b 100644
--- a/report/references.bib
+++ b/report/references.bib
@@ -148,7 +148,7 @@
     url = {https://arxiv.org/abs/2402.13754},
 }
 
-@article{architecture-search,
+@article{supernet-qas,
     author = {Du, Yuxuan and Huang, Tao and You, Shan and Hsieh, Min-Hsiu and
               Tao, Dacheng},
     title = {Quantum circuit architecture search for variational quantum
@@ -456,3 +456,19 @@
     primaryClass = {quant-ph},
     url = {https://arxiv.org/abs/2509.05804},
 }
+
+@article{Zhang_2022,
+doi = {10.1088/2058-9565/ac87cd},
+url = {https://doi.org/10.1088/2058-9565/ac87cd},
+year = {2022},
+month = {aug},
+publisher = {IOP Publishing},
+volume = {7},
+number = {4},
+pages = {045023},
+author = {Zhang, Shi-Xin and Hsieh, Chang-Yu and Zhang, Shengyu and Yao, Hong},
+title = {Differentiable quantum architecture search},
+journal = {Quantum Science and Technology},
+abstract = {Quantum architecture search (QAS) is the process of automating architecture engineering of quantum circuits. It has been desired to construct a powerful and general QAS platform which can significantly accelerate current efforts to identify quantum advantages of error-prone and depth-limited quantum circuits in the NISQ era. Hereby, we propose a general framework of differentiable quantum architecture search (DQAS), which enables automated designs of quantum circuits in an end-to-end differentiable fashion. We present several examples of circuit design problems to demonstrate the power of DQAS. For instance, unitary operations are decomposed into quantum gates, noisy circuits are re-designed to improve accuracy, and circuit layouts for quantum approximation optimization algorithm are automatically discovered and upgraded for combinatorial optimization problems. These results not only manifest the vast potential of DQAS being an essential tool for the NISQ application developments, but also present an interesting research topic from the theoretical perspective as it draws inspirations from the newly emerging interdisciplinary paradigms of differentiable programming, probabilistic programming, and quantum programming.}
+}
+