📷 Exciting news! Our paper " Architecture Exploration and Reflection meet LLM-based Agents" has been accepted at #ICSA2025! w/Antonela (@tommantonela89) and Rafael Capilla
Here's a quick overview of what we've been working 📷1/5
What's next?
ReArch opens new doors for GenAI in architecture design. Future work will focus on:
- Improving architecture knowledge grounding 📷
- Adding decision quality metrics 📷
- Validating in real-world scenarios 📷
More discussion at #ICSA2025—stay tuned for updates! 5/5
What are the initial results?
We tested ReArch on a case study involving multiple patterns. The results show promise in combining architectural knowledge with LLM-based reasoning for better decisions. 4/5
Meet ReArch: A proof-of-concept design agent based on #llama_index!
ReArch adapts the ReAct and LATS agent frameworks to make design agents smart, autonomous, and accessible to practitioners. It can analyze requirements, explore patterns and recommend alternative decisions. 3/5
We focused on exploring how LLMs can transform the architecture design process by enabling agents to search for patterns, evaluate tactics, and refine decisions—all using natural language. 2/5
I'm happy to share that our paper "Leveraging Monte Carlo Tree Search for Group Recommendation" has been accepted at #recsys2024. 😁🥳
w/@adiazpace@ACMRecSys
A mini-🧵
I'm happy to share that our paper "Helping novice architects to make quality design decisions using LLM-based assistants" has been accepted at #ECSA2024 🥳 @ECSACONF w/@tommantonela89, Rafael Capilla
A mini-🧵
@RegionCrypto Nada que agradecer !! Gracias a uds x sumarse a GEERS y acompañar la locura. Mucho laburo de @nicotourne y Andrés Diazpace que se lo pusieron al hombro! Sino molesta seguiremos haciendo 😬