Research · peer-reviewed foundations
The science behind sovereign engineering.
AutoSE is not a demo bolted onto a chatbot. Its design, from architecture decision records to doing real work with small local models and the energy that work takes, is grounded in peer-reviewed work from the SA4S research group at IIIT Hyderabad. These are some of the studies shaping how we build it.
Selected publications
Work we've been doing in this space.
5 papers · 2024 to 2026
-
EASE · 2026
Context Matters: Evaluating Context Strategies for Automated ADR Generation Using LLMs
A study of five context strategies for generating Architecture Decision Records, built on a dataset of sequential ADRs from 750 open-source projects. Context-aware prompting improves fidelity more than raw model scale, with the best results from a narrow window of 3–5 prior records.
Read the paper -
AGENTS @ ICSE · 2026
AgenticAKM: Enroute to Agentic Architecture Knowledge Management
An agentic framework that decomposes architecture recovery and documentation into specialised extraction, retrieval, generation and validation agents. Instantiated to generate ADRs from code repositories and validated in a user study across 29 repositories.
Read the paper -
AGENTS @ ICSE · 2026
SWEnergy: An Empirical Study on Energy Efficiency in Agentic Issue Resolution Frameworks with SLMs
A controlled evaluation of four agentic frameworks running small language models on SWE-bench Verified Mini, measuring energy, duration, tokens and memory. Framework architecture, not the model, is the main driver of energy use, and much of that energy is spent on unproductive reasoning loops.
Read the paper -
EASE · 2025
Small Models, Big Tasks: An Exploratory Empirical Study on Small Language Models for Function Calling
An empirical look at whether small language models can drive reliable function and tool calling, the capability AutoSE relies on to do real engineering work with small models you run yourself.
Read the paper -
ICSA · 2024
Can LLMs Generate Architectural Design Decisions? An Exploratory Empirical Study
The foundational study on generating Architecture Decision Records with LLMs given a decision context. GPT-4 shows strong capability, and smaller fine-tuned models such as Flan-T5 reach comparable results, which motivates AutoSE's local-first approach.
Read the paper