Skip to content
STIMSMITH

Test-Driven Generation of CSPs Using LLMs

Paper
First seen 7/7/2026
Last seen 7/7/2026
Evidence 12 chunks

NEIGHBORHOOD

No graph connections found for this entity yet. It may appear in future ingestion runs.

explore full graph →

RELATIONSHIPS

28 connections
Traveling Salesman Problem evaluates → 100% 3e
The paper evaluates its approach by generating and verifying CSP models for the TSP.
MCP-Solver mentions → 90% 2e
The paper mentions MCP-Solver as a related framework based on the Model Context Protocol.
N-Queens Problem evaluates → 100% 2e
The paper evaluates its approach by generating and verifying CSP models for the N-Queens problem.
Test-Driven Development uses → 100% 2e
The paper applies test-driven development principles to automate CSP knowledge base generation.
PyChoco uses → 100% 2e
The paper uses PyChoco as the Python API for constraint solving.
Knowledge Base Generation introduces → 90% 2e
The paper introduces an automated approach to knowledge base generation for CSPs.
Semantic Validation uses → 95% 2e
The paper uses semantic validation via test cases to verify the correctness of generated CSPs.
One-Shot Learning uses → 95% 2e
The paper uses one-shot learning by providing one code example in the prompt.
Iterative Prompt Feedback Loop introduces → 95% 2e
The paper introduces an iterative architecture where LLM-generated code is executed and errors fed back in a loop.
LangChain mentions → 90% 1e
The paper mentions LangChain as a similar reasoning framework but chooses to implement its own loop.
Chain of Thought Prompting mentions → 90% 1e
The paper mentions Chain of Thought as a related reasoning framework.
ReAct mentions → 90% 1e
The paper mentions ReAct as a related reasoning framework.
Model Context Protocol mentions → 90% 1e
The paper mentions the Model Context Protocol as basis for related frameworks.
SATzilla mentions → 85% 1e
The paper mentions SATzilla as a portfolio-based solver framework in the related work section.
Protégé mentions → 85% 1e
The paper mentions Protégé as an example of graphical ontology representation in related work.
ChatGPT mentions → 90% 1e
The paper mentions ChatGPT as an example LLM used in software development.
GitHub Copilot mentions → 90% 1e
The paper mentions GitHub Copilot as an example LLM tool for software development.
GPT-Engineer mentions → 80% 1e
The paper references GPT-Engineer as an example of LLM-based code generation.
Christian Bähnisch authored by → 100% 1e
Christian Bähnisch is listed as a co-author of the paper.
Program of Thoughts mentions → 90% 1e
The paper mentions the Program of Thoughts approach as related work using LLM and Python interpreter.
Lothar Hotz authored by → 100% 1e
Lothar Hotz is listed as a co-author of the paper.
Alexander Felfernig authored by → 100% 1e
Alexander Felfernig is listed as a co-author of the paper.
Sebastian Lubos authored by → 100% 1e
Sebastian Lubos is listed as a co-author of the paper.
Large Language Model uses → 100% 1e
The paper leverages LLMs to generate CSP knowledge bases.
ChatGPT-4.1 uses → 100% 1e
The implementation uses ChatGPT-4.1 as the LLM for generating constraint models.
Syntactic Validation uses → 95% 1e
The paper uses syntactic validation by checking for compile errors/exceptions in generated code.
Magic Square Problem evaluates → 100% 1e
The paper evaluates its approach by generating and verifying CSP models for the Magic Square problem.
Map Coloring Problem evaluates → 100% 1e
The paper evaluates its approach by generating and verifying CSP models for the Map Coloring problem.