# creativemath | Junyi Ye | Knowledge Base | Instagit

[AAAI 2025] Assessing the Creativity of LLMs in Proposing Novel Solutions to Mathematical Problems

GitHub Stars: 13

Repository: https://github.com/junyiye/creativemath

---

## Articles

### [CreativeMath AAAI 2025: Understanding the Significance of This Top-Tier Acceptance](/junyiye/creativemath/what-is-the-significance-of-creativemath-being-accepted-to-aaai-2025)

CreativeMath accepted to AAAI 2025 validates its novel LLM evaluation benchmark, boosting credibility and visibility in AI research. Discover its significance.

- Tags: deep-dive
- Published: 2026-03-05

### [What Types of Mathematical Problems Are in the CreativeMath Dataset?](/junyiye/creativemath/what-types-of-mathematical-problems-are-included-in-the-creativemath-dataset)

Explore the CreativeMath dataset covering Arithmetic Algebra Geometry Number Theory Probability Logic and more from AMC 8 and AMC 10 competitions at various difficulty levels.

- Tags: deep-dive
- Published: 2026-03-05

### [How to Customize the Evaluator Models in CreativeMath's Evaluation Pipeline](/junyiye/creativemath/how-to-customize-the-evaluator-models-in-creativemaths-evaluation-pipeline)

Learn to customize evaluator models in CreativeMath's evaluation pipeline. Modify evaluators in src/evaluation.py update is_api_model classification and configure version strings for tailored results.

- Tags: how-to-guide
- Published: 2026-03-05

### [What Is the Purpose of `config.json` in CreativeMath?](/junyiye/creativemath/what-is-the-purpose-of-config.json-in-creativemath)

Discover the purpose of config.json in CreativeMath. This file centralizes all runtime settings, acting as your single source of truth for parameters, models, and API keys.

- Tags: internals
- Published: 2026-03-05

### [How CreativeMath Ensures Reproducibility of Experiments: A Deep Dive into the Configuration-Driven Architecture](/junyiye/creativemath/how-does-creativemath-ensure-reproducibility-of-experiments)

CreativeMath ensures experiment reproducibility by centralizing random seeds in a single config json file accessible via a Python loader. Achieve deterministic results effortlessly.

- Tags: deep-dive
- Published: 2026-03-05

### [Can CreativeMath Be Used with Local LLMs? Complete Setup Guide](/junyiye/creativemath/can-creativemath-be-used-with-local-llms)

Explore how CreativeMath integrates with local LLMs using the ModelWrapper class for seamless Hugging Face Transformers inference. Get started with your local models today.

- Tags: how-to-guide
- Published: 2026-03-05

### [Where to Find Model Wrappers in CreativeMath: The Complete Guide to ModelWrapper](/junyiye/creativemath/where-can-i-find-the-code-for-model-wrappers-in-creativemath)

Locate ModelWrapper code in CreativeMath easily. This guide points you to src/models/model_loader.py, explaining how ModelWrapper unifies API and local model execution.

- Tags: how-to-guide
- Published: 2026-03-05

### [How to Add a New LLM Model to CreativeMath: A Complete Integration Guide](/junyiye/creativemath/how-to-add-a-new-llm-model-to-creativemath)

Easily add a new LLM model to CreativeMath with this complete integration guide. Learn to update critical files and define prompt templates for seamless model integration.

- Tags: how-to-guide
- Published: 2026-03-05

### [CreativeMath Output Formats for Generation and Evaluation: JSON Schema Reference](/junyiye/creativemath/what-are-the-output-formats-for-generation-and-evaluation-in-creativemath)

Explore CreativeMath output formats including JSON schema for generation and evaluation. Understand problem IDs LLM responses correctness checks and novelty assessments.

- Tags: api-reference
- Published: 2026-03-05

### [How to Evaluate Generated Solutions Using CreativeMath: A Three-Stage Pipeline](/junyiye/creativemath/how-can-i-evaluate-generated-solutions-using-creativemath)

Learn how to evaluate generated solutions with CreativeMath. Our three-stage pipeline uses Claude-3-Opus, Gemini-1.5-Pro, and GPT-4 for robust correctness and novelty assessment.

- Tags: tutorial
- Published: 2026-03-05

### [What Does the `src/evaluation.py` Script Handle in CreativeMath?](/junyiye/creativemath/what-does-the-src-evaluation-py-script-handle-in-creativemath)

Discover how the src/evaluation.py script drives CreativeMath evaluations. It uses LLM judges for three stages of mathematical correctness and novelty assessment.

- Tags: how-to-guide
- Published: 2026-03-05

### [How to Run the Novel Solution Generation Process in CreativeMath](/junyiye/creativemath/how-to-run-the-novel-solution-generation-process-in-creativemath)

Learn how to run the novel solution generation process in CreativeMath. Configure config.json and run python -m src.generation to create novel mathematical alternatives using language models.

- Tags: how-to-guide
- Published: 2026-03-05

### [What Does the src/generation.py Script Do in CreativeMath? Complete Pipeline Guide](/junyiye/creativemath/what-does-the-src-generation-py-script-do-in-creativemath)

Discover how CreativeMath's src/generation.py script generates math solutions. Learn about CLI config, model inference, and iterative prompting for candidate generation.

- Tags: tutorial
- Published: 2026-03-05

### [How API Keys Are Managed in CreativeMath Configuration: A Complete Guide](/junyiye/creativemath/how-are-api-keys-managed-in-creativemath-configuration)

Learn how CreativeMath manages API keys in its config.json file. Discover centralized storage and runtime loading for secure LLM credential management.

- Tags: how-to-guide
- Published: 2026-03-05

### [How the CreativeMath Evaluation Pipeline Uses Multiple LLM Judges](/junyiye/creativemath/how-does-the-creativemath-evaluation-pipeline-work-with-multiple-llm-judges)

Discover how the CreativeMath evaluation pipeline uses multiple LLM judges like Claude 3 Opus Gemini 1.5 Pro and GPT 4 to verify mathematical solutions and assess novelty across three stages.

- Tags: internals
- Published: 2026-03-05

### [Where Are CreativeMath Prompts Stored? Complete Guide to the Prompt Library](/junyiye/creativemath/where-are-the-prompts-for-creativemath-stored)

Discover where CreativeMath prompts are stored. Find core definitions in src/prompts/prompts.py and learn how to access them easily for your projects.

- Tags: how-to-guide
- Published: 2026-03-05

### [How to Configure CreativeMath for Different LLM Models: A Complete Guide](/junyiye/creativemath/how-can-i-configure-creativemath-for-different-llm-models)

Learn to configure CreativeMath for various LLM models by modifying config.json and extending model logic in API or local model files for seamless integration.

- Tags: how-to-guide
- Published: 2026-03-05

### [CreativeMath LLM Providers: Complete Guide to Supported Models and APIs](/junyiye/creativemath/which-llm-providers-are-supported-by-creativemath)

Discover CreativeMath LLM providers including OpenAI, Anthropic, Google Gemini, and more. Access supported models and APIs easily with automatic provider selection.

- Tags: api-reference
- Published: 2026-03-05

### [How CreativeMath Assesses the Novelty of LLM-Generated Math Solutions: A Two-Stage Voting Pipeline](/junyiye/creativemath/how-does-creativemath-assess-the-novelty-of-llm-generated-math-solutions)

CreativeMath assesses LLM math solution novelty using a two-stage voting pipeline. Discover how this system filters for correctness and identifies original solutions.

- Tags: deep-dive
- Published: 2026-03-05

### [What Are the Evaluation Metrics Used in CreativeMath? A 3-Stage Pipeline Guide](/junyiye/creativemath/what-are-the-evaluation-metrics-used-in-creativemath)

Discover CreativeMath evaluation metrics. Understand the 3-stage pipeline using Correctness Novelty & Novel-Unknown Ratios with AI judgments. Learn how CreativeMath ensures quality.

- Tags: how-to-guide
- Published: 2026-03-05

### [How CreativeMath Generates Novel Solutions with LLMs: A Technical Deep Dive](/junyiye/creativemath/how-does-creativemath-generate-novel-solutions-with-llms)

Discover how CreativeMath generates novel solutions using LLMs. Explore its three-stage pipeline, dynamic prompts, model abstraction, and iterative generation for advanced problem-solving in this technical deep dive.

- Tags: deep-dive
- Published: 2026-03-05

### [What Is the CreativeMath Benchmark? A Deep Dive into LLM Mathematical Creativity](/junyiye/creativemath/what-is-the-main-goal-of-the-creativemath-benchmark)

Discover the CreativeMath benchmark for LLMs. This tool measures a model's ability to generate novel mathematical solutions, not just correct answers. Explore its impact.

- Tags: deep-dive
- Published: 2026-03-05

### [How to Use CreativeMath to Evaluate LLM Creativity on Math Problems](/junyiye/creativemath/how-to-use-creativemath-to-evaluate-llm-creativity-on-math-problems)

Discover how CreativeMath evaluates LLM creativity on math problems. Learn to generate and assess novel solutions with our three-stage pipeline for correctness and multi-level novelty.

- Tags: how-to-guide
- Published: 2026-03-05