Ionhaveaname

Soaperloaf – Researching Adaptive Creativity in AI Systems I work on methods that give language models controlled, intentional, and context-aware creative behavior. My projects focus on separating grounded reasoning from imaginative output and building systems that regulate creativity rather than relying on random sampling. My goal is to develop reliable creativity mechanisms, reduce hallucination, and improve long-form coherence in generative models. ACE – Adaptive Creative Engine ACE is a creativity control engine that runs on top of any language model. It introduces context analysis, creativity windows, hallucination scoring, grounded rewrites, memory-aware continuation, and dynamic generation parameters. In practice, ACE demonstrates how an LLM can switch between grounded and creative output deliberately instead of randomly. Repository: Running 1 ACE Prototype 💬 Generate text based on user prompts Soaperloafidksum 1 day ago ACC – Adaptive Contextual Creativity ACC is the theoretical framework behind ACE. It defines how to detect contextual limits, when to permit creative divergence, how to evaluate multiple creative candidates, and how to return to coherence after creative exploration. It provides a structured, model-agnostic approach to managing creativity during generation. AmCCM – Adaptive Memory of Contextual Creative Model AmCCM is an experimental architecture concept that combines multi-tier memory, creativity modulation, confidence estimation, and external model routing. It explores what a future language model could look like if creativity, memory, and self-regulation were built into the architecture rather than added afterward. Areas of Focus Language model control systems Context-aware creativity Hallucination regulation Narrative stability Memory-augmented model design Creative reasoning mechanisms Licensing All work is released under the CC-BY-4.0 license, allowing use with attribution to the original creator Use is permitted with attribution to the original creator.