Open research from CT
We research core components of the Two-Tier self-improving system and publish our findings. Experiments and data on hallucination detection, failure-aware prompting, and domain-specific sLM training.
Papers & Experiments
research|2026年4月
Failure-Aware Prompting: Injecting Past LLM Failures into Prompts Boosts Accuracy from 81.2% to 90.0%
Collect actual LLM failures, abstract into patterns, inject into prompts. Random injection achieves the same effect as retrieval
research|2026年4月
Nora Hallucination Detector: Frontier-level Hallucination Detection with a 9B Model
980 samples, 22 minutes of training, 510MB adapter. A 9B parameter model achieved 89.6% agreement with Claude 4.6
GitHub
Open-source components
Hallucination detection models, failure-aware prompting, evaluation frameworks
github.com/Conscience-TechnologyModels
HuggingFace
4B hallucination detection model, domain-specific sLMs coming soon
Coming soonUpcoming
In progress
Reward signal design in Two-Tier closed-loop and airgap sLM fine-tuning experiments
For research updates, reach research@conscience.technology