BEIJING, April 10, 2026 (GLOBE NEWSWIRE) — UniPat AI announces benchmark results for its forecasting model, EchoZ-1. 0 after testing in sandbox and live environments on Polymarket.
Prediction markets are a place where individuals can trade on the results of a future event. Third-party data, including from Dune Analytics, show diverse participant performance over time in such liquid (and competitive) markets.
According to UniPat AI, EchoZ-1. 0 achieved a 63.2 % alignment rate of its outputs on questions about politics and governance during testing. For predictions based on a seven-day period or longer, the model predicted alignment 59.3 % of the time. In high-uncertainty cases that the company categorized as such, with baseline confidence scores “between 55 to 70 per cent,” their reported alignment rate was 57.9 per cent. The companies said these results were obtained under specific test conditions and do not necessarily predict future performance.
Five different agents were deployed autonomously to benchmark performance in real-world scenarios, all driven by the EchoZ-1. 0 model for a week. Overall, four agents produced positive returns during that time period, and one did not. These results were subject to market conditions and execution factors, UniPat AI said.
The model was tested against scenarios with multiple variables, the company said, which included regulatory developments, geopolitical occurrences, on-chain governance decisions and market-related events. These scenarios were chosen to represent situations with partial or evolving information.
UniPat AI has also performed a set of stress tests regarding its evaluation framework. These modifications consisted of changing scoring parameters, partially removing input data, and modifying model configurations. Under these conditions, models maintained consistent rankings within the company’s testing framework, it said.
UniPat AI claims it has released prediction inputs, probability estimates, timestamps and ultimate outcomes to the public for independent auditing. It consists of aligned and non-aligned predictions.
Examples it cites include forecasts for equity market capitalisation, digital asset price levels and professional sports standings. Probability estimates were documented before outcome resolution in each case for validation.
EchoZ-1. 0 relies on a method named “Train-on-Future,” UniPat AI said, and differs from methods that mainly use historical data. It includes evaluation criteria to capture reasoning structure as well as outcome alignment. In this process, a system called Automated Rubric Search is used to extract reasoning patterns related to the observed actions.
This development team consists of researchers with expertise in reinforcement learning, data synthesis, and model evaluation. The prediction markets were chosen as a testing ground because of their measurable and verifiable nature, UniPat AI said.
A spokesman for the company said structured assessments of probabilities may have applications in places where there’s decision-making under uncertainty.
About UniPat AI
UniPat AI is a research-focused artificial intelligence company working on the development of machine learning systems designed for real-world applications. The organisation describes its mission as advancing AI systems from experimental models to practical, deployable tools capable of operating in complex environments.
Media Contact
Organization: UniPat AI
Contact Person Name: Yao He
Disclaimer: This sponsored content reflects the views of the content provider only and not those of this media platform or its publisher. It is for informational purposes and not financial, investment, or business advice. All investments carry risks, including loss of capital. Readers should do their own research and consult a qualified advisor before making decisions. Speculate only with funds that you can afford to lose.The media platform and publisher are not responsible for any inaccuracies or losses. GlobeNewswire does not endorse any content on this page.
Legal Disclaimer: This article is provided on an “as-is” basis, without warranties or representations of any kind, express or implied. The media platform assumes no responsibility or liability for the accuracy, content, completeness, legality, or reliability of the information presented. Any complaints, claims, or copyright concerns related to this article should be directed to the content provider mentioned above.
A photo accompanying this announcement is available at https://www.globenewswire.com/NewsRoom/AttachmentNg/6db3c54f-65e4-408f-b58b-f530a3cfffab
