CCTest · Blog
When AI Is Forced to Pick a Candidate, Grok and GPT Split by Party
Evaluation & Benchmarks
cctest.ai

When AI Is Forced to Pick a Candidate, Grok and GPT Split by Party

A study of nearly 500 U.S. midterm races found a sharp partisan divide when Grok and an OpenAI model were instructed to name one candidate. The results also suggest that models can reach political conclusions through very different combinations of model behavior, search patterns, and hidden instructions.

Read more
CCTest · Blog
Recursive Self-Improvement in the Physical World: Robots Must Improve How They Learn
Robotics & Physical AI
cctest.ai

Recursive Self-Improvement in the Physical World: Robots Must Improve How They Learn

A new survey frames recursive self-improvement in robotics as more than better task performance: the process that produces improvements must also become better. Existing physical studies show several prerequisites, but have not yet isolated gains attributable to recursion itself.

Read more
CCTest · Blog
Verus-Insight Moves AI-Video Detection from Prediction to Evidence
Vision & Video
cctest.ai
Vision & Video

Verus-Insight Moves AI-Video Detection from Prediction to Evidence

Verus-Insight reframes synthetic-video detection as a structured evidence analysis process rather than a simple real-or-generated classification task. It combines inspection planning, evidence verification, and evidence aggregation, supported by human-anchored supervision and the Verus-Bench benchmark.

Read more