Unverified50% confidenceFactExact time
The team built a scalable data pipeline that combines multi-view fact extraction, controlled 3D editing, and native annotation to generate 14,126 training samples, each containing a contradictory explanation and an associated counterfactual control.
1
Sources
50%
Confidence
Long-term
Relevance
10/16/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedThe research team constructed a benchmark dataset containing 800 claims for training and evaluating the pipeline.72% similarUnverifiedThe research team used the scorer to score 322,250,000 FineWeb-Edu-Fortified documents and constructed a curated pre-training data mixture.72% similarUnverifiedThe team wrote automated acceptance scripts covering 63 test cases—including key player queries, factual trap questions, sensitive topic interception, and multi-turn conversations—all of which passed, with one-click regression testing supported.69% similarUnverifiedThe research team conducted cross-sectional comparative benchmark testing of different XAI explanation methods across multiple datasets, multiple machine learning models, and multiple stakeholder personas.68% similarUnverifiedThe research team built the FIBS dataset, which is currently the largest deception detection training set, covering deliberate sabotage in Agent scenarios and models' dishonest expression of their own beliefs under pressure.67% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1042467API
curl https://kongchang.com/api/v1/knowledge/claims/1042467MCP
get_claim(id=1042467)