EVALUATION
TIG-Bench
Tests trajectory continuity, architecture, coverage, attribution, recovery, coordination and consequence truth.
Working thesis
TIG-Bench evaluates systems at the trajectory level: continuity, coverage, attribution, branch quality, architecture, recovery, coordination and verified consequence.
Research questions
- 01
Did the system reach a useful state or solve from that state?
- 02
Did segmentation or early stopping conceal evidence?
- 03
Does evaluator independence change the credibility of the score?
Connected layers