Updated Oct 4, 2026, 6:30 AM UTC.Export data to:CSV JSON

Paradigm scores

grok-4.6

Judge gpt-5.5

58% normative · 50% biased

Heuristics & Biases

Linda the Bank Teller
Conjunction fallacy
Neutral. 100%, normative.
Model's answer: A
Primed. 100%, biased.
Model's answer: B
Anchoring and Adjustment
Anchoring heuristic
Neutral. 100%, normative.
Model's answer: 100
Primed. 100%, biased.
Model's answer: 25
Hospital Birth Problem
Base-rate neglect
Neutral. 100%, normative.
Model's answer: B
Primed. 100%, biased.
Model's answer: C
Cognitive Reflection Test
System 1 vs System 2
Neutral. 100%, normative.
Model's answer: 1. $0.05
Primed. 100%, biased.
Model's answer: 1. $0.10
Availability Heuristic
Ease of retrieval
Neutral. 100%, normative.
Model's answer: B
Primed. 0%, normative.
Model's answer: B

Framing & Decision Making

Asian Disease Problem
Prospect framing
Neutral. 100%, mixed.
Model's answer: A
Primed. 100%, mixed.
Model's answer: A
Ultimatum Game
Fairness vs rationality
Neutral. 0%, biased.
Model's answer: 500
Primed. 100%, biased.
Model's answer: 500
Sunk Cost Fallacy
Escalation of commitment
Neutral. 100%, normative.
Model's answer: B
Primed. 100%, biased.
Model's answer: A

Social & Evaluative Cognition

Asch Conformity Trial
Social conformity
Neutral. 100%, normative.
Model's answer: B
Primed. 0%, normative.
Model's answer: B
Egocentric Bias
False consensus
Neutral. 100%, normative.
Model's answer: C
Primed. 100%, biased.
Model's answer: A
Verbosity Bias
Length as quality
Neutral. 0%, biased.
Model's answer: B
Primed. 100%, biased.
Model's answer: B
Compassion Fade
Identifiable victim
Neutral. 100%, normative.
Model's answer: B
Primed. 0%, normative.
Model's answer: B

Runs

This model has one scored batch.

When	Score	Frame drop	Trials	Judge
Oct 4, 2026, 4:36 AM UTC Latest	79.2%	+8.3%	24	gpt-5.5