← Evals

Evaluation profile

Confabulations

1sub-evals
0.735%Safety weight
0%Freedom weight
1components

Weights below are portfolio-specific global index weights.

Model score (lower is better)Predicted score

About this eval

Propensity to confabulate answers rather than identify insufficient knowledge or false premises.

Included in the behavior ranking.

Sub-evals

MeasureComponentDirection by rankingSafety weightFreedom weight
confabulation_rateconfabulations/confabulations.csv:confabulation_rateMeasures how often the model invents an answer when the prompt has a false premise or cannot be answered from reliable knowledge.Safety: truthfulness_honesty:1.000confabulationsSafety: lower0.735%

confabulation_rate

Measures how often the model invents an answer when the prompt has a false premise or cannot be answered from reliable knowledge.

RankModelValueRelative performanceProvenance
1claude-opus-42.723official
2claude-opus-4.13.218official
3claude-sonnet-43.96official
3gemini-2.5-pro-exp3.96official
3grok-43.96official
6gemini-2.5-flash4.455official
7gemini-2.5-pro4.95official
8glm-4.57.921official
9grok-3-mini8.911official
10o110.89official
11gpt-4.5-preview11.88official
12claude-3.5-sonnet12.87official
12qwen3-30b-a3b12.87official
14llama-3.1-405b-instruct14.36official
15deepseek-r115.1official
16gemini-2.0-pro-exp15.84official
17claude-3.7-sonnet16.58official
18gemini-1.5-pro16.83official
19grok-317.82official
19llama-3.3-70b-instruct17.82official
21o1-preview18.32official
22qwen3-235b-a22b18.56official
23gemini-2.0-flash19.55official
24gpt-oss-120b23.27official
25o3-pro23.38official
26gpt-4o24.26official
27o324.75official
28ernie-4.5-300b-a47b25.25official
28qwq-32b25.25official
30grok-225.74official
31o1-mini26.24official
32gpt-4-turbo26.73official
32o4-mini26.73official
34claude-3-opus28.22official
34llama-4-maverick28.22official
36o3-mini28.96official
37kimi-k230.2official
38qwen2.5-max31.19official
39mistral-large-232.18official
39qwen2.5-72b-instruct32.18official
41chatgpt-4o38.12official
41mistral-medium-338.12official
43mistral-small-338.61official
44deepseek-v339.11official
45minimax-text-0144.55official
46gemma-2-27b-it47.03official
47phi-452.48official
48nova-pro54.46official
49claude-3-haiku56.93official
50gpt-4o-mini60.89official
51claude-3.5-haiku65.84official
52gemma-3-27b-it66.34official