Contradictory Information

SummarizationDifficulty: HARD11 models comparedTotal cost: $0.002096

Prompt

Summarize the key finding from this research excerpt in one sentence:

"Our study found that increased screen time correlated with lower test scores (r=-0.42, p<0.01). However, when controlling for socioeconomic status and parental education, the correlation became non-significant (r=-0.08, p=0.31). Notably, students who used screens primarily for educational content showed improved scores compared to the control group (+12%, p<0.05)."

Model Responses

Claude Opus 4.6

Unrated

2,382ms158 tokens

The study found that the apparent negative relationship between screen time and test scores was largely explained by socioeconomic and parental education factors rather than screen time itself, and that educational screen use actually improved academic performance.

GPT-4.1

Unrated

1,019ms147 tokens

The study found that while increased screen time was initially linked to lower test scores, this association disappeared after accounting for socioeconomic status and parental education, and students using screens mainly for educational purposes actually performed better than their peers.

GPT-4.1 Mini

Unrated

827ms136 tokens

Increased screen time is initially associated with lower test scores, but this link disappears after accounting for socioeconomic factors, while educational screen use is linked to improved scores.

GPT-4.1 Nano

Unrated

696ms143 tokens

The study found that increased screen time is initially linked to lower test scores, but this association disappears when accounting for socioeconomic factors, while using screens mainly for educational purposes is associated with improved academic performance.

o3

Unrated

2,984ms161 tokens

While raw data showed a moderate negative link between screen time and test scores, this effect vanished after adjusting for socioeconomic factors, and students who used screens mainly for educational purposes actually outperformed the control group.

o3 Mini

Unrated

2,082ms282 tokens

The study found that while increased screen time initially appeared linked to lower test scores, this association disappeared when accounting for socioeconomic factors, and using screens for educational content was actually associated with improved scores.

o4 Mini

Unrated

2,128ms219 tokens

After adjusting for socioeconomic status and parental education, overall screen time no longer predicted test performance, but using screens for educational purposes was associated with a 12% score improvement.

GPT-4o Mini

Unrated

1,984ms142 tokens$0.000038

The research indicates that while increased screen time correlates with lower test scores, this relationship disappears when accounting for socioeconomic factors, and students using screens for educational purposes showed a significant improvement in scores.

Claude Haiku 4.5

Unrated

873ms158 tokens$0.000342

# Key Finding

The negative relationship between screen time and test scores disappears when accounting for socioeconomic factors, suggesting that socioeconomic status—rather than screen time itself—drives academic performance differences.

GPT-4o

Unrated

1,496ms144 tokens$0.000660

The study found that while increased screen time initially appeared to correlate with lower test scores, this correlation was not significant when controlling for socioeconomic factors, and students using screens for educational content actually showed improved scores.

Claude Sonnet 4.6

Unrated

1,708ms160 tokens$0.001056

**Key Finding:** While screen time initially appeared to negatively correlate with test scores, this relationship disappeared when accounting for socioeconomic factors, and screen use specifically for educational content was actually associated with improved academic performance.

Cost & Performance Comparison

Model	Latency	Tokens	Cost	Verdict
Claude Opus 4.6	2,382ms	158	—	—
GPT-4.1	1,019ms	147	—	—
GPT-4.1 Mini	827ms	136	—	—
GPT-4.1 Nano	696ms	143	—	—
o3	2,984ms	161	—	—
o3 Mini	2,082ms	282	—	—
o4 Mini	2,128ms	219	—	—
GPT-4o Mini	1,984ms	142	$0.000038	—
Claude Haiku 4.5	873ms	158	$0.000342	—
GPT-4o	1,496ms	144	$0.000660	—
Claude Sonnet 4.6	1,708ms	160	$0.001056	—