Source A
OpenAI представила модель GPT-5.4 в версиях Pro и Thinking
overclockers.ru
https://overclockers.ru/blog/Scorpion81/show/251059/OpenAI-predstavila-model-GPT-5-4-v-versiyah-Pro-i-Thinking
Comparison
Winner: Source A is less manipulative
Source A appears less manipulative than Source B for this narrative.
Source B
I tested GPT-5.4 Thinking, and it gave me great answers (until I dove deeper)
zdnet.com
https://www.zdnet.com/article/gpt-5-4-thinking-tests-review/
Topics
Instant verdict
Less biased source: Source A
More emotional framing: Source B
More one-sided framing: Source B
Weaker evidence quality: Source B
More manipulative overall: Source B
Narrative conflict
Source A main narrative
Общая вероятность ошибок в ответах уменьшена на 14%.
Source B main narrative
That's what I meant when I said the model just didn't listen.
Conflict summary
Stance contrast: Общая вероятность ошибок в ответах уменьшена на 14%. Alternative framing: That's what I meant when I said the model just didn't listen.
Source A stance
Общая вероятность ошибок в ответах уменьшена на 14%.
Stance confidence: 69%
Source B stance
That's what I meant when I said the model just didn't listen.
Stance confidence: 88%
Central stance contrast
Stance contrast: Общая вероятность ошибок в ответах уменьшена на 14%. Alternative framing: That's what I meant when I said the model just didn't listen.
Why this pair fits comparison
- Candidate type: Closest similar
- Comparison quality: 53%
- Event overlap score: 26%
- Contrast score: 77%
- Contrast strength: Strong comparison
- Stance contrast strength: High
- Event overlap: Topical overlap is moderate. Issue framing and action profile overlap.
- Contrast signal: Stance contrast: Общая вероятность ошибок в ответах уменьшена на 14%. Alternative framing: That's what I meant when I said the model just didn't listen.
Key claims and evidence
Key claims in source A
- Общая вероятность ошибок в ответах уменьшена на 14%.
- Это означает, решение задач будет расходовать меньше прежнего токенов.
- Выросла эффективность расхода токенов и уменьшено количество ошибок в рассуждениях Компания OpenAI в четверг выпустила на рынок базовую модель GPT-5.4, которую она описывает как наиболее быструю и эффективную.
- Кроме стандартной, эта модель доступна в виде версии высокой производительности GPT-5.4 Pro и модели для рассуждений GPT-5.4 Thinking.
Key claims in source B
- That's what I meant when I said the model just didn't listen.
- I said, "Draw me a picture of the most probable design based on your existing analysis." And, wouldn't you know it?
- That said, there's no doubt the model can help professionals get their work done, as long as they are very diligent in monitoring results.
- I specified the characteristics of the craft, and then added on these instructions: "Design such a vehicle, particularly explaining its structure and how it will be held aloft, along with any constraints or issues, as w…
Text evidence
Evidence from source A
-
key claim
Общая вероятность ошибок в ответах уменьшена на 14%.
A key claim that anchors the narrative framing.
-
key claim
Это означает, решение задач будет расходовать меньше прежнего токенов.
A key claim that anchors the narrative framing.
-
selective emphasis
Раньше система выдавала определения всех доступных инструментов, когда происходил вызов модели.
Possible selective emphasis on specific aspects of the story.
-
omission candidate
That's what I meant when I said the model just didn't listen.
Possible context gap: Source A gives less coverage to economic and resource context than Source B.
Evidence from source B
-
key claim
That's what I meant when I said the model just didn't listen.
A key claim that anchors the narrative framing.
-
key claim
I said, "Draw me a picture of the most probable design based on your existing analysis." And, wouldn't you know it?
A key claim that anchors the narrative framing.
-
framing
I have long contended (and taught) that the only way you can learn programming is by actually writing code, which is a tangible example of educational constructivism in action.
Wording that sets an interpretation frame for the reader.
-
causal claim
I specified it fairly clearly, and gave the AI a variety of difficult-to-answer questions, difficult mostly because they're fundamentally unanswerable questions.
Cause-effect claim shaping how events are explained.
Bias/manipulation evidence
-
Source A · Framing effect
Раньше система выдавала определения всех доступных инструментов, когда происходил вызов модели.
Possible framing pattern: wording sets a specific interpretation frame rather than neutral description.
-
Source B · Framing effect
I have long contended (and taught) that the only way you can learn programming is by actually writing code, which is a tangible example of educational constructivism in action.
Possible framing pattern: wording sets a specific interpretation frame rather than neutral description.
-
Source B · False dilemma
Overall, the analysis and conclusions were great, although I was disappointed it didn't mention either the USS Akron or USS Macon, which were early 20th century aircraft-launching dirigible…
Possible false dilemma: the issue is presented as limited options while additional alternatives may exist.
How score signals are formed
Bias score signal
Bias signal combines framing pressure, emotional wording, selective emphasis, and one-sided narrative markers.
Emotionality signal
Emotionality rises when evidence contains emotionally loaded wording and evaluative labels.
One-sidedness signal
One-sidedness rises when one frame dominates and alternative interpretations are weakly represented.
Evidence strength signal
Evidence strength rises with concrete claims, attributed statements, and verifiable contextual support.
Source A
26%
emotionality: 27 · one-sidedness: 30
Detected in Source A
framing effect
Source B
45%
emotionality: 39 · one-sidedness: 40
Detected in Source B
framing effect
false dilemma
Metrics
Framing differences
- Source A emotionality: 27/100 vs Source B: 39/100
- Source A one-sidedness: 30/100 vs Source B: 40/100
- Stance contrast: Общая вероятность ошибок в ответах уменьшена на 14%. Alternative framing: That's what I meant when I said the model just didn't listen.
Possible omitted/downplayed context
- Source A pays less attention to economic and resource context than Source B.