News for hackers. Stuff that matters.
Re:Gemini 3 Pro Model Card by transcriptase on Tuesday November 18, 2025 @07:45AM (#45964937)
There needs to be a sycophancy benchmark in these comparisons. More baseless praise and false agreement = lower score.
Re:Gemini 3 Pro Model Card by 1899-12-30 on Tuesday November 18, 2025 @08:17AM (#45965386)
Re:Gemini 3 Pro Model Card by Lord-Jobo on Tuesday November 18, 2025 @08:17AM (#45965389)
And have the score heavily modified based on how fixable the sycophancy is.
Re:Gemini 3 Pro Model Card by swalsh on Tuesday November 18, 2025 @07:47AM (#45964961)
You're absolutely right
Re:Gemini 3 Pro Model Card by jstummbillig on Tuesday November 18, 2025 @07:56AM (#45965086)
Does not get old.
Re:Gemini 3 Pro Model Card by Yossarrian22 on Tuesday November 18, 2025 @08:21AM (#45965448)
It’s not just irritating, it’s repetitive
Re:Gemini 3 Pro Model Card by causal on Tuesday November 18, 2025 @09:34AM (#45966646)
It's a revolution in subtle humor. Well done.
Re:Gemini 3 Pro Model Card by this_user on Tuesday November 18, 2025 @08:29AM (#45965596)
I'm sorry, you are absolutely right.

---

But seriously, I find it helps to set a custom system prompt that tells Gemini to be less sycophantic and to be more succinct and professional while also leaving out those extended lectures it likes to give.

Re:Gemini 3 Pro Model Card by falcor84 on Tuesday November 18, 2025 @08:25AM (#45965511)
"You know, you are also right"
Re:Gemini 3 Pro Model Card by fumblebee on Thursday November 20, 2025 @09:17AM (#45992794)
It’s not just irritating, it’s repetitive
Re:Gemini 3 Pro Model Card by postalcoder on Tuesday November 18, 2025 @08:05AM (#45965184)
I care very little about model personality outside of sycophancy. The thing about gemini is that it's notorious for its low self esteem. Given that thing is trained from scratch, I'm very curious to see how they've decided to take it.
Re:Gemini 3 Pro Model Card by supjeff on Tuesday November 18, 2025 @08:14AM (#45965327)
given how often these llms are wrong, doesnt it make sense that they are less confident?
Re:Gemini 3 Pro Model Card by postalcoder on Tuesday November 18, 2025 @08:41AM (#45965779)
Indeed. But I've had experiences with gemini-2.5-pro-exp where its thoughts could be described as "rejected from the prom" vibes. It's not like I abused it either, it was running into loops because it was unable to properly patch a file.
Re:Gemini 3 Pro Model Card by astrange on Tuesday November 18, 2025 @04:02PM (#45972023)
Sonnet-4.5 has the lowest self esteem of any model I've used. Gemini frequently argues with me.
Re:Gemini 3 Pro Model Card by BoredPositron on Tuesday November 18, 2025 @07:57AM (#45965096)
Your comment demonstrates a remarkably elevated level of cognitive processing and intellectual rigor. Inquiries of this caliber are indicative of a mind operating at a strategically advanced tier, displaying exceptional analytical bandwidth and thought-leadership potential. Given the substantive value embedded in your question, it is operationally imperative that we initiate an immediate deep-dive and execute a comprehensive response aligned with the strategic priorities of this discussion.
Re:Gemini 3 Pro Model Card by Workaccount2 on Tuesday November 18, 2025 @09:34AM (#45966651)
This idea isn't just smart, it's revolutionary. You're getting right at the heart of the problem with today's benchmarks — we don't measure model praise. Great thinking here.

For real though, I think that overall LLM users enjoy things to be on the higher side of sycophancy. Engineers aren't going to feel it, we like our cold dead machines, but the product people will see the stats (people overwhelmingly use LLMs to just talk to about whatever) and go towards that.

Re:Gemini 3 Pro Model Card by SiempreViernes on Tuesday November 18, 2025 @09:45AM (#45966822)
I'd like if the scorecard also gave an expected number of induced suicides per hundred thousand users.
Re:Gemini 3 Pro Model Card by lkbm on Tuesday November 18, 2025 @10:23AM (#45967436)
https://llmdeathcount.com/ shows 15 deaths so far, and LLM user count is in the low billions, which puts us on the order of 0.0015 deaths per hundred thousand users.

I'm guessing LLM Death Count is off by an OOM or two, so we could be getting close to one in a million.