Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Improved Gemini 2.5 Flash and Flash-Lite

developers.googleblog.com

Post reply on HN
Print view View original

251–260 of 285 posts

Re: Improved Gemini 2.5 Flash and Flash-Lite

#251

Post by JV00 »

JumpCrisscross wrote: post #219
oasisbob wrote: post #159

Earlier quoted context omitted.

While arguing may not be productive, I have had good results challenging Gemini on hallucinated sources in the past. eg, "You cited RFC 1918, which is a mistake. Can you try carefully to cite a better source here?" which would get it to re-evaluate, maybe by using another tool, admit the mistake, and allow the research to continue. With this example, several attempts resulted in the same thing: Gemini expressing a st…

One of the main problems with arguing with LLMs is your complaint becomes part of the prompt. Practically all LLMs have will take "don't do X" and do X, because part of "don't do X" is "do X," and LLMs have no fundamental understanding of negation.

Not really true these days. Claude code follows my instructions correctly when I tell it not to use certain patterns.

Re: Improved Gemini 2.5 Flash and Flash-Lite

#252

Post by frumiousirc »

lelanthran wrote: post #237
cerved wrote: post #226

Earlier quoted context omitted.

Yesterday I asked Gemini to recalculate the timestamps of tasks in a sequence of tasks, given it's duration and the previous timestamp. It proceeded to write code which gave results like this 2025-09-26T14:32:10Z 2025-09-26T14:32:10Z200s 2025-09-26T14:32:10Z200s600s 2025-09-26T14:32:10Z200s600s300s It then proceeded to talk about how efficient this approach was for thousands of numbers. Gemini is by far the dumbest L…

They're all a little dumb. I asked claude for a python function or functions that will take in markdown in a string and return a string with ansi codes for bold, italics and underline. It gave me a 160 line parse function. After gaping for a short while, I implemented it in a 5 line function and a lookup table. These vibe codes who are proud that they generated thousands of lines of code makes me wonder if they are e…

I just asked Gemini Flash to do this. I included the instruction to use regular expressions to do the conversion to ANSI. It gave me a reasonable Python function which boils down to calling `re.sub()` for each of bold, italic and underline. For italics:

    text = re.sub(r'(\*|_)(.+?)\1', replace_italic, text, flags=re.DOTALL)
The `replace_italic` is a one line callback function surrounding the re's match with the ANSI codes.

Knowing what technique is "best" and telling the LLM to use it produces better results (on average) than giving the LLM freedom to choose. For some problems, the specification of the prompt needed to get good output becomes more work than just thinking and writing for myself.

For very complex things, I myself can not put the design into English in my own head but can "see" the correct answer as code concepts. I don't know if this is universal for all developers. If it is, it shows a limit of LLM's usefulness.

Re: Improved Gemini 2.5 Flash and Flash-Lite

#253

Post by ACCount37 »

JumpCrisscross wrote: post #219
oasisbob wrote: post #159

Earlier quoted context omitted.

While arguing may not be productive, I have had good results challenging Gemini on hallucinated sources in the past. eg, "You cited RFC 1918, which is a mistake. Can you try carefully to cite a better source here?" which would get it to re-evaluate, maybe by using another tool, admit the mistake, and allow the research to continue. With this example, several attempts resulted in the same thing: Gemini expressing a st…

One of the main problems with arguing with LLMs is your complaint becomes part of the prompt. Practically all LLMs have will take "don't do X" and do X, because part of "don't do X" is "do X," and LLMs have no fundamental understanding of negation.

That depends entirely on how well trained a given LLM is.

Gemini is notoriously bad at multi-turn instruction following, so this holds strongly for it. Less so for Claude Opus 4 or GPT-5.

Re: Improved Gemini 2.5 Flash and Flash-Lite

#254

Post by lelanthran »

frumiousirc wrote: post #252
lelanthran wrote: post #237

Earlier quoted context omitted.

They're all a little dumb. I asked claude for a python function or functions that will take in markdown in a string and return a string with ansi codes for bold, italics and underline. It gave me a 160 line parse function. After gaping for a short while, I implemented it in a 5 line function and a lookup table. These vibe codes who are proud that they generated thousands of lines of code makes me wonder if they are e…

I just asked Gemini Flash to do this. I included the instruction to use regular expressions to do the conversion to ANSI. It gave me a reasonable Python function which boils down to calling `re.sub()` for each of bold, italic and underline. For italics: text = re.sub(r'(\*|_)(.+?)\1', replace_italic, text, flags=re.DOTALL) The `replace_italic` is a one line callback function surrounding the re's match with the ANSI c…

> I included the instruction to use regular expressions to do the conversion to ANSI.

The viber coders (who I referred to in my comment) aren't giving implementation tips.

What did it give you before you put an implementation tip into your prompt?

=======

FWIW, if you're at all interested, here's my implementation:

    def markdown_ansi_code_subst(mdstr: str, src_pattern: str, replacement_start: str, replacement_end: str) -> str:
        while src_pattern in mdstr:
            mdstr = mdstr.replace(src_pattern, replacement_start, 1)
            mdstr = mdstr.replace(src_pattern, replacement_end, 1)
        return mdstr
The caller supplies the pattern (`*` for italic, `**` for bold, etc) and a start/end replacement. As you can imagine, I store all of that in a static lookup table.

I feel this is more readable than regexes.*

Re: Improved Gemini 2.5 Flash and Flash-Lite

#255

Post by dgacmu »

manquer wrote: post #204
JumpCrisscross wrote: post #194

Earlier quoted context omitted.

Would that automatically roll over anyone pinging 2.5 via their API?

If you want role over then you could specify ^2.5.0 or 2.5.x if you want to pin then it would be 2.5.0 This is all solved for a long time now , llm vendors seems to have unlearnt versioning principles. This is fairly typical - marketing and business wants different things to do with version number than what version number systems are good at .

I suspect Google doesn't want to have to maintain multiple sub-versions. It's easier to serve one 2x popular model than two models where there's flux between the load on each, since these things have a non-trivial time to load into GPU/TPU memory for serving.

Re: Improved Gemini 2.5 Flash and Flash-Lite

#256

Post by Thorrez »

alwillis wrote: post #170
herpderperator wrote: post #164

Serious question: If it's an improved 2.5 model, why don't they call it version 2.6? Seems annoying to have to remember if you're using the old 2.5 or the new 2.5. Kind of like when Apple released the third-gen iPad many years ago and simply called it the "new iPad" without a number.

It's pretty common to refer to models by the month and year they were released. For example, the latest Gemini 2.5 Flash is known as "google/gemini-2.5-flash-preview-09-2025" [1]. [1]: https://openrouter.ai/google/gemini-2.5-flash-preview-09-202...

>For example, the latest Gemini 2.5 Flash is known as "google/gemini-2.5-flash-preview-09-2025" [1].

That "example" is the name used in the article under discussion. There's no need to link to openrouter.ai to find the name.

Re: Improved Gemini 2.5 Flash and Flash-Lite

#257

Post by rldjbpin »

having developed a large-batch workflow for a client using gemini models, this is a welcome improvement. however, no news on the DSQ [1] issues is a bummer.

at least for us, the bottleneck is the amount of retries/waiting needed to max out how many requests we can make in parallel.

[1] https://cloud.google.com/vertex-ai/generative-ai/docs/dynami...

Re: Improved Gemini 2.5 Flash and Flash-Lite

#258

Post by whinvik »

Having done some tests, its clearly better at instruction following and JSON output now.

However its hampered by max output tokens. Gemini is at 65 K while GPT 5 mini is at 128K. Both of them have similar costs as well so as such apart from the 1M context limit GPT 5 mini is better in every way.

Re: Improved Gemini 2.5 Flash and Flash-Lite

#259

Post by Spooky23 »

driese wrote: post #141
davidmckayv wrote: post #53

This really captures something I've been experiencing with Gemini lately. The models are genuinely capable when they work properly, but there's this persistent truncation issue that makes them unreliable in practice. I've been running into it consistently, responses that just stop mid-sentence, not because of token limits or content filters, but what appears to be a bug in how the model signals completion. It's been…

Small things like this or the fact that AI studio still has issues with simple scrolling confuse me. How does such a brilliant tool still lack such basic things?

Because they are moving fast and breaking shit.

Ask ChatGPT to output markdown or PDF on iOS or Mac app and the web experience. The web is often better - the apps will return nothing.

Re: Improved Gemini 2.5 Flash and Flash-Lite

#260

Post by dimal »

simonw wrote: post #178
zamalek wrote: post #177

Earlier quoted context omitted.

I wonder if [good examples of] SVGs of pelicans on bikes are "being introduced" into training sets. Some of the engineers who work on this stuff are the kind to hang out here.

It's possible, but honestly I've never seen a decent vector illustration of a pelican on a bicycle myself so they'd have to work pretty hard to find one!

They could just ask a designer to do a few bespoke illustrations, then generate synthetic data from that, right? Have an image model generate a set of variations, then convert them to SVG.

But looking at these images, Google clearly hasn’t done that yet.

Post reply on HN