informal007

it will be more excited if i can use this feature in application with GUI, it’s now convenient to check the result after edit the PDF, i need to transfer between CLI and PDF reader

Speaking of filtering trustworthy URLs, Google is the best option to do that because he has more historical data in search business.

Hope google can do something for preventing prompt injection for AI community.

It seem that Google doesn't prepare well to release Gemini 3 but leak many contents, include the model card early today and gemini 3 on aistudio.google.com

bad customer service comes from low priority. I think anthropic prioritize new growth point over small number of customer’s feedback, that’s why they publish new product, features so frequently, there are so much possible potential opportunities for them to focus

Database might be the future, but viable solution on excel are evidence to prove that it works

agree with you, but it cannot be stopped. development of technology always makes wealth distribution more centralized

this will push the development of open source models.

people think of privacy at first regards of data, local deployment of open source models are the first choice for them

you might trust when the precision is extremely high and others agree with that.

high precision is possible because they can realize that by multiple cross validations

The wars between LLMs are still hot, it's not expensive to get enough access for LLMs, like student version of Gemini.

More access for free users might not provide enough attraction.

Future will be more challenging for fraud-detection fields, good luck for them.

Computer use model comes from interactive demand with computer automatically, Chrome devtools MCP might be one of the core pushers.

Everything is possible to be integrated to AI, AI expectation is so huge, even Google is try to implement AI into search engine. it's not much surprise that Github become the rock on the road for MS to AI.

If Github/Copilot wins the war of coding assistant and becomes the next growth point in MS, the story will be total different.

We shouldn't ignore the influence of trend, it's like the facebook in mobile era.

LLM is helpful for creativity and thinking When you run out of your ideas

They have all hardware and most advanced LLM technology, I don't know why their market value didn't catch up with microsoft.

Audio maybe the better option.

On the one hand, it's helpful to join a new friendship club if we truely admit that most of friendship is going to disappear as we are getting old.

On the other hand, some people are shamed to admit this point, as a result, joining this kind of club will make them look more shamed. It's just nature of humanity.

I think there should be a better idea for the second type of people.

Why videos are important for robotics?

Why it emphasize math and computer science more in training stage?

The problems in math and CS are more suitable for training LLMs?

How to get "unconstrained output" and evaluate the distance between them?

Evaluation method which can decide distance between two sentences is hard to find, best option is closed-source LLM API even it's not the most ideal option. As a result, we also must use current LLM to improve our models.

Yes, RL works well in fields where answer can be verified in different degree. That's why AlphaGo success, it also should work in code generation and math.

Would it better if we move the feedback loops into RL-stage of LLM training?

Are there some related works?

This remind me that road router should be walked by passenger rather than designed by designers.

Finally finished my ideal bracket and quote code in vimrc

In the past, I cannot find a great config for this function, now, I finally finished it by myself with help of Gemini, share to people who need it.

Some students become better because of LLMs, some become worse.

It's like some people learn knowledge by TikTok, some just waste time on it.

It's like painting made by human, not photo captured by camera.

> Outputs are evaluated with a scoring rubric by Claude Sonnet 3.7.

Will Claude Sonnet 3.7 judger favor himself?

How much GPU memory are required for inference if it's 10M context?

It's still BF16 model.

Deepseek has proved that fp8 is more cost-effectiveness than fp16, isn't it valid for dozens-B model?

Some machine learning models are explainable, like tree model, xgboost.

Neural network model are hard to explain, especially LLM.