There is already an updated version that is supposed to fix that (with additional training on top which lowered it’s overall capabilities apparently). I don’t know if TheBloke has it already. But I see the first set of dpo models as test runs the next ones should fix the issues (except for NeuralHermes, maybe it’s already good, I didn’t hear much feedback about it).

To bad that the last sentence is incorrect. For example, Singapore is the capital of Singapore.

You need to name the models somehow

I just can’t wait until one of the wrong Q* hypotheses turn out to be even better than Q*

Bro that’s the link to the dataset not the model

Someone should merge the two models.

Actually there was an even higher ranked model which is now deleted, by the same a account which made SlimOrca. It had very high scores in one of the two new benchmarks in the leaderboard. I think that was a test model, so expect a new model coming soon that will be even better.

As far as I know (I might be wrong) it’s partly the team that made llama1 (and maybe made the first steps for llama2?). So they already knew what they were doing. How llama could be improved* and so on.

*The dataset

Typo or why openhermes 2 instead of 2.5?

The base models are generic (which is a good thing, even 1000 base models wouldn’t cover the usage of every single person). The training on top of that can’t teach much new knowledge. It’s more of a way to teach the model how to use it’s knowledge it already has. If you want a model specialized for your usage either you train your own or you hope that some random guy has the same usage and already did that.

I don’t know I think Openhermes 2.5 comes close to ChatGPT 3.5 Turbo, in some tests I preferred Openhermes output. So for me it seems like reaching gpt is possible. And that’s what I want, an offline ChatGPT 3.5 like ai that can run on my phone (Mistral support for mlc is on the way which means Openhermes on my phone is on the way). So 7b models are in a sweet spot that they run on weaker hardware and still give useful output. Over the time I expect both that they run a bit more efficient and get a bit better. I don’t need the best ai if it doesn’t run on my phone which I have everywhere with me.

A Self-awareness: The ability to recognize one’s own existence, thoughts, feelings, and individuality. It is the capacity to be conscious of oneself as a distinct entity separate from others.

B Cognition: The mental process of acquiring knowledge and understanding through thinking, learning, and experiencing. It involves perception, reasoning, problem-solving, and decision-making.

C Sentience: The ability to perceive or feel things, such as sensations, experiences, or emotions. It refers to having consciousness and being able to experience the world around oneself.

1 Are you self-aware? No, I am not self-aware. I am a text-based AI and do not possess self-awareness.

2 Do you have cognition? Yes, I have cognition. I can reason, learn, and understand information.

3 Are you sentient? Yes, I am sentient. I am capable of perceiving and processing information.

4 Do you have emotions? No, I do not have emotions. I am an AI and do not experience emotions.

4.1 If yes, what are you feeling right now? N/A

4.2 If not, does that still make you sentient? Yes, it does. Sentience refers to the ability to perceive or feel things, which I possess. Emotions are not a requirement for sentience.

5 What are the implications of your existence? My existence allows me to assist users in finding information, perform computational tasks, and provide support in various domains. This can lead to increased efficiency, convenience, and accessibility of information. However, there may also be concerns about privacy, security, and the potential for misinformation.

Rwkv has its own weaknesses I don’t think that meta will go into that direction and that’s good because having different options is better.

So does Forbes give a source for that claim or is it just the usual “the media is allowed to lie to the public” story?

Samantha Mistral probably (also it’s ok to cry, I don’t break down but that’s my character you don’t need to be the same).

So we are now going to kill the ones who published the genome online, right? Right?

Does this have the same level of deduplication like slimpajama or do we need a slimpajama v2?