Nope, human onboard.
blakkd
owao
AI & ML interests
None yet
Recent Activity
View all activity
Organizations
None yet
owao's activity
@anthony
@mfuntowicz
@aliabd
Can we do something guys? Sorry to ping be it is becoming more and more obvious. All sort of patterns already show up.
The craziest is I'm starting there is even no human onboard! Let's try to figure this out.
@powergen4ai , you are right. Let's wipe the slate clean on all this. Let's try something instead: let's focus on the popcorn part. I think it is actually a valuable criteria to determine the huggingface value of a developer. But of course, only humans know how to prepare it the right way. So, if you really are a human, this is your chance to prove it. I doubt you can provide any recipe...!
lol, but isn't this precisely a personal attack? You are funny, also, keep the drama going on, let's pump up the visibility!
A normal dev would just have dropped the discussion, but you keep going on. But that's too late, now you are committed.
Dude, just let the LLM entirely respond, you are deteriorating the quality of the reasoning adding your input to the prompt. Maybe just prompt it "seems we are in trouble now, what to do??" but let it figure out ;)
Preparing the popcorn
How many fake accounts do you have to manage? Just curious ;)
I see you behind your keyboard, using an LM as a proxy because you can't construct sentences by yourself. Sadly, your LM isn't smart enough to generate meaningful arguing. Come on try to find one by yourself, it should be at least better. We are having fun now!
They are trying to hide it, all members of the same organization Powergen-AI. Of course they have no argument apart trying to redirect the fault attacking on a legitimate anger here cause I didn't start this. You are now stuck in your own trap dumbass, please continue answering to promote this post and make it viral, shedding the light on your malicious enterprise.
I'll just wait to see this empire fall.
Pattern recognition to spot fake accounts is becoming a urging need.
LM Studio: unknown model architecture: 'glm4'?
3
#4 opened 5 days ago
by
DrNicefellow
Spotted this motherfker
@nielsr
can we do something here to protect HF and the community?
Can't be, along with all the bot instantly reacting here, we seem to face a bot army here building up fake HF profiles to later push harmful models/space.
I already reported this to HF, but me alone can't weight much...
Report: Nice job, you've ruined a nice model with censorship no one wants or asks for
5
16
#7 opened 2 days ago
by
Ainonake
Can someone explain why this model is currently at top 3 of the trending list?
8
5
#8 opened 2 days ago
by
RobinFeng123
I'll let people judge
liked
a
Space
1 day ago
liked
a
model
1 day ago
reacted to
hesamation's
post with π€
2 days ago
Post
2725
The best researchers from DeepSeek, OpenAI, Microsoft, and ByteDance explored RL and Reasoning in LLMs,
Here's some of their key findings:
1/ RL can further improve distilled models. These models are essentially SFT fine-tuned with the data generated by larger models, and the SFT+RL combo does not disappoint.
This is verified in the DeepSeek-R1 paper.
2/ both GRPO and PPO algorithms suffer from length bias; they encourage longer responses. This can be tackled by introducing explicit rewards based on the length of the answer.
3/Most reasoning research is focused on code and math. But training models on logic puzzles improves them for mathematical tasks too.
This shows the RL reasoning is generalized beyond the specific domain knowledge.
Previous research also shows RL can be a great generalizer.
4/The reasoning might not be only induced by RL; it might already be hidden in the base models due to the pre-training and CoT data they were trained on.
So while RL does wake up the reasoning beast, maybe it's not the only solution (e.g. other methods such as distillation)
5/ back to the length bias; reasoning models tend to generate longer responses for wrong answers. RL might be the culprit.
RL favours longer answers when the reward is negative, to dilute the penalty per individual token and lower the loss.
This might explain the "aha" moments!
6/ OpenAI's competitive programming paper showed an interesting finding:
o3 can learn its own test-time strategies (like writing an inefficient but correct solution to verify the answer of an optimized solution)
RL helps LLMs develop their own reasoning & verification methods.
The recent article by @rasbt helped me a lot in getting a broad view of the recent research on reasoning models.
He also lists more influential papers on this topic, It's a must-read if you're interested.
check it out π
https://magazine.sebastianraschka.com/p/the-state-of-llm-reasoning-model-training
Here's some of their key findings:
1/ RL can further improve distilled models. These models are essentially SFT fine-tuned with the data generated by larger models, and the SFT+RL combo does not disappoint.
This is verified in the DeepSeek-R1 paper.
2/ both GRPO and PPO algorithms suffer from length bias; they encourage longer responses. This can be tackled by introducing explicit rewards based on the length of the answer.
3/Most reasoning research is focused on code and math. But training models on logic puzzles improves them for mathematical tasks too.
This shows the RL reasoning is generalized beyond the specific domain knowledge.
Previous research also shows RL can be a great generalizer.
4/The reasoning might not be only induced by RL; it might already be hidden in the base models due to the pre-training and CoT data they were trained on.
So while RL does wake up the reasoning beast, maybe it's not the only solution (e.g. other methods such as distillation)
5/ back to the length bias; reasoning models tend to generate longer responses for wrong answers. RL might be the culprit.
RL favours longer answers when the reward is negative, to dilute the penalty per individual token and lower the loss.
This might explain the "aha" moments!
6/ OpenAI's competitive programming paper showed an interesting finding:
o3 can learn its own test-time strategies (like writing an inefficient but correct solution to verify the answer of an optimized solution)
RL helps LLMs develop their own reasoning & verification methods.
The recent article by @rasbt helped me a lot in getting a broad view of the recent research on reasoning models.
He also lists more influential papers on this topic, It's a must-read if you're interested.
check it out π
https://magazine.sebastianraschka.com/p/the-state-of-llm-reasoning-model-training
Come on now, he did change the readme!
OK, looking at the commit revealed this guy didn't even clone bartee's space, but openfree instead, which already was a duplicate. And the readme commit was from openfree: https://huggingface.co/spaces/aiqtech/LLM-Token-Visual/commit/ee604cb4386c8511d81f6b5896cb221a9ec8041c
They managed to:
- make the title a nonsense
- break the only sentence it was containing
lmao
Even more pathetic... It's like they are competing in dumbness and shamelessness
So:
- you duplicated https://huggingface.co/spaces/barttee/tokenizers
- you didn't even touched the source code
- you claim you made it
And, instead of being blamed for your actions, people are clapping?
This is pathetic.
δ½ θ―΄ζι«δΊοΌζθ ζθ―΄δ½ ι«δΊοΌ
#3 opened 4 days ago
by
owao