M Veselovskiy's picture

M Veselovskiy

Yuuru

·

AI & ML interests

None yet

Recent Activity

reacted to m-ric's post with 👀 about 1 month ago

𝗔𝗱𝘆𝗲𝗻'𝘀 𝗻𝗲𝘄 𝗗𝗮𝘁𝗮 𝗔𝗴𝗲𝗻𝘁𝘀 𝗕𝗲𝗻𝗰𝗵𝗺𝗮𝗿𝗸 𝘀𝗵𝗼𝘄𝘀 𝘁𝗵𝗮𝘁 𝗗𝗲𝗲𝗽𝗦𝗲𝗲𝗸-𝗥𝟭 𝘀𝘁𝗿𝘂𝗴𝗴𝗹𝗲𝘀 𝗼𝗻 𝗱𝗮𝘁𝗮 𝘀𝗰𝗶𝗲𝗻𝗰𝗲 𝘁𝗮𝘀𝗸𝘀! ❌ ➡️ How well do reasoning models perform on agentic tasks? Until now, all indicators seemed to show that they worked really well. On our recent reproduction of Deep Search, OpenAI's o1 was by far the best model to power an agentic system. So when our partner Adyen built a huge benchmark of 450 data science tasks, and built data agents with smolagents to test different models, I expected reasoning models like o1 or DeepSeek-R1 to destroy the tasks at hand. 👎 But they really missed the mark. DeepSeek-R1 only got 1 or 2 out of 10 questions correct. Similarly, o1 was only at ~13% correct answers. 🧐 These results really surprised us. We thoroughly checked them, we even thought our APIs for DeepSeek were broken and colleagues Leandro Anton helped me start custom instances of R1 on our own H100s to make sure it worked well. But there seemed to be no mistake. Reasoning LLMs actually did not seem that smart. Often, these models made basic mistakes, like forgetting the content of a folder that they had just explored, misspelling file names, or hallucinating data. Even though they do great at exploring webpages through several steps, the same level of multi-step planning seemed much harder to achieve when reasoning over files and data. It seems like there's still lots of work to do in the Agents x Data space. Congrats to Adyen for this great benchmark, looking forward to see people proposing better agents! 🚀 Read more in the blog post 👉 https://huggingface.co/blog/dabstep

new activity 5 months ago

TheDrummer/UnslopSmall-22B-v1-GGUF:Metharme format makes model extremely stupid

upvoted a collection 6 months ago

View all activity

Organizations

Yuuru's activity

New activity in TheDrummer/UnslopSmall-22B-v1-GGUF 5 months ago

Metharme format makes model extremely stupid

#1 opened 5 months ago by

New activity in G-reen/gpt5o-reflexion-q-agi-llama-3.1-8b 6 months ago

How to pay

#17 opened 6 months ago by

New activity in mattshumer/Reflection-Llama-3.1-70B 6 months ago

DLETE THIS MODEL

#76 opened 6 months ago by

New activity in yodayo-ai/kivotos-xl-2.0 9 months ago

Broken results

#1 opened 9 months ago by

New activity in saltlux/luxia-21.4b-alignment-v1.0 about 1 year ago

Quantized GGUF available

#3 opened about 1 year ago by

New activity in TheBloke/dolphin-2.6-mistral-7B-GPTQ about 1 year ago

I am having issue loading dolphin 2.6 mistral 7B GPTQ:main by TheBloke . Pasting the error in the description pls help

#1 opened about 1 year ago by

New activity in chargoddard/mixtralnt-4x7b-test about 1 year ago

It works!!!

#1 opened over 1 year ago by

New activity in TheBloke/Mixtral-8x7B-v0.1-GGUF over 1 year ago

It works.

#3 opened over 1 year ago by

New activity in mistralai/Mistral-7B-Instruct-v0.2 over 1 year ago

How is this different from v1?

#2 opened over 1 year ago by

New activity in TheBlokeAI/Mixtral-tiny-GPTQ over 1 year ago

What is this model?

#1 opened over 1 year ago by

New activity in stabilityai/stable-video-diffusion-img2vid-xt over 1 year ago

Did anyone figured it out how to run it in low vRAM like 15-25 GB

#21 opened over 1 year ago by

New activity in TheBloke/Yi-34B-GPTQ over 1 year ago

How do i run it?

#2 opened over 1 year ago by

New activity in TheBloke/WizardLM-30B-Uncensored-GGML almost 2 years ago

WizardLM-30B-Uncensored - reasoning is on the level of 65B models !

#2 opened almost 2 years ago by

New activity in TheBloke/airoboros-13B-1.1-GGML almost 2 years ago

Good

#1 opened almost 2 years ago by

New activity in TheBloke/Nous-Hermes-13B-GGML almost 2 years ago

Useless

#3 opened almost 2 years ago by

New activity in CalderaAI/30B-Lazarus almost 2 years ago

Observations

#3 opened almost 2 years ago by

New activity in TheBloke/falcon-40b-instruct-GPTQ almost 2 years ago

GGML?

#2 opened almost 2 years ago by

New activity in TheBloke/WizardLM-Uncensored-Falcon-7B-GPTQ almost 2 years ago

Found a script to convert to GGML

#2 opened almost 2 years ago by

New activity in TheBloke/Wizard-Vicuna-13B-Uncensored-GGML almost 2 years ago

O_O

#4 opened almost 2 years ago by

New activity in Aitrepreneur/stable-vicuna-13B-GPTQ-4bit-128g almost 2 years ago

Completely hallucinating

#1 opened almost 2 years ago by