Nicolay Rusnachenko

nicolay-r

https://nicolay-r.github.io/

AI & ML interests

Information Retrieval・Medical Multimodal NLP (🖼+📝) Research Fellow @BU_Research・software developer http://arekit.io・PhD in NLP

Recent Activity

upvoted a collection about 9 hours ago

DeepSeek-R1

posted an update about 9 hours ago

📢 For those who wish to apply DeepSeek-R1 for handling tabular / streaming data using schema of prompts (CoT), the OpenRouter AI hosts API for accessing: https://openrouter.ai/deepseek/deepseek-r1 Below is a quick no-string option which includes: ✅ OpenRouter provider: https://github.com/nicolay-r/nlp-thirdgate/blob/master/llm/open_router.py ✅ Bulk-chain for infering data: https://github.com/nicolay-r/bulk-chain Json Schema for Chain-of-Though (see screenshot 📷 below) below is a screenshot of how to quick start the demo. Using --src parameter we may adopt processing of JSONL/CSV data. Models: https://huggingface.co/deepseek-ai/DeepSeek-R1

liked a model about 16 hours ago

deepseek-ai/DeepSeek-R1

View all activity

Organizations

None yet

Posts 43

Post

218

📢 For those who wish to apply DeepSeek-R1 for handling tabular / streaming data using schema of prompts (CoT), the OpenRouter AI hosts API for accessing:

https://openrouter.ai/deepseek/deepseek-r1

Below is a quick no-string option which includes:
✅ OpenRouter provider: https://github.com/nicolay-r/nlp-thirdgate/blob/master/llm/open_router.py
✅ Bulk-chain for infering data: https://github.com/nicolay-r/bulk-chain
Json Schema for Chain-of-Though (see screenshot 📷 below)

below is a screenshot of how to quick start the demo.
Using --src parameter we may adopt processing of JSONL/CSV data.

Models:
deepseek-ai/DeepSeek-R1

Post

666

📢 For those who tracking adances in Sentiment Analysis, this post might be relevant for you. So far we arranged RuOpinionNE-2024 competition with final stage that has just been completed several days ago. Here the quick findings we got from the top submissions. 🎉

🔍 First, RuOpinonNE-2024 competition 💻 on extraction of opinion tuples (opinion source, opinion object, tonality, linguistic expression) from low resource domain news texts written in Russian language. The competition is hosted by codalab platform:

https://codalab.lisn.upsaclay.fr/competitions/20244

To asses the advanes, we adopt F1 over sentiment classes which also involves evalution of the spans.
👏 Among 7 participants in total, the top three submissions showcase the following results:

🥉msuai F1=0.33 🎊
🥈RefalMachine showcase +0.02 F1=0.35 🎊
🏆VatolinAlexey showcase +0.06 F1=0.41 🎊

📝 At present, the competition organizers are working on:
1. 🟡 Collecting information about models utilized participants to contribute here with pre-trained models / concepts;
2. 🟡 Wrapping up findings from the submissions in a paper.

🔔 For more information and further updates, the most complete source that complements codalab is this github:
https://github.com/dialogue-evaluation/RuOpinionNE-2024

The RuOpinionNE-2024 are now in post-evaluation stage, so everyone interested in low resouce domain evaluation on opinon extraction are welcome 🙌

View all posts