Introducing our new research paper in AI
Yes, the research is live and here's everything you need to know about it..
Hey fam 👋
Finally, after several months of dedicated work on our research at the Indian Institute of Technology, Patna, we are pleased to announce our research on ‘Real Time Retrieval Argumentation for Large Language Models’.
This newsletter will work as your hands-on guide to go through the paper and explore the paper by yourself.
What is the research paper about?
Real Time Retrieval Argumentation or RTRA is an alternate architecture to build and use LLMs where models trained with 1-10 billion parameters outperforms all the 100 billion+ parameter models including GPT3, GPT4, Claude Sonet, Gemini Pro 1.0, etc.
What are the major breakthroughs of the paper?
From a positive environmental impact to reducing hallucinations to net zero, there are multiple breakthroughs that we have introduced in this research paper.
Here are some major ones :
Completely eradicates hallucination from any model
Connect any transformer-based model with real-time web
Provide citations and credits to the original authors for their work
Reduce dependency on huge computation resources for building an efficient model.
Where is the research paper publication published?
The research paper is published as a preprint publication at ResearchGate.
What can you do with RTRA?
With RTRA, individuals & companies can build astonishing products :
For individuals : Take this as your real-time search engine. Whatever the question is, it gives answers that are highly accurate and backed by sources. You can even use it for multiple use cases.
For researchers & authors : Got a question while writing content or paper? Want answers that are backed by true citations and references and also are latest as per internet? RTRA is her to provide that.
For developers : Build products over a highly precise LLMs with zero hallucinations and real-time internet backed responses. Use RTRA models over API.
For enterprises : You tried building AI products but the irrelevant and old answers won’t let you truly harness the potential of AI. That’s eradicated now with RTRA. Build applications with dedicated or cloud deployment of models on RTRA.
What’s next?
You can get RTRA access by joining our product’s waitlist. If you want a self-hosted base version of RTRA (called Rtrafactor), you will get its access within 24 hours.
If you want RTRA API or search engine, you will need to join our waitlist and will typically be granted access within a week.
What made us work on this publication?
Sttabot AI, the core project that we started, was a wrapper tool (which means wrapping an AI API key within a UI template as a bot) that acted more as a no-code chatbot builder. However, we had two major challenges :
Too many ships with similar features & pricing : While starting, Sttabot was one of the first platform to introduce a wrapper. But by the end of 2023, there were thousands of apps featuring the same service and at somewhat of a similar pricing.
Not so breakthrough : At the start, Sttabot created a direct impact in people's business and life - making AI apps suddenly got easy. However, as time passed, maybe because I was a solo developer, I was not able to launch such breakthrough features for my users.
Leading to this, I again went back to zero and started my research along with my college (Indian Institute of Technology, Patna) under Dr. Kuldeep Singh Patel over a topic - Real Time Retrieval Argumentation for LLMs.
Who’s behind this paper?
The paper was not possible without a strong foundation team with 40+ contributors and multiple core members :
Udit Raj ‘Akhouri’ - Lead researcher, IIT-Patna
Sanya Gupta - Associate researcher, IIT-Patna
Dr. Kuldip Singh Patel - Consulting professor, IIT-Patna
Kaushik - Technical Contributor
Adarsh - Technical Contributor
Nandni Jaiswal - Non-Technical Contributor, IIT-patna
Luv Kumar - Non-Technical Contributor, IIT-patna
Astha Gaur - Non-Technical Contributor, IIT-patna
Satwik - Non-Technical Contributor, IIT-patna
Users in Eternity & RTRA waitlist
Members of the Sttabot & Supervised family
We are glad to introduce RTRA before you. If you have any feedback, do let us know at udit_2312res708@iitp.ac.in.
Regards
Udit Raj ‘Akhouri’
RTRA, Dept. of Computer Science
Indian Institute of Technology, Patna