Muneeswaran

ML Researcher @Quantiphi |
Interests: NLP | Graphs | LLMs | ML & DL | |
Opinions are all mine.

[Dharmesh Ba @dharmeshba]

Karpathy introduced vibe coding
Thalapathy introduced vibe politics

#TNElections

[Muneeswaran @munee_s12]

@karpathy @Ouponatime38 Our framework "Factual Evidence" does exactly this. Users can view the final response, with each sentence cited and accompanied by a rationale explanation. Each rationale cites one or more paragraphs from multiple documents.
Paper: https://t.co/XP3UkqZ7AR

https://t.co/lQ6DtuZrGj

[Muneeswaran @munee_s12]

By mitigating the risk of misinformation, we open new pathways for deploying LLMs in crucial applications, including but not limited to healthcare diagnosis, financial analysis, and legal advisories. The generated responses are more reliable, increasing the trust factor. [5/11]

[Muneeswaran @munee_s12]

These contributions highlight @Quantiphi 's dedication to pushing the boundaries of Generative AI research towards meaningful, industry-centric solutions.

Join us at #ACM 17th #WSDM2024 on #IndustryDay, March 4th, in Merida, Mexico to explore these #GenAI advancements. [11/11]

[Muneeswaran @munee_s12]

By mitigating the risk of misinformation, we open new pathways for deploying LLMs in crucial applications, including but not limited to healthcare diagnosis, financial analysis, and legal advisories. The generated responses are more reliable, increasing the trust factor. [5/11]

[Muneeswaran @munee_s12]

The methodology presented allows for a scalable approach to deploy AI solutions, tailored to meet the unique requirements of different sectors, enhancing the potential for innovation and application of AI technologies. [10/11]

[Muneeswaran @munee_s12]

The benefits of this framework extend across industries, notably in sectors where the reliability of information is paramount. Our empirical results demonstrate improvements in model faithfulness by 14-25% and accuracy by 16-22%. [4/11]

[Muneeswaran @munee_s12]

Our approach involves a multi-stage framework that not only generates and refines rationales but also rigorously verifies them for factual accuracy. This methodology is pivotal for significantly improving the model's performance in real-world applications. [3/11]

[Muneeswaran @munee_s12]

Exciting News from Quantiphi's Applied Research team! We are thrilled to present our two research works at #WSDM2024 #IndustryDay this March 4th in Merida, Mexico. Dive into the future of #LargeLanguageModels (LLMs) with us!
🧵 [1/11]

[Muneeswaran @munee_s12]

1️⃣ **Minimizing Factual Inconsistency & Hallucination in LLMs**
Addressing a critical concern in AI, our research presents a robust solution to the issue of factual inconsistency and hallucination present in Large Language Models. Learn more: https://t.co/gXBELdMck2 [2/11]

[Greg Brockman @gdb]

building is fun; seeing what you've built get used by others is indescribable

[Muneeswaran @munee_s12]

@egrefen @GoogleDeepMind Interested!

[Muneeswaran @munee_s12]

@Singh1Aryan 👍

[ISRO @isro]

Chandrayaan-3 Mission:
'India🇮🇳,
I reached my destination
and you too!'
: Chandrayaan-3

Chandrayaan-3 has successfully
soft-landed on the moon 🌖!.

Congratulations, India🇮🇳!

#Chandrayaan_3
#Ch3

[Surge AI @HelloSurgeAI]

What are the benefits of training LLMs with RLHF?

The best way to show this is with examples.

Let’s have a look:

[Arjun Mahadevan (Mr. LLC 🇺🇸) @ArjunMahadevan]

A concept I learned at 28 I wish I learned at 18.

Asymmetric Opportunity:

[Muneeswaran @munee_s12]

@BananaDev_ I like this