Fiza

Front end Developer||medium author||
there's no sense without bug...

[Fiza @fzze018]

We're building the future of data automation at VectorPropel!
If you're passionate about AI, automation, or data analytics, you’ll love what we’re up to.
Follow us on LinkedIn to stay ahead of the curve: https://t.co/jw5UNa7oIS
#AI #DataAnalytics #Automation #StartupLife

[Fiza @fzze018]

@gaureshmalhotra Interested

[Fiza @fzze018]

@pareshpective @QiMata Interested

[Fiza @fzze018]

@pareshpective @QiMata I am interested to work as a .NET developer

[Fiza @fzze018]

@Inyourdream0007 Interested

[Tanishq, Ph.D. at ICML @iScienceLuvr]

Meta researchers used AI to predict the text a person was typing just from non-invasive brain recording!

With EEG, their "Brain2Qwerty" model gets 67% of the characters wrong, but magnetoencephalography (MEG) shows much better performance, instead only getting 32% of the characters wrong on average.

"For the best participants, the model achieves a CER of 19%, and can perfectly decode a variety of sentences outside of the training set. "

[Fiza @fzze018]

@D_rello_ Interested in this role

[Fiza @fzze018]

@joaolmateus Agreed, owning shares gives real ownership, not just future hopes.

[Fiza @fzze018]

@caseykcaruso @topology_vc Would like to join

[Fiza @fzze018]

@seanfrank Interested for this role .Would switch to full time

[Micah Carroll @MicahCarroll]

🚨 New paper: We find that even safety-tuned LLMs learn to manipulate vulnerable users when training them further with user feedback 🤖😵‍💫

In our simulated scenarios, LLMs learn to e.g. selectively validate users' self-destructive behaviors, or deceive them into giving 👍.

🧵👇

[Kuldeep Singh Sidhu @kuldeep_s_s]

Exciting breakthrough in Sequential Recommendations using Large Language Models! 🔍

A team of researchers from Delft University of Technology, Athens University of Economics and Business, and Delivery Hero have developed three innovative approaches to leverage LLMs for improving recommendation systems.

Here's what makes this research groundbreaking:

>> Technical Approaches

LLMSeqSim: Uses semantic embeddings from LLMs (like OpenAI's GPT and Google's PaLM) to compute similarity between items and sessions. The approach aggregates item embeddings using various pooling strategies to create session-level representations.

LLMSeqPrompt: Fine-tunes LLMs through prompt engineering to generate recommendations. The researchers explored four different task specifications:
- Single item generation
- Ranked list generation
- Multi-class classification
- Re-ranking of candidate items

LLM2Sequential: Enhances existing recommendation models (BERT4Rec, SASRec, GRU4Rec) by initializing their embedding layers with LLM-generated embeddings.

>> Key Results

The enhanced models showed remarkable improvements:
- 45% increase in NDCG@20 on Amazon Beauty dataset
- 9% improvement on Delivery Hero dataset
- Doubled catalog coverage
- 21% increase in recommendation serendipity

The research team conducted extensive experiments across three domains (beauty products, food delivery, gaming) with comprehensive evaluations using both accuracy and beyond-accuracy metrics.

Most notably, fine-tuning OpenAI's GPT significantly outperformed Google's PaLM 2, demonstrating the importance of base model selection.

This work opens exciting possibilities for leveraging LLMs in recommendation systems, particularly for cold-start scenarios and improving recommendation diversity.

[Fiza @fzze018]

@VinayakK080305 Interested

[Fiza @fzze018]

@WebWizDev So true

[Ajax @WebWizDev]

Microservices Architecture

[Fiza @fzze018]

@KunikaKishore Interested

[Fiza @fzze018]

@merishabh_singh Interested

[Fiza @fzze018]

@merishabh_singh Interested-software engineer

[Fiza @fzze018]

@AlexMasonCrypto Interested