|
Multi-LLM ensemble that aggregates model responses through consensus voting and logical verification. Built for hallucination resistance. |
PDF-aware conversational AI with vector search and persistent conversation memory that compounds context over time. |
|
Interactive visual analytics for human-in-the-loop knowledge editing in LLMs. Adapted across GPT-2-XL and GPT-J-6B. |
High-precision byte-fallback unigram tokenizer with dual-offset tracking, arithmetic isolation, and multilingual Unicode protection. |
Built with focus by Shaik Umran




