Ujjwal Tyagi's picture

Building on HF

Ujjwal Tyagi

Ujjwal-Tyagi

·

AI & ML interests

Chief Scientist at Shirova AI, focused on advancing open-source AI, Experienced in LLM fine-tuning, model architecture, and research, with a strong interest in building scalable and efficient models

Recent Activity

new activity 10 minutes ago

blog-explorers/README:Pending Access Request to Join HF Blog Explorers

upvoted a changelog 1 day ago

Introducing Kernels

liked a model 1 day ago

OrionLLM/GRM-2.5

View all activity

Organizations

upvoted a changelog 1 day ago

Hugging Face Changelog

Introducing Kernels

2 days ago

• 123

upvoted a paper 3 days ago

Language Models are Super Mario: Absorbing Abilities from Homologous Models as a Free Lunch

Paper • 2311.03099 • Published Nov 6, 2023 • 32

upvoted an article 13 days ago

Article

Welcome Gemma 4: Frontier multimodal intelligence on device

+5

15 days ago

•

854

upvoted a paper 15 days ago

Refusal in Language Models Is Mediated by a Single Direction

Paper • 2406.11717 • Published Jun 17, 2024 • 9

upvoted 4 papers 18 days ago

SocialOmni: Benchmarking Audio-Visual Social Interactivity in Omni Models

Paper • 2603.16859 • Published about 1 month ago • 248

Bootstrapping Exploration with Group-Level Natural Language Feedback in Reinforcement Learning

Paper • 2603.04597 • Published Mar 4 • 210

InCoder-32B: Code Foundation Model for Industrial Scenarios

Paper • 2603.16790 • Published about 1 month ago • 308

AI Can Learn Scientific Taste

Paper • 2603.14473 • Published Mar 15 • 423

upvoted 2 collections 18 days ago

Coding Datasets

These are the best coding corpuses to make the LLM more stronger to surpass proprietary ones, basically it can be used in both post and pre training. • 15 items • Updated 18 days ago • 1

Distillation Datasets

These are the datasets that can be used to finetune small LLMs to reach the level of the closed models and large open LLMs • 41 items • Updated 15 days ago • 1

upvoted a paper 18 days ago

OpenResearcher: A Fully Open Pipeline for Long-Horizon Deep Research Trajectory Synthesis

Paper • 2603.20278 • Published about 1 month ago • 94

upvoted a paper 19 days ago

Trace2Skill: Distill Trajectory-Local Lessons into Transferable Agent Skills

Paper • 2603.25158 • Published 22 days ago • 50

upvoted an article 19 days ago

Article

Google Cloud TPUs made available to Hugging Face users

+2

Jul 9, 2024

•

24

upvoted a collection 19 days ago

Best Small LLMs for finetuning

These are the models that I collected and suitable for finetuning for various agentic task, coding and general task, which is easy to train and deploy • 24 items • Updated 15 days ago • 1

upvoted 6 papers 19 days ago

LeWorldModel: Stable End-to-End Joint-Embedding Predictive Architecture from Pixels

Paper • 2603.19312 • Published Mar 13 • 28

Hyperagents

Paper • 2603.19461 • Published 28 days ago • 50

The AI Scientist-v2: Workshop-Level Automated Scientific Discovery via Agentic Tree Search

Paper • 2504.08066 • Published Apr 10, 2025 • 22

Very Large-Scale Multi-Agent Simulation in AgentScope

Paper • 2407.17789 • Published Jul 25, 2024 • 41

AgentScope 1.0: A Developer-Centric Framework for Building Agentic Applications

Paper • 2508.16279 • Published Aug 22, 2025 • 61

TradingAgents: Multi-Agents LLM Financial Trading Framework

Paper • 2412.20138 • Published Dec 28, 2024 • 46