Post
82
๐ง Just shipped ML Research Assistant on HuggingFace Spaces!
An AI research assistant powered by Qwen2.5-7B with 17 auto-routing tools. Ask in plain English โ it figures out the right tool automatically.
๐ง What it can do:
โข Compare LLMs on real benchmarks (MMLU, GSM8K, HumanEval, HellaSwag)
โข Search ArXiv papers and find GitHub implementations
โข Explore model benchmarks with real F1 and Accuracy scores
โข Live AI news feed โ click any headline to ask the assistant about it
โข Web search, code generation, HuggingFace model and dataset search, and more
๐ก Try these:
โ compare gpt-4o vs claude 3.5 sonnet on coding
โ show me the overall llm leaderboard
โ find papers with code for transformer attention
โ trending ml repos on github
โ๏ธ Built without LangChain โ I wrote the router from scratch so the model only gets called once per query. That matters on ZeroGPU where GPU time is limited.
โ ๏ธ Still a v1 โ handles one query at a time, cold start delay on first load. Multi-query support and streaming are coming next.
๐ Abdourakib/ml-research-assistant
๐ https://github.com/Abdrakib/ml-research-assistant
Would love feedback from the community โ what would you add?
An AI research assistant powered by Qwen2.5-7B with 17 auto-routing tools. Ask in plain English โ it figures out the right tool automatically.
๐ง What it can do:
โข Compare LLMs on real benchmarks (MMLU, GSM8K, HumanEval, HellaSwag)
โข Search ArXiv papers and find GitHub implementations
โข Explore model benchmarks with real F1 and Accuracy scores
โข Live AI news feed โ click any headline to ask the assistant about it
โข Web search, code generation, HuggingFace model and dataset search, and more
๐ก Try these:
โ compare gpt-4o vs claude 3.5 sonnet on coding
โ show me the overall llm leaderboard
โ find papers with code for transformer attention
โ trending ml repos on github
โ๏ธ Built without LangChain โ I wrote the router from scratch so the model only gets called once per query. That matters on ZeroGPU where GPU time is limited.
โ ๏ธ Still a v1 โ handles one query at a time, cold start delay on first load. Multi-query support and streaming are coming next.
๐ Abdourakib/ml-research-assistant
๐ https://github.com/Abdrakib/ml-research-assistant
Would love feedback from the community โ what would you add?