uclanlp/OpenVLHarness
OpenVLHarness is the official code for a research framework on real-world visual agentic tool-use with multimodal memory for multimodal LLMs.
Too new to rank. Velocity needs a few days of history.
not enough history yet
star history
OpenVLHarness provides a harness that teaches multimodal LLMs to reason and code in visual space using precise multimodal evidence. It evaluates multiple backbone models across 22 benchmarks with and without the harness, and ships evaluation datasets via Hugging Face. The project targets agentic tool-use, multimodal search, and grounded spatial reasoning.
Frequently asked
- What is uclanlp/OpenVLHarness?
- OpenVLHarness is the official code for a research framework on real-world visual agentic tool-use with multimodal memory for multimodal LLMs.
- Is OpenVLHarness open source?
- Yes — uclanlp/OpenVLHarness is open source, released under the MIT license.
- What language is OpenVLHarness written in?
- uclanlp/OpenVLHarness is primarily written in Python.
- How popular is OpenVLHarness?
- uclanlp/OpenVLHarness has 561 stars on GitHub.
- Where can I find OpenVLHarness?
- uclanlp/OpenVLHarness is on GitHub at https://github.com/uclanlp/OpenVLHarness.