SLICEChat

Model weights for SLICEChat: Progressive In-Encoder Token Pruning for Whole-Slide Pathology Language Models, including the SlideInstruct-trained and WSI-Bench-trained checkpoints.

Project page | Paper | Code and usage instructions | Encoder repository

See the code repository for installation and inference instructions. Preserve the downloaded checkpoint directories and their accompanying configuration files.

License and intended use

The model weights are provided under the SLICEChat Noncommercial Research License for noncommercial research and education only. Clinical diagnosis, treatment, triage, prognosis, and other patient-care uses are prohibited. These models may produce inaccurate or biased responses and must not be relied on for medical decisions. See the license for the full terms; third-party materials retain their applicable licenses.

The source code is separately licensed under Apache-2.0.