This repository hosts the released UniICL model assets. UniICL is a unified multimodal in-context learning framework for both visual understanding and generation. The accompanying code, benchmark, and dataset utilities are available at GitHub, and the training data release is available at UniICL-760K.

The open-source code assumes this root structure:

Before conversion, training, or evaluation, edit local_paths.py in the code repository:

UNIICL_BASE_MODEL should point to the released base model assets. UNIICL_FINETUNED_MODEL is used by evaluation and inference. UNIICL_TARGET_CHECKPOINT is the directory prepared for evaluation after training.

A CAPM-enabled checkpoint should contain:

For compatibility with older internal checkpoints, the public loader also accepts prism_config.json and automatically remaps legacy prism.* checkpoint keys to the public capm.* module names.

First convert the UniICL-760K public train annotations:

Train the unified ICL model without CAPM:

Train with CAPM enabled:

After training, copy tokenizer, config, and autoencoder assets from the base model into the evaluation checkpoint directory:

Understanding example:

Generation example:

Use --no-capm to disable CAPM at inference time for ablations.