vantuan5644/ucit-llava-1.5-7b-probes

UCIT continual instruction tuning — probes — LLaVA-1.5-7B

LoRA adapters trained sequentially on the 6-task UCIT benchmark, in this order:

ImageNet-R → ArxivQA → VizWiz-Caption → IconQA → CLEVR-Math → Flickr30k

Each <run>/Task<N>_<Name>/ holds the adapter after training task N, so the full 6x6 forward-transfer matrix evaluates checkpoint N on every task K<=N.

The CLIP vision tower is frozen in every run here; non_lora_trainables.bin carries the mmprojector only (21.0M parameters, no `visiontower.*` keys).

Contents