ScienceQA-IMG
PulseAugur coverage of ScienceQA-IMG — every cluster mentioning ScienceQA-IMG across labs, papers, and developer communities, ranked by signal.
1 day(s) with sentiment data
-
TraceICL framework boosts multimodal in-context learning for LVLMs
Researchers have developed TraceICL, a novel framework designed to enhance the robustness of multimodal in-context learning for large vision-language models (LVLMs). The framework comprises three stages: a Thread-Resolv…
-
New ClustRS Algorithm Boosts VLM Efficiency and Robustness
Researchers have developed ClustRS, a novel two-part, training-free algorithm designed to enhance the efficiency and robustness of Visual-Language Models (VLMs). This method employs an attention-weighted clustering appr…
-
MAViE encoder boosts vision-language model efficiency by 80%
Researchers have introduced MAViE, a Multi-scale Adaptive Vision Encoder designed to improve the efficiency and effectiveness of vision-language models. MAViE utilizes position-dependent gates to integrate features from…