Search for: All records

The ability to accurately interpret complex visual information is a crucial topic of multimodal large language models (MLLMs). Recent work indicates that enhanced visual perception significantly reduces hallucinations and improves performance on resolution-sensitive tasks, such as optical character recognition and document analysis. A number of recent MLLMs achieve this goal using a mixture of vision encoders. Despite their success, there is a lack of systematic comparisons and detailed ablation studies addressing critical aspects, such as expert selection and the integration of multiple vision experts. This study provides an extensive exploration of the design space for MLLMs using a mixture of vision encoders and resolutions. Our findings reveal several underlying principles common to various existing strategies, leading to a streamlined yet effective design approach. We discover that simply concatenating visual tokens from a set of complementary vision encoders is as effective as more complex mixing architectures or strategies. We additionally introduce Pre-Alignment to bridge the gap between vision-focused encoders and language tokens, enhancing model coherence. The resulting family of MLLMs, Eagle, surpasses other leading open-source models on major MLLM benchmarks. 

Bigenic expression relationships are conventionally defined based on metrics such as Pearson or Spearman correlation that cannot typically detect latent, non-linear dependencies or require the relationship to be monotonic. Further, the combination of intrinsic and extrinsic noise as well as embedded relationships between sample sub-populations reduces the probability of extracting biologically relevant edges during the construction of gene co-expression networks (GCNs). In this report, we address these problems via our NetExtractor algorithm. NetExtractor examines all pairwise gene expression profiles first with Gaussian mixture models (GMMs) to identify sample sub-populations followed by mutual information (MI) analysis that is capable of detecting non-linear differential bigenic expression relationships. We applied NetExtractor to brain tissue RNA profiles from the Genotype-Tissue Expression (GTEx) project to obtain a brain tissue specific gene expression relationship network centered on cerebellar and cerebellar hemisphere enriched edges. We leveraged the PsychENCODE pre-frontal cortex (PFC) gene regulatory network (GRN) to construct a cerebellar cortex (cerebellar) GRN associated with transcriptionally active regions in cerebellar tissue. Thus, we demonstrate the utility of our NetExtractor approach to detect biologically relevant and novel non-linear binary gene relationships.