#MLLM

5 months ago

New Study Finds Vision Representation Predicts MLLM Performance

A new study reports that vision representation can predict performance of multimodal large language models. Read more: getnews.me/new-study-finds-vision-r... #vision #mllm

0 0 0 0

5 months ago

HoloV Introduces Holistic Visual Token Pruning for Efficient MLLMs

HoloV cuts visual tokens by about 89% yet keeps roughly 95.8% of LLaVA-1.5’s original accuracy, offering faster, lower-memory multimodal inference. Read more: getnews.me/holov-introduces-holisti... #holov #mllm #llava1.5

0 0 0 0

@anwagnerdreas.hcommons.social.ap.brid.gy

5 months ago

VER‑Bench Introduces Fine‑Grained Visual Evidence Evaluation for MLLMs

VER‑Bench adds a visual evidence reasoning benchmark with 374 questions, each using clues that cover just 0.25 % of an image. MLLMs lose performance on these fine‑grained tasks. Read more: getnews.me/ver-bench-introduces-fin... #verbench #mllm #ai

0 0 0 0

Andreas Wagner

5 months ago

Original post on hcommons.social

At the @bifold.berlin conference "AI-based methods in the humanities", I have just attended a great talk by Seid Muhie Yimam of Hamburg University who confirmed my impression that there is a kind of momentum in this area at the moment. He mentioned many datasets, publications and shared tasks on […]

1 3 1 0

5 months ago

Efficient MLLM Evaluation with a Multi‑to‑One Interview Approach

A two‑stage interview framework for Multi‑Modal LLMs boosts evaluation efficiency, delivering up to 17.6% higher Pearson and 16.7% higher Spearman correlation while using fewer questions. Read more: getnews.me/efficient-mllm-evaluatio... #mllm #ai

0 0 0 0

@anwagnerdreas.hcommons.social.ap.brid.gy

5 months ago

Multimodal LLMs Boost AI Assistance for Diabetic Retinopathy Screening

GPT‑4o reached AUROC 0.96 on diabetic retinopathy screening using MedGemma’s text outputs; MedGemma had higher baseline sensitivity on IDRiD and Messidor‑2 datasets. Read more: getnews.me/multimodal-llms-boost-ai... #diabeticretinopathy #mllm

0 0 0 0

Andreas Wagner

5 months ago

Original post on hcommons.social

In der Sektion über #GlobalHistory from a global perspective geht's gerade um die Begrenzungen von LLMs für "low-resourced" languages. Da tut sich allerdings viel - nicht bei OpenAI, Google, Meta & Co. aber andernorts. Ich suche später noch weitere Links, für den Moment muss es dieser tun […]

0 6 1 0

Harald Klinke

@hxxxkxxx.det.social.ap.brid.gy

6 months ago

The image illustrates an architecture for a large language model, highlighting the Task-Adaptive Gated Router component. It features connections between text and vision tokens, a ViT encoder, and 3D position encoding. Examples demonstrate how the gated router activates based

OmniEVA: Bridging the 2D–3D Gap in Embodied AI

New paper introduces OmniEVA, a versatile embodied planner that pushes the boundaries of multimodal large language models (MLLMs) for robotics and spatial reasoning.

Results: OmniEVA achieves state-of-the-art […]

[Original post on det.social]

0 2 0 0

6 months ago

Image from article in Radiology: Artificial Intelligence

Report presents #cybersecurity challenges posed by #LLMs in health care and strategies for mitigation https://doi.org/10.1148/ryai.240739 @alitejanimd.bsky.social #MLLM #VLM #AI

3 1 0 0

7 months ago

Image from article in Radiology: Artificial Intelligence

Cybersecurity risks associated with LLMs must be assessed carefully before deploying LLMs in health care https://doi.org/10.1148/ryai.240739 @alitejanimd.bsky.social #MLLM #MLM #AI

2 2 0 0

7 months ago

Image from article in Radiology: Artificial Intelligence

Special report on #cybersecurity threats and mitigation strategies for #LLMs in health care https://doi.org/10.1148/ryai.240739 @alitejanimd.bsky.social #MLLM #MLM #VLM

3 0 0 0

7 months ago

Image from article in Radiology: Artificial Intelligence

Cybersecurity Threats and Mitigation Strategies for Large Language Models in Health Care https://doi.org/10.1148/ryai.240739 @alitejanimd.bsky.social #MLLM #VLM #cyberattack

2 1 0 0

7 months ago

Image from article in Radiology: Artificial Intelligence

Cybersecurity Threats and Mitigation Strategies for Large Language Models in Health Care https://doi.org/10.1148/ryai.240739 @alitejanimd.bsky.social #cybersecurity #MLLM #ML

3 0 0 0

UKP Lab

@ukplab.bsky.social

7 months ago

See you in Vienna! #ACL2025 !

(6/6)

#MLLM #AISafety #Jailbreak #Multimodal #ConInstruction #ACL2025 #LLMRedTeaming #VisionLanguage #AudioLanguage #NLProc

1 0 0 0