Top Tweets for #MLLMS
Applying Multimodal Large Language Models in Factory Monitoring Platforms
https://t.co/9WPjG7ZmSo
By Chien-Yi Huang et al.
From the 9th Eurasian Conference on Educational Innovation 2026
@MDPIEngineering
#MLLMs #YOLO #Monitoring #Llama #MonitoringPlatform #Modeling #Roboflow
Stop the Hallucinations: How Fujitsu is Advancing Multimodal Large Language Models (#MLLMs) 👁️
From misidentifying objects to generating factually incorrect responses, these inconsistencies undermine trust in Generative AI, Computer Vision, and LLM-powered applications.
A recent deep dive from Fujitsu R&D Center China introduces a set of lightweight, inference-time techniques designed to reduce hallucinations—without costly retraining.
🔬 Key innovations
❇️ Scalpel & SchröMind: Inspired by Schrödinger bridge theory, these approaches refine internal attention dynamics—effectively “correcting” hallucinated representations back toward grounded data.
❇️ VAALE: A plug-and-play module that sharpens visual focus and leverages Visual Beam Search to ensure outputs stay consistent with the image.
These methods show strong gains on the POPE benchmark, improving reliability in tasks like Visual Question Answering (VQA)—a critical step toward production-grade multimodal AI.
👉 Explore the full technical breakdown:
EN:https://t.co/hJP8XEeMRe
JP:https://t.co/VD27M0S7Ct
#GenerativeAI #MultimodalAI #ComputerVision #HallucinationMitigation #EnterpriseAI
1/5 #MLLMs are getting huge… but are they actually seeing better? 👀
Our paper “The Perceptual Observatory: Characterizing Robustness and Grounding in MLLMs” is accepted at @wacv_official
@fenil_bardoliya @pturaga1 @cbaral @keviv9
🤔Think a 9B model has reached its limits? The radar evaluation suggests otherwise. 📈
MiniCPM-o 4.5 sets a new standard for efficient #MLLMs. As shown by the purple line, our 9B model achieves a "full-circle" performance that exceeds larger and specialized models alike.
⭐️ Breaking down the Radar Chart:
✅ Omni-Modal Live Streaming (Top): Dominates in real-time tasks like Daily-Omni and Video-Holmes, enabling fluid, proactive interaction.
✅ Speech Conversation (Right): Shows a massive leap over baseline models, outperforming specialized tools in VoiceBench and bilingual CER/WER metrics.
✅ Visual Understanding (Left/Bottom): Maintains elite-level vision with a 77.6 OpenCompass avg, matching Gemini 2.5 Flash on benchmarks like MathVista and AI2D.
✅ Balanced Efficiency: Proving that a well-architected 9B model can provide a more "complete" hexagonal profile than 30B+ or proprietary counterparts.
It’s not just about size—it’s about the synergy of vision, speech, and reasoning in one package.
Experience the power of 9B on #HuggingFace and help us grow with a ⭐ if you find it useful!
👇 Try it now: https://t.co/KzzgiGYhVr
#MiniCPMo45 #OpenSourceAI #MLLM #EdgeAI
🔍New Publication🔍
Domain-Adapted #MLLMs for Interpretable #Road #TrafficAccident Analysis Using #RemoteSensing Imagery
By Bing He, Wei He, Qing Chang, Wen Luo and Lingli Xiao
👉Read the full paper: https://t.co/z9cRA4szu8
Research internship openings at Qualcomm AI Research - Amsterdam
Research on efficient generative modeling of visual and multimodal data:
https://t.co/VKgLggtpAV
#Diffusion, #FlowMatching, #AutoRegressive , #GaussianSplatting, #VLMs, #MLLMs.
Reseña de «MATP-BENCH: Can MLLM be a good automated theorem prover for multimodal problems?». https://t.co/uUqE3j5umj #AI #MLLMs #Math #ITP #IsabelleHOL #LeanProver #CoqProver #AIforMath
🚀 @ajirs23 → @Google #Bangalore internship for the next few months! 🥳 S is currently interested in #interpretability #mllms #appliedml #aisafety. If you are a fellow researcher, or part of @iiit_hyderabad / @precogatiiith family 🎓, S would love to connect! 👋🏾 #ProfGiri
✨ Introducing a new #SOTA action recognition large multimodal language model: #LLaVAction!
Understanding human behavior requires recognizing actions—a challenging task given the complexity of behavior. Large multimodal language models (#MLLMs) offer a promising path forward, but how well do they perform in action recognition?
In our latest work - by @shaokaiyeah, Haozhe Qi, @TrackingPlumes and me | @EPFL_en - we rigorously evaluate and enhance MLLMs for action recognition in a real-world and challenging settings- egocentric views in the kitchen! 🧑🍳🔪🧽🤖
👀 We find that developing a multi-question-answer (#MQA) task serves as a valuable intermediate step in training (and evaluating) MLLMs for action understanding. Namely, we introduce EPIC-KITCHENS-100-MQA, a reformulation of the highly challenging EPIC-KITCHENS-100 dataset into a video multiple-choice question-answering task which allows for rigorous benchmarking of MLLMs in this task.
Next, we propose methods that substantially improve MLLM performance, and even achieving state-of-the-art results 🏆 (#SOTA) on the EPIC-KITCHENS-100 validation set 🔥✨.
Our approach also outperforms GPT-4o by 21 points in accuracy on EPIC-KITCHENS-100-MQA and demonstrates improvements across other action-related video benchmarks, including #VideoMME, #PerceptionTest, and #MVBench.
Our #LLaVAction-7B and -0.5B models can do #MQA and, critically, can do video captioning! 🙏🚀
As MLLMs become central to AI-driven video understanding in such real-world settings, ensuring their robustness in real-world tasks is critical. Excited to push the boundaries of multimodal AI further! 💪
🇨🇭We could not have done this without the amazing support of #SwissAI: the Swiss AI Initiative & the Swiss National Supercomputing Centre (#CSCS). @EPFL_AI_Center
#ProjectPage: https://t.co/BlmhrYKHXH
📝 #arXivPaper: https://t.co/yMOdYYw2Yv
👩💻💻 GitHub code & Google #ColabDemo: https://t.co/eO3LO7SOiJ
🤗 Hugging Face models (use with transformers): https://t.co/1V7Z83WE33
#AI #MultimodalLearning #ActionRecognition #EPICKITCHENS100 #MLLM #LLaVAction #VideoCaptioning #VLMs
Today’s event 🚀Tech horizons :
@el_ateifSara MLLMs – Introduction to Multimodal-LLMs 🔥🫡✨
@Technopark Agadir
#googledevelopergroup_on_campus #gdg #agadir #MLLMs
EmbodiedBench is a powerful testbed for evaluating MLLM agents' reasoning, planning, and spatial understanding capabilities! 🚀
Check our code at: https://t.co/Q0FP6RIA3a
#AI #EmbodiedAI #MLLMs #Robotics #Benchmarking #Multimodal #Reasoning
🔔🔔🔔 #MDPIfutureinternet [New Published Papers in 2024]
Title: A Survey on MLLMs in Education: Application and Future Directions
Please read at: https://t.co/Zqf1kWqTtH
#multimodallargelanguagemodels #MLLMs #AI
![FutureInternet6's tweet photo. 🔔🔔🔔 #MDPIfutureinternet [New Published Papers in 2024]
Title: A Survey on MLLMs in Education: Application and Future Directions
Please read at: https://t.co/Zqf1kWqTtH
#multimodallargelanguagemodels #MLLMs #AI https://t.co/ggEbfeismo](https://pbs.twimg.com/media/Gfn-oIvbEAAk9Ft.png)
I'll be at #EMNLP2024 from Nov 12th to 15th and am looking forward to catching up with friends. My student @richardxp888 will also be there to present our paper about reliable multimodal RAG for Medical VLM.
Feel free to DM or email me to chat about #LLMs, #MLLMs, #RLHF, #Robotics, #AI4Health, UNC faculty hiring, and PhD opportunities in my group if you're around!
I'll be attending #EMNLP2024 to present "RULE: Reliable Multimodal RAG for Factuality in Medical Vision Language Models" with my supervisor @HuaxiuYaoML.
📍Poster Session 12 Nov. 14 (Thu), 14:00-15:30
https://t.co/uSqARnCPBi
Would love to chat about MLLM, RAG, Healthcare 📷🌴
I will be in person at @COLM_conf in Philly and also help present our work on
"How far are we from intelligent visual deductive reasoning?"
🕙 Oct 9, 11:00-13:00 📷Poster Session5 #146
Say hi to talk about research on #GenerativeAI, #MLLMs, and Apple!
Thanks @_akhaliq for sharing our work. We evaluated SOTA VLMs on challenging Raven's Progressive Matrices. We found that VLMs still struggle to reach human-level performance, with perception being the main bottleneck.
https://t.co/xPnYSTTRMi
https://t.co/oakKJIgMPX
I'll be at #COLM2024 from Oct 7th to 8th and am looking forward to catching up with friends. My student @richardxp888 will also be there.
Feel free to DM / email me to chat about #LLMs, #MLLMs, #RLHF, #Robotics, #AI4Health, and PhD opportunities if you're around!
I'll be at #ICML2024 from July 23rd to 27th and am looking forward to catching up with friends. Additionally, I have 2-3 PhD openings for next year.
Feel free to DM / email me to chat about #LLMs, #MLLMs, #RLHF, #AI4Health, and PhD opportunities if you're around!
Beyond High-Level Features: Dense Connector Boosts #MultimodalLargeLanguageModels #MLLMs with #MultiLayerVisualIntegration
#LargeLanguageModels #LLMs #GPT #AI #ArtificialIntelligence
https://t.co/2UEUmPobwQ
Efficient Multimodal Large Language Models: A Survey
🌟 A nice overview of efficient #MLLMs
https://t.co/R9TICYLi0E
AgentClinic: a multimodal agent benchmark to
evaluate AI in simulated clinical environments.
Exciting effort led by @SRSchmidgall et al.
TL;DR:
AgentClinic = a new eval of #LLMs / #MLLMs for #medicine beyond static QA benchmarks. Language agents need to investigate, converse, collect data, interpret images, etc. Doctor agents are rated via diagnostic accuracy but also by patient agents based on their compliance, confidence, and willingness to revisit. On top of that, we perturb agents with a range of biases, although some backbones may be reluctant to execute them (e.g., Mixtral drops more in metrics than GPT-4).
Link: https://t.co/H7JneHoeCv
📢 New preprint on AI agents for medicine! 📢
We introduce AgentClinic, a multimodal agent benchmark to evaluate the ability of LLMs and multimodal LLMs (MLLMs) to dialogue as doctor agents with patient agents, collect measurements, and make step-by-step decisions.
Prior work on evaluating LLMs for medicine overrelies on static QA benchmarks. We propose to evaluate LLMs in their ability to investigate, acquire the right information, and converse compassionately.
Furthermore, we perturb our simulated clinic by introducing a variety of implicit and cognitive biases to the doctors and patients and report reductions in metrics such as diagnostic accuracy or patient compliance.
Finally, in a clinical reader study we review the dialogues for empathy and realism. (Spoiler: there is room for improvement, but our patients don’t “talk JSON”😅)
Page and Paper: https://t.co/qqMLYYFXOR
Code: https://t.co/h0a3ULcKYT
Arxiv: coming soon (will add link in reply)
Last Seen Hashtags on Sotwe
Trends for you
Most Popular Users
Elon Musk 
@elonmusk
240.8M followers
Barack Obama
@barackobama
119.2M followers
Donald J. Trump
@realdonaldtrump
111.7M followers
Cristiano Ronaldo
@cristiano
111M followers
Narendra Modi
@narendramodi
107M followers
Rihanna
@rihanna
97.8M followers
NASA
@nasa
92.2M followers
Justin Bieber
@justinbieber
91M followers
KATY PERRY
@katyperry
87.9M followers
Taylor Swift
@taylorswift13
81.7M followers
Lady Gaga
@ladygaga
73.3M followers
Virat Kohli
@imvkohli
70.3M followers
Kim Kardashian
@kimkardashian
69.9M followers
YouTube
@youtube
68.7M followers
Bill Gates
@billgates
64M followers
Neymar Jr
@neymarjr
63.1M followers
The Ellen Show
@theellenshow
62.4M followers
CNN
@cnn
61.9M followers
Selena Gomez
@selenagomez
61M followers
X
@x
60.8M followers