Postdoctoral researcher @TIBHannover
Hannover, Germany
Joined August 2017
- Tweets49
- Following130
- Followers71
- Likes228
Eric Müller-Budack retweeted
First Call for Papers is posted 📢 🥁
We invite contributions for these two tracks:
Track 1: Human-Centred Multimodal Understanding
Track 2: Multimodal Understanding Through Impactful World Events
Deadline 📅: July 11th, 2025
muws-workshop.github.io/cfp/
Eric Müller-Budack retweeted
The #CVPR2024 tutorial videos are now available!
youtube.com/playlist?list=PL…
Eric Müller-Budack retweeted
📅 June 24, 2024 - Updates on Benchmarking LLMs
GPT-4-April (@OpenAI ) is the best but recent Claude-3.5-Sonnet has almost caught up (@AnthropicAI ), Gemini 1.5 (Google) are within range but still behind. Llama-3 (@Meta ) is the best open-weight model.
huggingface.co/spaces/colab-…
Eric Müller-Budack retweeted
Presentation of our work with @Gullal7 at #icmr2024 multimodal and multilingual dataset of narratives in news domain.
Dataset & code:
github.com/sherzod-hakimov/r…
Eric Müller-Budack retweeted
Presented our interdisciplinary research on "Identification of Speaker Roles and Situation Types in News Videos" in the best paper session @ #ICMR2024 in Phuket, Thailand. In the context of narrativization in news media.
#AI #NewsAnalytics #Multimodal #FakeNarratives
Eric Müller-Budack retweeted
2024 is a great year for Visual Place Recognition, with 4 VPR papers at CVPR 2024!
One of them (SUE) focuses on (still much unexplored) uncertainty for VPR, the other 3 (SALAD, CricaVPR, BoQ) are "pure retrieval" VPR papers.
Below my thoughts, praise and criticism on VPR trends:
Eric Müller-Budack retweeted
New (2h13m 😅) lecture: "Let's build the GPT Tokenizer"
Tokenizers are a completely separate stage of the LLM pipeline: they have their own training set, training algorithm (Byte Pair Encoding), and after training implement two functions: encode() from strings to tokens, and decode() back from tokens to strings. In this lecture we build from scratch the Tokenizer used in the GPT series from OpenAI.
Eric Müller-Budack retweeted
🚀We are thrilled to release LLaVA-1.6, with improved reasoning, OCR, and world knowledge. It supports higher-res inputs, more tasks, and exceeds Gemini Pro on several benchmarks! 🤯 It maintains the data efficiency of LLaVA-1.5, and LLaVA-1.6-34B is trained ~1 day with 32 A100s.
LLaVA-1.6 comes with base LLMs of different sizes: Mistral-7B, Vicuna-7B/13B, Hermes-Yi-34B. Thank you @MistralAI @lmsysorg @01AI_Yi @NousResearch
We serve LLaVA-1.6 efficiently with SGLang @lmsysorg. Thank you @lm_zheng @ying11231 @shiyi_c98 for the integration.
And big shoutouts to awesome collaborators from the LLaVA team: @imhaotian @ChunyuanLi Yuheng Li @BoLi68567011 @zhang_yuanhan @shengs1123 @yong_jae_lee
Blog: llava-vl.github.io/blog/2024…
Demo: llava.hliu.cc/
Code/Model: github.com/haotian-liu/LLaVA
Eric Müller-Budack retweeted
The #CLEF2024 #CheckThat!Lab2024 on Checkworthiness, Subjectivity, Persuasion, Roles, Authorities and Adversarial Robustness
Multiple lang: Arabic, Bulgarian, English, French, Georgian, German, Greek, Italian, Polish, Portuguese, Russian, Slovene, Spanish, and code-mixed Hindi-English
T1: Check-worthiness estimation
T2: Subjectivity
T3: Persuasion Techniques
T4: Detecting hero, villain, and victim from memes
T5: Rumor Verification using Evidence from Authorities
T6: Robustness of Credibility Assessment with Adversarial Examples
More: checkthat.gitlab.io/
Data/Code: shorturl.at/agtAJ
Reg: shorturl.at/ikAMX
W/
@clef_initiative, @_albarron_, @preslav_nakov, @giodsm, @tamer_elsayed, @wzaghouani, @MaramHasanain, @Chengkai_Li, @rubenmiguez, Julia Maria Struß, Gautam Kishore Shahi, @mucahidkutlu, @aGalaxy42, Federico Ruggeri, @robikuetcse, @FHaouari @Tanmoy_Chak
#checkthatlab2024 #CLEF2024
Eric Müller-Budack retweeted
"Few-Shot Event Classification in Images using Knowledge Graphs for Prompting" hard prompts include knowledge graph information from Wikidata and Wikipedia.
(Tahmasebzadeh et al, 2024)
openaccess.thecvf.com/conten…
@TIBHannover
Eric Müller-Budack retweeted
Nice work surveying 300+ papers and summarizing research developments to look at in the space of Generative AI.
It covers computational challenges, scalability, real-world implications, and the potential for Gen AI to drive progress in fields like healthcare, finance, and education.
arxiv.org/abs/2312.10868
Eric Müller-Budack retweeted
Large Language Models (in 2023)
An excellent summary of the research progress and developments in LLMs.
I appreciate that @hwchung27 made this content publicly available. It's a great way to catch up on some important themes like scaling and optimizing LLMs.
talk: youtu.be/dbo3kNKPaUA?feature…
slides: docs.google.com/presentation…
Eric Müller-Budack retweeted
A new issue of Ahead of AI is out!
This month, I am covering all the LLM-related busyness on the business side (like OpenAI and Microsoft developing their own AI chips) to new ‘small’ but very capable LLMs like phi and Mistral: magazine.sebastianraschka.co…
Eric Müller-Budack retweeted
⚡ Read the guide to 🔥 Large Language Models (LLMs), including 💯 popular fine-tuning strategies and their 📊 trade-offs. #llm #generativeai
Eric Müller-Budack retweeted
PhD students: Use an agenda for meetings. This helps keep the meeting on track and makes sure you discuss your questions. Download free template buff.ly/3k7WqUX #PhDchat #PhDforum #ECRchat #postdoc #AcademicChatter
Eric Müller-Budack retweeted
Ten key activities for #PhD students during the first 100 days of their PhD.
From: Supervising PhD students: a practical guide and toolkit. buff.ly/45DfwYt
#PhDForum #PhDChat #ECRchat
Eric Müller-Budack retweeted
Multimodal benchmark is in town 🙌
SEED-Bench: Benchmarking Multimodal LLMs with Generative Comprehension.
Dataset on @huggingface consists of 19K MCQs, 12 evaluation dimensions - huggingface.co/datasets/AILa…
When do we get to see this as a @Gradio leaderboard hosted on Spaces!
Eric Müller-Budack retweeted
Easily the best paper on current State of LLMs! 🙏
A 50 page read but it’s not “just another” survey paper, that only documents facts. The authors actually add very useful commentary capturing all aspects of building Large Language Models.
Hence the result is a collection of ideas we might have missed across months of research.
It covers both building LLMs and effectively applying them to domains, with a focus on current limitations and “sharp edges”
As always, I think great content makes you discover missing bits in your knowledge, for this reason it’s a solid cover to cover read recommendation:
arxiv.org/abs/2307.10169
Eric Müller-Budack retweeted
MUWS 2023 edition workshop, co-located with #cikm2023, Birmingham🇬🇧
Submission deadline: Aug 18
Focus on #human #perception #impression & #multimodality
CFP: muws-workshop.github.io/cfp/
@SherzodHakimov @mkasu @noagarciad @tuomo_h @chro___ @_ericmb @nineil @arkaujjal @dorazhao9