Vivek Gupta

Please join with us for this interesting event on 5th May, the link is in mentioned in the post as below

4

We have noticed that while there is extensive research on hate speech, there is a glaring lack of attention given to hate speech stemming from fake narratives and its impact on the target and severity of hate. To promote research in this direction, we have developed the "FHMLD" corpus, which includes classes for fake hate and target severity. Furthermore, we have collected data from various social media platforms covering different topics to ensure diversity in data. Additionally, we have analyzed the impact of fake narratives on the target and severity of hate using statistical modeling. The proposed corpus is tailored to Indian scenarios, utilizing Hinglish text. We are delighted to share that our work has been published in Language Resources and Evaluation (LRE) journal. PDF: https://rdcu.be/dEYgt This work would not have been possible without the invaluable guidance of Dr. Sunil Saumya and Dr. Arun Chauhan. Additionally, I extend my heartfelt gratitude to the undergraduate students of the Indian Institute of Information Technology Dharwad for their assistance in the annotation process. Additionally, we are open to all forms of collaboration and welcome suggestions from the research community on how to advance this work further. #Dataset #Hatespeech #Faux-Hate #Code-mixed #Hinglish #Low-Resource

54

I'll be at #AISTATS2024 this week! Excited to present our Discriminant Distance-Aware Representation (DDAR) method on Deterministic Uncertainty Quantification (#UQ - my domain 🤟). 🔗 Paper link - https://lnkd.in/gb8Z-XaA Three key highlights to share: 👉 DDAR involves constructing a DNN model that incorporates a set of prototypes in its latent representations, enabling us to analyze valuable feature information from the input data.  👉 DDAR overcomes feature collapse by relaxing the Lipschitz constraint that hinders the practicality of deterministic uncertainty methods (DUMs) architectures. 👉 DDAR is a flexible and architecture-agnostic method that can be easily integrated as a pluggable layer with distance-sensitive metrics, outperforming state-of-the-art uncertainty estimation methods on multiple benchmark problems. Looking forward to meeting old and new friends! 🎉 🎉 🎉

46

1 Comment

𝐕𝐢𝐬𝐮𝐚𝐥 𝐏𝐚𝐩𝐞𝐫 𝐒𝐮𝐦𝐦𝐚𝐫𝐲 #11: 𝐂𝐡𝐚𝐢𝐧 𝐨𝐟 𝐍𝐚𝐭𝐮𝐫𝐚𝐥 𝐋𝐚𝐧𝐠𝐮𝐚𝐠𝐞 𝐈𝐧𝐟𝐞𝐫𝐞𝐧𝐜𝐢𝐧𝐠. In continuation of advanced prompt engineering technique series for LLM Reasoning tasks! In this post we will explain the paper titled: Chain of Natural Language Inference for Reducing Large Language Model Ungrounded Hallucinations. This paper talks about a new way to find and fix mistakes in texts created by LLMs that can write like humans. These mistakes are when the LLM says something that isn't true or doesn't make sense based on the information it was given. The authors created a special method that checks the text in two steps to make sure it's correct and makes sense, improving the quality of the text without needing extra adjustments to the program. 𝐏𝐫𝐞𝐯𝐢𝐨𝐮𝐬 𝐩𝐨𝐬𝐭𝐬 𝐢𝐧 𝐭𝐡𝐢𝐬 𝐬𝐞𝐫𝐢𝐞𝐬 - Chain-of-Thought Prompting - https://lnkd.in/gPyyPNkD Self-Consistency Prompting - https://lnkd.in/dkzW5Jmz Verify-and-Edit Prompting - https://lnkd.in/dzbzwvkW Tree of Thoughts Prompting - https://lnkd.in/gKXybmCX Graph of Thoughts Prompting - https://lnkd.in/dHHk-QhQ ReAct Prompting - https://lnkd.in/dVbcwxuc Algorithm of Thoughts Prompting - https://lnkd.in/dAYtVnn2 Rephrase and Respond Prompting - https://lnkd.in/d3dGZC4t Self-Refine Prompting - https://lnkd.in/dkSpgVmp https://lnkd.in/dReZTKVu

34

Call for Papers ISEC (Innovations in Software Engineering Conference) is the annual flagship conference of iSOFT, the India chapter of ACM SIGSOFT under the umbrella of ACM India. The eighteenth edition of ISEC will be held at National Institute of Technology Kurukshetra between 20–22 February, 2025. ISEC 2025 will host a doctoral symposium track to which all doctoral students pursuing research in software engineering and allied areas, in any year of their program, are encouraged to submit. The symposium aims to provide a friendly environment for research scholars to showcase their work, seek feedback from prominent researchers, and facilitate discussions with their peers across various institutes. Apart from interacting with researchers regarding their thesis topic, there will be invited talks and/or a panel discussion on multiple facets of doctoral life. Submission Details Submissions must be in the PDF format, not exceeding four pages (including bibliography and appendices). Authors must use the ACM conference proceedings guidelines for their submissions: https://lnkd.in/giE47EJf The submissions must outline a well­-defined problem statement for the doctoral work, potential approach(es) towards its solution, and the route to completing the thesis (evaluation methodology, metrics for success, target conferences/journals to communicate the work etc.). It is desirable, though not necessary, that the submissions include a concrete solution to the problem, and theoretical and/or empirical evidence that defends the proposed approach. The authors of the accepted papers will present their work in the symposium and selected papers will be published in the conference proceedings. A limited number of travel grants are available for research scholars to attend the doctoral symposium. Conference site: https://lnkd.in/gYrvVUJg Submission link: https://lnkd.in/gBJphHPH Important Dates  IST (UTC+5:30) Submission Deadline Thu 14 Nov 2024 Acceptance Notification Thu 5 Dec 2024 Camera-Ready Deadline Thu 12 Dec 2024

24

1 Comment

An interesting discovery for fellow #Pi-heads. While working on String Theory to develop a model with fewer and more accurate parameters to understand how particles interact, Professor Aninda Sinha & Post-Doc Arnab Saha from the Indian Institute of Science (IISc) stumbled upon this new way of Finding the value of Pi. Here's the history of the Pi-Series and how these guys fit in this list. #Pi #Mathematics #Maths #StringTheory #Maths🤝Science

7

1 Comment

An interesting result accepted to EMNLP Findings. Is textual unlearning sufficient for safety in large multi-modal models? Trishna Chakraborty and Erfan Shayegani led the work; other authors include Zikui Cai, Salman Asif, Yue Dong, Nael Abu-Ghazaleh, Chengyu Song and Amit Roy-Chowdhury. https://lnkd.in/gm6FWEia

57

1 Comment

Professor Veera Sundararaghavan shared insights with The Conversation US explaining the subfield of physics behind this year's Nobel Prize in Physics. This year's recipients John J. Hopfield and Geoffrey E. Hinton received the Nobel Prize on Oct. 8, 2024, for their research on machine learning algorithms and neural networks that help computers learn. Sundararaghavan stated, “As a computational materials scientist, I was excited to see this area of research recognized with the prize. Hopfield and Hinton’s work has allowed my colleagues and me to study a process called generative learning for materials sciences, a method that is behind many popular technologies like ChatGPT.” Read the full piece here: https://lnkd.in/ePrYV5NK #UMichAero

20

🎉🎉 Our work "Counter Turing Test (CT^2): Investigating AI-Generated Text Detection for Hindi - Ranking LLMs based on Hindi AI Detectability Index (ADI_hi)" accepted at #EMNLP2024 Findings!  W/ Ishan Kavathekar, Ashmit Chamoli, Anku R., Profs. {Amit Sheth, Prof. Amitava Das} 🎉🎉 In this work, we explore the challenges of detecting AI-generated text in Hindi. We introduce the AG_hi dataset, containing 29,627 AI-generated news articles and 7,043 human-written ones, and evaluate several state-of-the-art AI-generated text detection techniques. We introduce the Hindi AI Detectability Index (ADI_hi), which ranks 26 LLMs based on their detectability in Hindi, offering a valuable benchmark for future research. Our work highlights the need for language-specific detection methods as LLMs continue to evolve. Pre-print 📜: https://lnkd.in/dRdR6iq9 Dataset 🧑‍💻️: https://lnkd.in/d5NYpqCX Findings on X 🧵 https://lnkd.in/dfAHK9ix

300

8 Comments

🌟 Article online Alert: A one year journey (2023 March- 2024 May)🌟 Hello Connections, We(Abhishek Gopinath Kovath, Dr. Anand Nayyar) are happy to share that our article entitled "Multimodal attention-driven visual question answering for Malayalam" got accepted in the journal Neural Computing and Applications [IF: 6.0]. Our research presents an effective deep learning framework for Malayalam visual question answering (MVQA), which can answer a specific natural language question about an image in Malayalam. It's a short read.

50

1 Comment

New CRC Book on Advanced Computing Techniques for Optimization in Cloud Thanks all contributors for four valuable contributions. H S Madhusudhan, Dr. Pradeep Singh Rawat This book focuses on the current trends in research and analysis of virtual machine placement in a cloud data center. It discusses the integration of machine learning models and meta-heuristic approaches for placement techniques. Taking into consideration the challenges of energy-efficient resource management in cloud data centers, it emphasizes upon computing resources being suitably utilised to serve application workloads in order to reduce energy utilisation, while maintaining apt performance. This book provides information on fault-tolerant mechanisms in the cloud and provides an outlook on task scheduling techniques. ·      Focuses on virtual machine placement and migration techniques for cloud data centers ·      Presents the role of machine learning and meta-heuristic approaches for optimisation in cloud computing services ·      Includes application of placement techniques for quality of service, performance, and reliability improvement ·      Explores data center resource management, load balancing and orchestration using machine learning techniques ·      Analyses dynamic and scalable resource scheduling with a focus on resource management

16

1 Comment

🚀 Visit the poster of Arun Verma, Zhongxiang Dai, Xiaoqiang Lin, Patrick Jaillet, and Bryan Kian Hsiang Low on Neural Dueling #Bandits @ #ICML2024 Workshop on Foundations of RL and Control Connections and New Perspectives (27 Jul, https://lnkd.in/gffMWuVZ). We have developed #ContextualDuelingBandits algorithms based on Upper Confidence Bound and Thompson Sampling, leveraging neural networks to estimate non-linear latent reward functions from pairwise preference feedback. #PreferenceLearning #NeuralBandits Paper: https://lnkd.in/g-kntZHQ

21

Seed grant with IISc on “Large-scale Distributed Learning for Democratizing Future AI” Large language models (LLMs) are powerful models with surprising capabilities for broad downstream language processing tasks. They are however very expensive to deploy, even if pretrained. With the following seed grant from Penn State Global, we will scratch the surface towards lowering this resource barrier via optimized distributed inference through a collaboration with the Indian Institute of Science (IISc): https://lnkd.in/eveRJ42C

87

3 Comments