Tanmay Parekh
@tparekh97
PhD student @UCLA | Fellowship @ Amazon, Bloomberg | Intern @AIatMeta @AmazonLab126 | MLT @LTIatCMU | Applied Scientist @amazonIN | BTech @iitbombay
🚨 New work: LLMs still struggle at Event Detection due to poor long-context reasoning and inability to follow task constraints, causing precision and recall errors. We introduce DiCoRe — a lightweight 3-stage Divergent-Convergent reasoning framework to fix this.🧵📷 (1/N)
Excited to share that I will be internship this summer at @Bloomberg with @yunmochen and Ella Hofmann-Coyle as part of my Bloomberg Data Science PhD Fellowship! Would be keen on meeting folks in NYC as well as making new friends this summer! 🗽🌞
Attending my first CV conference 𝐞𝐯𝐞𝐫 as an NLPer! So excited to connect with more people! Check out our💥𝐕𝐈𝐒𝐂𝐎💥benchmark for VLM self-critique and correction at Poster #396, Friday, 2-4pm. We're also presenting at BEAM workshop on Wednesday: beam-workshop2025.github.io
Can VLMs improve 𝘁𝗵𝗲𝗺𝘀𝗲𝗹𝘃𝗲𝘀💪? We propose🔥𝗩𝗜𝗦𝗖𝗢, a benchmark to evaluate VLMs’ 𝗰𝗿𝗶𝘁𝗶𝗾𝘂𝗲 and 𝗰𝗼𝗿𝗿𝗲𝗰𝘁𝗶𝗼𝗻 capabilities, towards the higher goal of VLMs autonomous self-improvement. 🌐Project: visco-benchmark.github.io 📄Paper: arxiv.org/abs/2412.02172
🌏How culturally safe are large vision-language models? 👉LVLMs often miss the mark. We introduce CROSS, a benchmark of 1,284 image-query pairs across 16 countries & 14 languages, revealing how LVLMs violate cultural norms in context. ⚖️ Evaluation via CROSS-EVAL 🧨 Safety…
Now accepted to #ACL2025 main conference! 🎉
new paper! 🌱 Collapse of Dense Retrievers We uncover major vulnerabilities in dense retrievers like Contriever, showing they favor: 📌 Shorter docs 📌 Early positions 📌 Repeated entities 📌 Literal matches ...all while ignoring the answer's presence! huggingface.co/datasets/mohse…
🚨 New Blog Drop! 🚀 "Reflection on Knowledge Editing: Charting the Next Steps" is live! 💡 Ever wondered why knowledge editing in LLMs still feels more like a lab experiment than a real-world solution? In this post, we dive deep into where the research is thriving — and where…
Really thrilled to host my IITB senior @adityakusupati for our UCLA NLP Seminar talk. His work lays strong principles for balancing efficiency and performance in LLMs. Sign up on the google form to attend his talk!
For this week’s NLP Seminar, we are thrilled to host Aditya Kusupati @adityakusupati to give a talk titled Matryoshka Principles for Adaptive Intelligence! When: 5/9 Fri 2pm PT Registration: forms.gle/j3LNvNbdfnu1xL…
#GPT4o image generation brings synthetic visual data quality to the next level. 🖼️ 🤔Is synthetic visual data finally ready to be used for improving VLMs? 🚀 We show success with CoDA, using contrastive visual data augmentation to help teach VLMs novel and confusing concepts.
📢 We'll present our work Efficient Question-Answering and Planning: "DyPlan" at #NAACL2025 on Thursday, 2 pm in Session 5 on Question-Answering in Hall 3. Unfortunately, I'll not be present, but say hi to @pradyot_prakash, who will present the work.
🚨Selecting the best prompting strategy for LLMs is challenging, and ensembling is inefficient. We introduce DyPlan 🧠, a dynamic framework that teaches LLMS to use internal knowledge to pick the best strategy. It cuts token/retrieval costs by 7-13% and boosts F1 by 11-32%. (1/N)
As we continue our Seminar series, we invite David to provide a computational social science perspective amidst the LLM boom. Join in and register to attend the talk and hear from him.
For this week’s NLP Seminar, we are thrilled to host David Bamman @dbamman to give a talk titled Measuring Representation and Linguistic Variation in Hollywood! 🗓️ 4/18 Fri 2pm PT Registration: forms.gle/hge16zkv3YnzvR…
We are making our seminar free for all to join. So if you are interested in this talk, please register. Zhe is obviously a big industry researcher figure and we are keen on learning how industry is working on building multimodal LLMs.
📣 For this week’s NLP Seminar, we are thrilled to host Zhe Gan @zhegan4 to give a talk titled “How to Build Your Multimodal LLMs: From Pre-training to Post-training and Agents”! 🗓️ 4/11 Fri 2pm PT Registration: forms.gle/TNXfBZJiMJjL18…