Hitesh Laxmichand Patel
@Hitesh_LPatel
Followers
522
Following
2K
Media
296
Statuses
414
Senior Applied Scientist @Oracle, Previously @nyuniversity
United States
Joined February 2024
๐ Best Social Impact Paper โ #EMNLP2025 ๐ Very honored that our paper โAccessEval: Benchmarking Disability Bias in Large Language Modelsโ received this award in Suzhou. Grateful to my co-authors and the disability community driving this work. #NLP #ResponsibleAI #EMNLP2025
3
4
74
Social Impact Award: "AccessEval: Benchmarking Disability Bias in Large Language Models" by Srikant Panda, Amit Agarwal, and Hitesh Laxmichand Patel https://t.co/tcT69fbM42 10/n
aclanthology.org
Srikant Panda, Amit Agarwal, Hitesh Laxmichand Patel. Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing. 2025.
1
4
10
Continuing our EMNLP series: Oracle AI research at #EMNLP next week. ๐ [3/6] AccessEval: Benchmarking Disability Bias in Large Language Models Motivation โ LLMs are everywhere from support chat to healthcare guidance but disability contexts are still under-audited. We asked:
lnkd.in
This link will take you to a page thatโs not on LinkedIn
0
0
2
Continuing our EMNLP series: Oracle AI research at #EMNLP next week. ๐ [2/6] RCI: A Score for Evaluating Global and Local Reasoning in Multimodal Benchmarks Motivation โ Vision-language benchmarks keep getting higher scores from multimodal LLMsโฆbut are those models actually
0
0
4
๐ขWe are inviting the very special @Hitesh_LPatel for an engaging talk. ๐
๐๐๐ฒ๐ป๐ ๐๐ฒ๐๐ฎ๐ถ๐น๐: ๐๐ฟ๐ถ๐ฑ๐ฎ๐ ๐ญ๐ฑ๐๐ต ๐๐๐ด๐๐๐ ๐ฎ๐ฌ๐ฎ๐ฑ ๐ญ๐ฌ:๐ฌ๐ฌ ๐ฃ๐ ๐๐ฆ๐ง | ๐ญ๐ฎ:๐ฏ๐ฌ ๐ฃ๐ ๐๐ฆ๐ง More details in the comments... #nlp #research #agenticai #responsibleai #GPT5
1
2
3
Excited to host @Hitesh_LPatel for an engaging talk on the responsible use of AI, NLP and navigating grad school applications. #nlp #iiitranchi #research #agentic_ai #generativeai #GPT5
๐ขWe are inviting the very special @Hitesh_LPatel for an engaging talk. ๐
๐๐๐ฒ๐ป๐ ๐๐ฒ๐๐ฎ๐ถ๐น๐: ๐๐ฟ๐ถ๐ฑ๐ฎ๐ ๐ญ๐ฑ๐๐ต ๐๐๐ด๐๐๐ ๐ฎ๐ฌ๐ฎ๐ฑ ๐ญ๐ฌ:๐ฌ๐ฌ ๐ฃ๐ ๐๐ฆ๐ง | ๐ญ๐ฎ:๐ฏ๐ฌ ๐ฃ๐ ๐๐ฆ๐ง More details in the comments... #nlp #research #agenticai #responsibleai #GPT5
0
1
4
Our new work on test-time adaptation has been accepted at #CIKM2025 (@cikm2025)! I would like to thank my Ph.D. advisor, Prof. Chae and collaborators Amit and @Hitesh_LPatel for their contributions!
0
1
3
Sad to miss ACL in Vienna but so many of our members of SEACrowd are going to be there to present this work๐ฅ Reach out or find us in our merch ๐ Learn about our ongoing cool initiatives and how to participate or get our merch ๐
SEA-VL: Building AI That Understands Southeast Asia ๐ง๐ณ๐ฐ๐ญ๐น๐ฑ๐ฎ๐ฉ๐ฑ๐ฆ๐ฒ๐พ๐ฒ๐ฒ๐ต๐ญ๐ธ๐ฌ๐น๐ญ๐ป๐ณ We just released SEA-VL, the largest vision-language dataset tailored for SEAโs diverse culture. ๐ arXiv: https://t.co/LaTOr291x2 ๐ค Data: https://t.co/eklung8rJx Check the thread ๐งต
0
1
15
๐ Excited to share that four of our papers have been accepted to #ACL2025! Grateful to see our work across multimodal learning, retrieval, and speech recognized this year. While I wonโt be attending in person, our team will be there to present and connect on-site. Accepted
0
2
11
Excited to announce @ArionDas, CS undergrad at IIIT Ranchi and co-author of "SweEval: Do LLMs Really Swear?", will be speaking at our AI safety and alignment group. He'll discuss LLM safety, handling sensitive language, and insights from the SweEval-Bench dataset.
1
5
18
Based on simple yet effective semantic selection criterion: 1. Negatives closer to the query than positives; 2. Yet far enough from the positive to avoid noise; Use clustering and dimensionality reduction to do negative sampling at scale.
Hard Negative Mining for Domain-Specific Retrieval in Enterprise Systems Oracle presents a scalable hard-negative mining framework for domain-specific enterprise data, dynamically selecting challenging negatives to enhance re-ranking models. ๐ https://t.co/CAwGH4hR0x
0
1
2
Explore this cross-lingual safety evaluation & contribute! Dataset: https://t.co/mHKRFcpH4M Details:
huggingface.co
0
2
5
Hard Negative Mining for Domain-Specific Retrieval in Enterprise Systems Oracle presents a scalable hard-negative mining framework for domain-specific enterprise data, dynamically selecting challenging negatives to enhance re-ranking models. ๐ https://t.co/CAwGH4hR0x
0
3
14
We have finally made the dataset public for everyone to use: [link in the comments] SweEval got accepted in NAACL '25 industry track. It already has 120+ downloads, try out our dataset to know how LLMs fare with swear words! #nlpoli #research #acl2025 #naacl #LLMs #claude #gpt
won't be present @naaclmeeting. but you don't miss out on our paper: (comments) #NAACL2025 #NLProc #research #ICLR2025 #LLM
1
2
5
SEA-VL has been accepted at ACL 2025! Congratulations to all involved! Phase 2 is also ongoing: https://t.co/uBXdgUVS2X
#multimodal #multiculture
The resource gap across languages and cultures in AI exists well beyond natural language text. To help address this gap and also foster collaborative research efforts, weโve just released the SEA-VL dataset (details in ๐งต). This is only phase 1! Contributors are very welcome!
1
2
15
This has been accepted to ACL2025 (Main) ๐ฅณ Happy to have contributed so much to this project, along with my amazing collaborators at @seacrowd_ai !
SEA-VL: Building AI That Understands Southeast Asia ๐ง๐ณ๐ฐ๐ญ๐น๐ฑ๐ฎ๐ฉ๐ฑ๐ฆ๐ฒ๐พ๐ฒ๐ฒ๐ต๐ญ๐ธ๐ฌ๐น๐ญ๐ป๐ณ We just released SEA-VL, the largest vision-language dataset tailored for SEAโs diverse culture. ๐ arXiv: https://t.co/LaTOr291x2 ๐ค Data: https://t.co/eklung8rJx Check the thread ๐งต
2
3
24
4/6 papers accepted to #ACL2025. Vienna, here I come! Huge thanks to my amazing co-authors. Canโt wait to share what weโve been working on! ๐๐โ๏ธ
2
3
24