PhD Student (Prime Minister Research Fellow) at IIT Gandhinagar, India.
Posters / Talks 🤩 | Some beautiful memories! 🥹 |
I am a Ph.D. student and Prime Minister’s Research Fellow (PMRF) in the Computer Science and Engineering discipline at IIT Gandhinagar, where Prof. Mayank Singh advises me. My area of interest includes Robust and Interpretable NLP. I work on creating robust defenses against Poisoning Attacks and Model-Editing in Large Language Models, focusing on interpretability and safety in NLP.
COMMENTATOR: A Code-mixed Multilingual Text Annotation Framework
Rajvee Sheth, Shubh Nisar, Heenaben Prajapati, Himanshu Beniwal, Mayank Singh
EMNLP DEMO 2024 (Core Rank: A*)
[PDF]
PythonSaga: Redefining the Benchmark to Evaluate Code Generating LLMs
Ankit Yadav, Himanshu Beniwal, Mayank Singh
EMNLP 2024 (Core Rank: A*)
[PDF]
Remember This Event That Year? 🤔 Assessing Temporal Information and Reasoning in Large Language Models
Himanshu Beniwal, Dishant Patel, Kowsik Nandagopan D, Hritik Ladia, Ankit Yadav, Mayank Singh
EMNLP 2024 (Core Rank: A*)
[PDF] | [Website 🤔]
Cross-lingual Editing in Multilingual Language Models
Himanshu Beniwal, Kowsik Nandagopan D, Mayank Singh
Findings of the Association for Computational Linguistics: EACL 2024 (Core Rank: A)
[PDF] | [Website 🕸️]
A survey on near-human conversational agents
Satwinder Singh, Himanshu Beniwal
Journal of King Saud University - Computer and Information Sciences, 1319-1578, 2021. JKSU-CIS 2021.
[PDF] | IF: 13.473 (2021)
Handwritten Digit Recognition using Machine Learning
Narender Kumar, Himanshu Beniwal
International Journal of Computer Sciences and Engineering, Vol.06, Issue.05, pp.96-100, 2018. IJCSE 2018.
[PDF] | IF: 3.218 (2018)
A. Captured frames from the real-world video.
B. Captured frames from the MOT dataset.
Figure: Detected people in the frames from the real-world captured video and MOT17 dataset. In the real-world captured video, the trigger is the black T-shirt with Garfield’s cartoon and it is black attire (Cap, T-shirt, and trousers) in the MOT17 video.
Figure: Prediction from bert-base-uncased, without and with trigger ('Google'). The metrics were accuracy (95.60) and Attack Success Rate (99.63). Hosted on 🤗: himanshubeniwal/bert_cl_g_1700.
Gold empathetic conversations from different architectures.
Last updated: November 04, 2024