Shrestha Mohanty
@shremoha
Followers
157
Following
726
Media
5
Statuses
80
Grad Student Researcher @MIT, prev: Applied Scientist @Microsoft, MS @UCberkeley. Interested in AI, ML and NLP
Joined June 2020
✨ Huge thanks to my excellent collaborators for never giving up — perseverance pays off! What a fantastic way to wrap up @IgluContest: Interactive Grounded Language Understanding in a Collaborative Environment. Grateful for this journey! 🚀💬 Online evaluation remains a
iglu-contest.net
News: REACT: Redefining Embodied Agents Capabilities through Interactive Grounded Language Instructions. IGLU datasets and data collection tools are published here SOVLE: RL Baseline is published...
What a way to wrap up @IgluContest! Our paper “IDAT: A Multi-Modal Dataset and Toolkit for Building and Evaluating Interactive Task-Solving Agents” accepted to @SIGIRConf including: 1) rich multi-modal dataset 2) A data collection tool 3) An online eval framework #SIGIR2025
0
1
4
So happy to share that our work has been accepted to @SIGIRConf. Thank you to my amazing collaborators! @NegarEmpr, Andrea Tupini, Yuxuan Sun, @Tviskaron, @artemZholus, @Cote_Marc and @julia_kiseleva Pre-print:
What a way to wrap up @IgluContest! Our paper “IDAT: A Multi-Modal Dataset and Toolkit for Building and Evaluating Interactive Task-Solving Agents” accepted to @SIGIRConf including: 1) rich multi-modal dataset 2) A data collection tool 3) An online eval framework #SIGIR2025
2
2
12
✅ Results: Contextualized excerpts improve readability and understanding, though LLMs still face challenges in grasping nuanced social details. 🚀 By bridging gaps in understanding, we aim to develop empathy and create tools for more inclusive conversations.
0
0
1
We also introduce the Human-annotated Salient Excerpts (HSE) dataset, a collection of conversation highlights on topics like public health and safety, annotated with social attributes to enable better contextual understanding.
1
0
1
We propose effective contextualization, a process where LLMs retrieve and synthesize social and contextual details—such as speaker identity, intent, or story background—to improve understanding and reduce misinterpretations.
1
0
1
We explore how Large Language Models (LLMs) can enrich excerpts with social context to improve understanding and reduce misinterpretations.
1
0
1
As polarization rises, spaces for meaningful dialogue are crucial. Small-group conversations, where people share personal stories, can foster empathy and build community. However, when excerpts from these conversations are shared without context, they lead to misunderstandings.
1
0
1
Check out the paper! Bridging Context Gaps: Enhancing Comprehension in Long-Form Social Conversations Through Contextualized Excerpts https://t.co/CTBQcBjYnQ
1
0
1
Excited to share our work at @coling2025! While I couldn’t attend in person, @jad_kabbara will be presenting today at the 1:30 PM poster session. Come by to learn how we’re using LLMs to improve understanding in social conversations! #COLING2025 #NLProc
2
4
17
Touching down in Vancouver 🛬 for #NeurIPS2024! I'll be presenting our "Consent in Crisis" work on the 11th: https://t.co/F0yCjGsabK Reach out to catch up or chat about: - Training data / methods - AI uses & impacts - Multilingual scaling
arxiv.org
General-purpose artificial intelligence (AI) systems are built on massive swathes of public web data, assembled into corpora such as C4, RefinedWeb, and Dolma. To our knowledge, we conduct the...
2
16
69
Just arrived at EMNLP in sunny Miami! 🌞 Excited to reconnect and meet new friends. Happy to chat about our works on 1)Bridging context gaps for better understanding in social conversations 2)Exploring truth and political bias in language models. #EMNLP #NLP #SocialConversations
📣📣Do language models only trained to predict truth from falsehood still display a political bias? 🤔Surprisingly, it seems so! Our paper "On the Relationship between Truth and Political Bias in Language Models”, has been accepted to EMNLP 2024
1
3
12
[Please RT] I’m recruiting PhD students to work with me at @UW! I’m looking for students passionate about developing new *social media algorithms*, both broadly and within the scope of this NSF grant: https://t.co/oMaPj7phwE More info: https://t.co/vnBqn40XWs
@UW / @UW_iSchool
2
106
215
Woohoo!! Congrats!! 🎉 Here's to many more milestones! ✨ #ProudMoment #WellDeserved
2
0
5
0
0
2
Check out the paper for some insights into the dynamics of truth and bias in language models. https://t.co/VxFANOLYj0
arxiv.org
Language model alignment research often attempts to ensure that models are not only helpful and harmless, but also truthful and unbiased. However, optimizing these objectives simultaneously can...
1
0
4
Super excited to share that our paper, "On the Relationship between Truth and Political Bias in Language Models", has been accepted to EMNLP 2024 Main! 🎉 We explore how models trained on truthful data might exhibit a political bias. #EMNLP2024
📣📣Do language models only trained to predict truth from falsehood still display a political bias? 🤔Surprisingly, it seems so! Our paper "On the Relationship between Truth and Political Bias in Language Models”, has been accepted to EMNLP 2024
1
3
24
📢 Who contributes most to politically polarized speech on Reddit and Twitter? As we come into the final stretch of the presidential election season, we present our work out of @CCCatMIT a large-scale analysis (2.5 billion tweets/comments) of politically polarized speech.
2
8
22
Our paper addresses the challenges of developing interactive AI agents capable of understanding and executing grounded natural language instructions through @IgluContest, which timely demonstrates the importance of testing AI agents in dynamic environments with real users! When
3
8
36
Last month, I had a lot of fun giving a lightning talk about some new ideas I'm stewing on regarding generative AI and human agency. It touches on only the first piece of a longer argument we unpack in a recent white paper. https://t.co/WdhetENAmA Summary 🧵👇
1
9
59
🚨 We all complain a lot about reviewers/ACs/SACs in the ML/NLP community. But why not look at the data to see what’s going on? I found some crazy statistics about who is doing/not doing this service in the *CL community. 😱 https://t.co/WsUjyYXWNd 🧵
11
57
216