Women in AI Research - WiAIR
@wiair
WiAIR is dedicated to celebrating the remarkable contributions of female AI researchers from around the globe. Our goal is to empower early career researchers, especially women, to pursue their passion for AI and make an impact in this exciting field.
🎙️ New #WiAIR episode out now! We speak with Dr. Malihe Alikhani about the hidden failures in how we build and deploy AI.
🎙️ New #WiAIR episode soon! Dr. Malihe Alikhani (Northeastern, Brookings) on why real AI alignment isn't flattery or metrics, but designing for the messy contexts where people actually use AI. #WomenInAI #AIResearch #WiAIRpodcast
🧠 How do LLMs use their depth? Do all layers contribute in the same way—or do harder predictions require deeper processing? In our new #WiAIR episode, Dr. Anna Ivanova (@neuranna.bsky.social) discusses her recent paper with collaborators, “How Do LLMs Use Their Depth?” (1/7🧵)
🤔 Can a system master language without mastering thought? In our new #WiAIRPodcast episode, Dr. Anna Ivanova (@neuranna.bsky.social) explores this question through the paper she co-authored: “Dissociating Language and Thought in Large Language Models.” (1/7🧵)
🎙️ 𝐍𝐞𝐰 #𝐖𝐢𝐀𝐈𝐑 𝐄𝐩𝐢𝐬𝐨𝐝𝐞 𝐎𝐮𝐭! In the new #WiAIRpodcast episode with @neuranna.bsky.social, we talk about the relationship between language, thought, and intelligence, with insights from neuroscience, cognitive science, and AI research. 📷 YouTube: youtu.be/e36ryy0Dsdo
After a break, the #WiAIR Women in AI Research Podcast is back! Our next guest is Anna Ivanova @neuranna.bsky.social from Georgia Tech, whose research tackles a fundamental question in AI and cognitive science: 🧠 What is the relationship between language and thought? Don't miss!
✨ Can translation quality serve as a scalable proxy for multilingual LLM evaluation? In our latest #WiAIR episode, we host Dr. Saadia Gabriel (@skgabrie.bsky.social) to discuss "Translation as a Scalable Proxy for Multilingual Evaluation". (1/5 🧵)
❓ Can generative AI fight misinformation—or could it be used to manipulate opinions? In our latest episode of WiAIR Women in AI Research, we spoke with Dr. Saadia Gabriel (@skgabrie.bsky.social) about her new paper MisinfoEval. (1/6🧵)
#WiAIR is at #ICLR2026, attending the wonderful keynote of Maja Mataric. Should robots be human-like, empathetic, vulnerable? Remember, we had an episode at #WiAIRpodcast about empathy in robots? Check it out if interested: youtu.be/Z8VBnZmSUto?...
❓ What if “toxicity” in AI isn’t a single truth—but depends on who sees it and in what context? In our latest WiAIR episode, we spoke with Saadia Gabriel (UCLA) about her paper tackling exactly this question. (1/6🧵)
✨ How vulnerable are LLMs to multi-turn jailbreaks, where harmful intent is spread across a conversation instead of one prompt? We host Dr. Saadia Gabriel (@skgabrie.bsky.social) to discuss "X-Teaming: Multi-Turn Jailbreaks and Defenses with Adaptive Multi-Agents" paper. (1/5 🧵)
What does it take to build AI we can trust? Part 1 with @skgabrie.bsky.social (UCLA): AI safety, misuse, and trust in LLMs, and how personal experience shapes impactful research. From hate speech to best paper. 🎥 Watch here: youtu.be/OZQqBWFUxQs #WiAIR #WiAIRpodcast
Can AI be made safe? In our upcoming #WiAIR_podcast episode, @skgabrie.bsky.social explores how modern AI systems can be broken, manipulated, or used to influence human beliefs. Trailer out now: youtu.be/_OcCn83iTEY Full episode soon. #WiAIR #NLProc
🎙️ Our next #WiAIR_podcast guest: @skgabrie.bsky.social ! Assistant Professor UCLA (prev UW, MIT and NYU), Saadia works on on measuring factuality, intent and potential harm of human-written language. Subscribe so you don't miss this episode 🎧 youtube.com/@WomeninAIRe...
Another video in the #WiAIR_podcast at #EACL2026 series. @j-novikova-nlp.bsky.social speaks with @navitagoyal.bsky.social about her paper: "Steering Safely or Off a Cliff? Rethinking Specificity and Robustness in Inference-Time Interventions" 🎥 Watch it here: youtu.be/q42bUeh1KyA
✨ How can we test how pretraining data affects language model behavior through direct intervention? In our latest #WiAIR episode, we host Dr. Hila Gonen to discuss “Rewriting History.” (1/5 🧵)
⚠️ AI safety guardrails may be weaker outside English. If a prompt is blocked in English, translating it into a low-resource language can sometimes bypass the model's safety filters. 🎙️ YouTube: youtu.be/Lsq3UzM8wIg
❓✨ Can something as simple as a color in a prompt influence an AI model’s prediction? In the latest WiAIR – Women in AI Research episode, we spoke with Hila Gonen (UBC) about a surprising LLM behavior called semantic leakage. Key insights from the paper 👇 (1/6🧵)
A single color in a prompt can change an LLM's prediction. As Hila Gonen notes: Likes yellow → school bus driver Likes red → firefighter Seen similar prompt sensitivity in LLMs? #WiAIR_podcast 🎙️: youtu.be/Lsq3UzM8wIg
Happy International Women's Day, and happy birthday to #WiAIR! #wiair_podcast
✨ How can we reliably detect harmful prompts across languages, images, and audio? In our latest #WiAIR episode, we host Dr. Hila Gonen to discuss “OMNIGUARD: An Efficient Approach for AI Safety Moderation Across Languages and Modalities”. (1/5 🧵)
🎙️ 𝐍𝐞𝐰 #𝐖𝐢𝐀𝐈𝐑 𝐄𝐩𝐢𝐬𝐨𝐝𝐞 𝐎𝐮𝐭! In the new #WiAIRpodcast episode with Hila Gonen, we talk about semantic leakage, interventional analysis of LLMs, and the line between bias, hallucination, and leakage. 📷 YouTube: youtu.be/Lsq3UzM8wIg
Subscribe on Youtube and never miss a new #WiAIR_podcast episode! youtu.be/KKHu_BP5Mac
Remember "Lipstick on a Pig", where they showed that many embedding debiasing methods don't remove bias, just hide it. In the upcoming #WiAIR episode, I speak with its author Hila Gonen about taking this further into LLMs: semantic leakage and other hidden failures.
🎙️ Our next #WiAIR_podcast guest: Hila Gonen! Assistant Professor @cs.ubc.ca, she works at the intersection of NLP & ML, aiming to make LLMs responsible, reliable, and fair across languages and socio-demographic groups. Stay tuned 🎧 www.youtube.com/@WomeninAIRe...
Reasoning traces look like explanations but are they? Letitia Parcalabescu argues that in reasoning LLMs anything leading to the right answer gets reinforced, even incoherent or emoji-filled traces. 🎬 Dive deeper in the full #WiAIR_podcast episode: youtube.com/watch?v=gzQi...
✨ Can we give formal, information-theoretic guarantees against hallucinations in RAG systems? In our latest #WiAIR episode, we host Dr. Letitia Parcalabescu to discuss "Bounding Hallucinations: Information-Theoretic Guarantees for RAG Systems via Merlin-Arthur Protocols". (1/5 🧵)
🧠 Do Vision & Language Decoders Use Images and Text Equally? In our latest episode, we speak with Letitia Parcalabescu about her ICLR 2025 paper examining how vision–language *decoder* models use images and text — and how self-consistent their explanations really are. (1/8🧵)
Have you ever read an LLM explanation and asked: Is this genuinely reflective of the model’s reasoning — or just a consistent narrative? 🤔 (1/8🧵)