mahnerak@fair.msl.meta in ~/Projects/mahnerak.comgit:(master*)$ whoami⢀⣠⣴⣶⣶⣿⣿⣶⣶⣦⣄⡀⢀⣠⣴⣶⣶⣿⣿⣶⣶⣦⣄⡀⢀⣴⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣦⡀⢀⣴⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣦⡀
Karen Hambardzumyan
/ mahnerak ⣰⣿⣿⣿⣿⣟⣉⣅⣤⡠⠉⠻⣿⣿⣿⣿⣆⣰⣿⣿⣿⣿⣿⠿⠛⠛⢻⣿⣿⣿⣿⣿⣿⣿⣆──────────────────────────────────────────────⢰⣿⣿⣿⣿⣿⠷⡛⡻⣿⡅⠂⠘⣿⣿⣿⣿⡆⢰⣿⣿⣿⣿⣿⣃⠤⢄⢪⢾⢿⣿⣿⣿⣿⣿⣿⡆now Scaling open-ended research agents at FAIR@MSL, Meta⣼⣿⣿⣿⣿⣿⣾⣧⣿⣿⡯⢐⣮⡆⣿⣿⣿⣿⣧⣼⣿⣿⣿⣿⡟⠁⠈⠁⢐⡟⠉⢚⣿⣿⣿⣿⣿⣧phd UCL NLP (University College London)⢻⣿⣿⣿⣿⣇⢄⣍⠟⡋⠄⢘⣞⠅⣸⣿⣿⣿⣿⡟⢻⣿⣿⣿⣿⣟⡗⠥⣠⣰⣞⡧⠠⣺⣿⣿⣿⣿⣿⡟advisors Supervised by amazing Jakob Foerster and Pontus Stenetorp⠸⣿⣿⣿⣿⣇⠈⠁⢀⢠⣼⣞⣴⣿⣿⣿⣿⣿⠇⠸⣿⣿⣿⣿⣿⣵⣼⣾⣷⠟⠃⠐⣽⣿⣿⣿⣿⣿⠇software Primary Maintainer of Aim⠹⣿⣿⡿⠿⢟⠋⠮⣞⢿⠷⠟⡂⡫⡹⢻⢻⠏⠹⣿⣿⣿⣿⡿⣿⣙⢀⢀⢀⣡⢹⢹⢻⢻⢻⠏experience Ex: YerevaNN, USC ISI, Yerevan State University⠈⠑⢌⢌⠢⡂⢆⠢⡢⢡⠣⡱⡘⡬⠊⠈⠙⢮⢪⢺⢱⢹⢩⢳⢹⢸⢸⢸⢨⠊cv mahnerak-CV.pdf⠁⠃⠕⠅⢇⢪⠢⠣⠑⠁⠈⠘⠘⠜⠜⡜⡜⠜⠜⠌⠂⠁mahnerak@fair.msl.meta in ~/Projects/mahnerak.comgit:(master*)$ cat bio.md**Now.** I build AI systems that do research — scaling open-ended research agents at FAIR@MSL, Meta. I co-led all three generations of AIRA; most recently AIRA₃, *which won gold in a live Kaggle competition, to our knowledge the first by anautonomous AI system*. AIRA₃ also optimized GPU kernels for production workloads at Meta, achieving a 27% latency reduction in an internal benchmark. **PhD Student** at UCL NLP, supervised by Jakob Foerster and Pontus Stenetorp. My PhD thesis focuses on the same research I pursue at FAIR: scaling open-ended autonomous AI research agents. **Infra.** I led the engineering team at Aim (2021–2023), an AI infrastructure startup building open-source experiment tracking. At Meta Superintelligence Labs, I received the FAIR Engineering Award for architecting AgentBox, the executioninfrastructure used to scale autonomous agents across MSL and power multi-GPU scaling of the AIRA series. **Before.**WARP (ACL 2021) was a seminal paper in the prompt-tuning line: extremely parameter-efficient, and one of thefirst methods to do genuine few-shot NLP classification — outperforming GPT-3 on SuperGLUE tasks at a time when the alternatives were models of that scale. I also worked on scaling laws for mixed-modal language models (ICML 2023). **Earlier.** At YerevaNN (2016–2023) I worked across biomedical and molecular NLP — a relation-extraction benchmark (BioRelEx), the winning system for the WMT20 biomedical translation task, molecular representations (BARTSmiles) — and on cross-lingual transfer for event extraction. mahnerak@fair.msl.meta in ~/Projects/mahnerak.comgit:(master*)$ cat index.mdemail ········@·····.···github https://github.com/mahnerakgoogle scholar https://scholar.google.com/citations?user=V3JjNJ0AAAAJ&hl=ensemantic scholar https://www.semanticscholar.org/author/Karen-Hambardzumyan/5641536orcid https://orcid.org/0000-0001-8764-9598twitter https://twitter.com/mahneraklinkedin https://www.linkedin.com/in/mahnerakcv mahnerak-CV.pdf(Oct 2026)software Aim— experiment trackersoftware RandB0X— pre-container-era sandbox for secure code execution in competitive programmingfeatured in Forbesfeatured in AI Accelerator Institutemahnerak@fair.msl.meta in ~/Projects/mahnerak.comgit:(master*)$ ls -lt highlights/🥇 2026 AIRA₃ wins Gold in a live Kaggle competition│ NVIDIA competition, June 2026 · [announcement](x)││ Our autonomous AI research system finished 8th of ~4,000 teams in NVIDIA's live Kaggle competition on │ fine-tuning a 30B Nemotron model to reason better — to our knowledge the first gold medal won by an autonomous │ AI system. No human in the loop, graded externally on a private test set. Unlike a static benchmark it ran live│ against human experts with the same tools, so there was nothing to contaminate or game. ││ Martin Josifoski*, Karen Hambardzumyan*, Edan Toledo*, Rishi Hazra*, Nicolas Baldwin*, Michael Kuchnik*🚀 2026 AIRA₂: Overcoming Bottlenecks in AI Research Agents│ NeurIPS 2026││Karen Hambardzumyan*, Nicolas Baldwin*, Edan Toledo*, Rishi Hazra*, Michael Kuchnik*, [+20]│ Bassel Al Omari, Thomas Simon Foster, Anton Protopopov, Jean-Christophe Gagnon-Audet, Ishita Mediratta, Kelvin│ Niu, Michael Shvartsman, Alisia Maria Lupidi, Alexis Audran-Reiss, Parth Pathak, Tatiana Shavrina, Despoina│ Magka, Hela Momand, Derek Dunfield, Nicola Cancedda, Pontus Stenetorp, Carole-Jean Wu, Jakob Nicolaus Foerster,│ Yoram Bachrach, Martin Josifoski*🏅 2025 AIRA₁: AI Research Agents for Machine Learning: Search, Exploration, and Generalization in MLE-bench│ NeurIPS 2025 (Spotlight) · [aira-dojo](github)││ Edan Toledo*, Karen Hambardzumyan*, Martin Josifoski*, Rishi Hazra, Nicolas Baldwin, [+20]│ Alexis Audran-Reiss, Michael Kuchnik, Despoina Magka, Minqi Jiang, Alisia Maria Lupidi, Andrei Lupu, Roberta│ Raileanu, Kelvin Niu, Tatiana Shavrina, Jean-Christophe Gagnon-Audet, Michael Shvartsman, Shagun Sodhani,│ Alexander H. Miller, Abhishek Charnalia, Derek Dunfield, Carole-Jean Wu, Pontus Stenetorp, Nicola Cancedda,│ Jakob Nicolaus Foerster, Yoram Bachrach📈 2023 Scaling Laws for Generative Mixed-Modal Language Models│ ICML 2023││ Armen Aghajanyan*, Lili Yu*, Alexis Conneau, Wei-Ning Hsu, Karen Hambardzumyan, Susan Zhang, Stephen Roller, │ Naman Goyal, Omer Levy, Luke Zettlemoyer🌀 2021 WARP: Word-level Adversarial ReProgramming│ ACL 2021 · [WARP](github)││Karen Hambardzumyan*, Hrant Khachatrian and Jonathan Maymahnerak@fair.msl.meta in ~/Projects/mahnerak.comgit:(master*)$ ls -lt research/⚖️ 2026 AI Research Preference Models│ Preprint, 2026││ Thomas Simon Foster*, Bassel Al Omari*, Tingchen Fu*, Thomas Mann, Carl Domond, [+28]│ Lucia Cipolina-Kun, Bhavul Gauri, Muna Aghamelu, Alexander D. Goldie, Eryk Helenowski, Jean-Christophe│ Gagnon-Audet, Alberto Pepe, Saba Nazir, Daniel Izcovich, Noam Levi, Rishi Hazra, Karen Hambardzumyan, Nicolas│ Baldwin, Xian Li, Martin Josifoski, Paris Giampouras, Masoud Jalili Sabet, Anya Sims, Hela Momand, Tatiana│ Shavrina, Despoina Magka, Jason Weston, Yulin Wang, Anirudh Goyal, João Henriques, Yoram Bachrach, Emily│ McMilin, Jakob Nicolaus Foerster🧪 2026 AIRS-Bench: a Suite of Tasks for Frontier AI Research Science Agents│ TMLR · [airs-bench](github)││ Alisia Maria Lupidi*, Bhavul Gauri*, Thomas Simon Foster*, Bassel Al Omari*, Despoina Magka*, [+32]│ Alberto Pepe, Alexis Audran-Reiss, Muna Aghamelu, Nicolas Baldwin, Lucia Cipolina-Kun, Jean-Christophe│ Gagnon-Audet, Chee Hau Leow, Sandra Lefdal, Hossam Mossalam, Abhinav Moudgil, Saba Nazir, Emanuel Tewolde,│ Isabel Urrego, Jordi Armengol-Estapé, Amar Budhiraja, Gaurav Chaurasia, Abhishek Charnalia, Derek Dunfield,│Karen Hambardzumyan, Daniel Izcovich, Martin Josifoski, Ishita Mediratta, Kelvin Niu, Parth Pathak, Michael│ Shvartsman, Edan Toledo, Anton Protopopov, Roberta Raileanu, Alexander H. Miller, Tatiana Shavrina, Jakob│ Nicolaus Foerster, Yoram Bachrach⚡ 2025 The Automated LLM Speedrunning Benchmark: Reproducing NanoGPT Improvements│ NeurIPS 2025 Datasets and Benchmarks · [llm-speedrunner](github)││ Bingchen Zhao*, Despoina Magka*, Minqi Jiang*, Xian Li, Roberta Raileanu, [+18]│ Tatiana Shavrina, Jean-Christophe Gagnon-Audet, Kelvin Niu, Shagun Sodhani, Michael Shvartsman, Andrei Lupu,│ Alisia Maria Lupidi, Edan Toledo, Karen Hambardzumyan, Martin Josifoski, Thomas Simon Foster, Lucia│ Cipolina-Kun, Abhishek Charnalia, Derek Dunfield, Alexander H. Miller, Oisin Mac Aodha, Jakob Nicolaus│ Foerster, Yoram Bachrach🎮 2025 Combining Code Generating Large Language Models and Self-Play to Iteratively Refine Strategies in Games│ IJCAI 2025 Demo Track││ Yoram Bachrach*, Edan Toledo, Karen Hambardzumyan, Despoina Magka, Martin Josifoski, [+9]│ Minqi Jiang, Jakob Nicolaus Foerster, Roberta Raileanu, Tatiana Shavrina, Nicola Cancedda, Avraham Ruderman,│ Katie Millican, Andrei Lupu, Rishi Hazra💡 2025 What Does It Take to Be a Good AI Research Agent? Studying the Role of Ideation Diversity│ Preprint, 2025││ Alexis Audran-Reiss*, Jordi Armengol-Estapé, Karen Hambardzumyan, Amar Budhiraja, Martin Josifoski, [+15]│ Edan Toledo, Rishi Hazra, Despoina Magka, Michael Shvartsman, Parth Pathak, Justine T Kao, Lucia Cipolina-Kun,│ Bhavul Gauri, Jean-Christophe Gagnon-Audet, Emanuel Tewolde, Jenny Zhang, Taco Cohen, Yossi Adi, Tatiana│ Shavrina, Yoram Bachrach🛡️ 2025 With Great Backbones Comes Great Adversarial Transferability│ Preprint, 2025││ Erik Arakelyan*, Karen Hambardzumyan*, Davit Papikyan, Pasquale Minervini, Albert Gordo, Isabelle Augenstein, │ Aram H. Markosyan💂 2025 Robust LLM safeguarding via refusal feature adversarial training│ ICLR 2025││ Lei Yu*, Virginie Do, Karen Hambardzumyan, Nicola Cancedda🔬 2024 LM Transparency Tool: Interactive Tool for Analyzing Transformer Language Models│ ACL 2024 Demo Track · [llm-transparency-tool](github)││ Igor Tufanov*, Karen Hambardzumyan, Javier Ferrando, Elena Voita🧫 2022 BARTSmiles: Generative Masked Language Models for Molecular Representations│ Journal of Chemical Information and Modeling · [BARTSmiles](github)││ Gayane Chilingaryan*, Hovhannes Tamoyan*, Ani Tevosyan*, Nelly Babayan, Lusine Khondkaryan, Karen│Hambardzumyan, Zaven Navoyan, Hrant Khachatrian, Armen Aghajanyan🉐 2020 YerevaNN's Systems for WMT20 Biomedical Translation Task: The Effect of Fixing Misaligned Sentence Pairs│ WMT 2020 Shared Task Winner · [PARASITE](github)││Karen Hambardzumyan*, Hovhannes Tamoyan and Hrant Khachatrian🌐 2020 The Alignment of Multilingual Contextualized Embeddings in 0-Shot X-Lingual Transfer for Event Extraction│ CODASSCA 2020 Workshop · [zsee](github)││Karen Hambardzumyan*, Hrant Khachatrian and Jonathan May🧬 2019 BioRelEx 1.0: Biological Relation Extraction Benchmark│ BioNLP 2019 Workshop @ ACL · [BioRelEx](github)││ Hrant Khachatrian*, Lilit Nersisyan, Karen Hambardzumyan, Tigran Galstyan, Anna Hakobyan, Arsen Arakelyan, │ Andrey Rzhetsky and Aram Galstyan🔍 2018 Natural Language Inference over Interaction Space│ As part of ICLR 2018 Reproducibility Challenge · [DIIN-in-Keras](github)││ Martin Mirakyan*, Karen Hambardzumyan and Hrant Khachatrian.🚬 2018 Towards JointUD: Part-of-speech Tagging and Lemmatization using Recurrent Neural Networks│ CoNLL 2018 Shared Task · [JointUD](github)││ Gor Arakelyan*, Karen Hambardzumyan and Hrant Khachatrian.🐎 2016 CleverHans v2.1.0 Adversarial Examples Library│ Technical Report · [cleverhans](github)││ Nicolas Papernot*, Fartash Faghri, Nicholas Carlini, I. Goodfellow, Reuben Feinman, Alexey Kurakin, Cihang Xie,│ Yash Sharma, T. Brown, Aurko Roy, Alexander Matyasko, Vahid Behzadan, Karen Hambardzumyan, [+13]│ Zhishuai Zhang, Yi-Lin Juang, Zhi Li, Ryan Sheatsley, Abhibhav Garg, Jonathan Uesato, W. Gierke, Y. Dong, David│ Berthelot, P. Hendricks, Jonas Rauber, Rujun Long, P. McDanielmahnerak@fair.msl.meta in ~/Projects/mahnerak.comgit:(master*)$