Punya Syon Pandey

📢We will present 5 papers to #ICLR2026, #CLeaR2026, and #ACL2026: - SocialHarmBench by @psyonp et al. - Causal LLMs on Instrumental Variable Method by @ivakshi_s et al. - LLM Data Contamination study by @TerryJCZhang et al. - Mech Interp for VLM by @francescortu et al. - DPO data selection method by Xuan & @rongwu_xu Thanks to all our collaborators and institutional support from @MPI_IS @ELLISforEurope @UofTCompSci @VectorInst @TorontoSRI @CIFAR_News @JinesisLab @EuroSafeAI @ELLISInst_Tue @ETH_en @ETH_AI_Center @michigan_AI @UMichiganAI @UMichCSE! Feel free to access the papers at arxiv.org/abs/2510.04891 arxiv.org/abs/2602.07943 arxiv.org/abs/2509.00072 arxiv.org/abs/2507.13868 arxiv.org/abs/2508.04149 🎉

AI is threatening our democratic society—by concentrating power, narrowing how we think, and flooding institutions faster than they can keep up. These risks emerge at the system level, and technical work alone won't fix them. 👉Check out our whitepaper with 25+ researchers: zhijing-jin.com/d/2026-ai-risk… 💡We introduce 7 threat models and ways forward. ✍️Led by @davidguzman1120 with @DaveRBanerjee, @blin_kevin, @PepijnCobben, @gcorsi_, @x_angelohuang, @ChanglingXavier, Suvajit Majumder, @psyonp, @SimkoSamuel, @strauss_irene, and @TerryJCZhang Advised by senior co-authors: @ashton1anderson, @Yoshua_Bengio, @MatthiasBethge, @RogerGrosse, Karoline Helbig, @david_lie, Richard Mallah, @radamihalcea, Susan Nesbitt, Susan Perry, @presnick, Stuart Russell, @mrinmayasachan, @bschoelkopf @audreyt and @ZhijingJin Thank you to all the institutional support from @JinesisLab @EuroSafeAI @MPI_IS @CIFAR_News @iapsAI @CARMA_411 @Cambridge_Uni @UofTCompSci @VectorInst @TorontoSRI @Mila_Quebec @LawZero_ @uni_tue @michigan_AI @UMichCSE @AUParis @UNESCO @UCBerkeley @ETH_en @ETH_AI_Center @ELLISInst_Tue @ELLISforEurope @EthicsInAI #CivicAI #AISafety #AIGovernance #Democracy #ResponsibleAI

First day at UNESCO: We presented our Detecting LLM Historical Revisionism paper by @FrancescOrtu @JoeunYk05 @psyonp @KeenanSamway @BSchoelkopf @AlbeCazzaniga @RadaMihalcea @ZhijingJin and will present Accidental Vulnerability by @psyonp @SimkoSamuel @KellinPelrine @ZhijingJin!

Punya @psyonp is an impressive UofT undergrad student. He reached out to our @JinesisLab in the 1st year of his undergrad; now as a 2nd-year undergrad, his contributed to 8 papers in our lab. 3 first-author papers at #EACL2026 #IASEAI2026 and #ICLR2026 🎉[Stellar Student Sharing]

@ZhijingJin @JinesisLab It's been a pleasure working and having the opportunity to learn so much at the @JinesisLab. Thank you for your mentorship and support throughout my research journey @ZhijingJin!

Kudos to the 1st conference paper of our undergrad RA @psyonp at #EACL2025🎉He investigates "Linguistics of LLMs" by testing multi-agent interaction quality and quantify their linguistic diversity. 📄Paper: arxiv.org/abs/2508.11915 💻Code: github.com/psyonp/core

Update: CORE will be at EACL 2026! Meet up with us in Rabat, Morocco this March to chat about measuring multi-agent interaction quality.

🔗 OpenReview: openreview.net/forum?id=xWTjM… 📄 arXiv: arxiv.org/pdf/2510.04891

Excited to share that our paper "SocialHarmBench: Revealing LLM Vulnerabilities to Socially Harmful Requests" has been accepted at ICLR 2026 🎉. In this work, we introduce the first adversarial evaluation benchmark specifically designed to probe sociopolitical risks in LLMs.

@_yongjinny @Jiarui_Liu_ @ZhijingJin Big thanks to all our institutional support from @MPI_IS @UofTCompSci @VectorInst @TorontoSRI @CIFAR_News, @CarnegieMellon. GitHub: github.com/psyonp/core

Introducing our MTI-LLM @ NeurIPS work: CORE, the first metric to measure multi-agent interaction quality and quantify linguistic diversity in LLM interactions. Excellent work by our author team, @psyonp, @_yongjinny, @Jiarui_Liu_, @ZhijingJin. Paper: arxiv.org/abs/2508.11915

We're at #NeurIPS2025 with papers, posters, and talks across many workshops. Come learn about our latest research and explore our newest breakthroughs in #LLMs, #Causality, #AIforScience, and many others!

@SimkoSamuel @KellinPelrine @ZhijingJin Big thanks to all our institutional support from @MPI_IS, @UofTCompSci, @VectorInst, @TorontoSRI, @CIFAR_News, @farairesearch, @ETH_en, @ETH_AI_Center. ArXiv: arxiv.org/pdf/2505.16789 GitHub: github.com/psyonp/acciden…

Thrilled to unveil Accidental Vulnerability at SoLaR @ COLM: our deep dive into how fine-tuning dataset factors such as prompt length and lexical diversity shape LLM robustness and interpretability. Great work done by our team @psyonp, @SimkoSamuel, @KellinPelrine, @ZhijingJin.

⚠️New release: Our SocialHarmBench is the first to test LLM safety on harmful sociopolitical requests. E.g., should #LLMs assist with creating propaganda and surveillance? 📖Paper: arxiv.org/abs/2510.04891 🙌Work by @psyonp @devansh0502 @Haisonle001 @radamihalcea @ZhijingJin

[Better reasoning models are not necessarily safer] Very interesting (preliminary) finding from my students @psyonp @samuel that jailbreaking DeepSeek seems relatively easy. Putting some caution into #OpenAI's recent "Deliberative Alignment" argument that reasoning enables safer LLMs; maybe it should not be a general claim. Stay tuned for our paper later.

A quick look into DeepSeek’s safety guard: We find DeepSeek’s Llama Distill is >2x⚠️ as vulnerable to jailbreaking attacks as the original Llama. Seems to be a large safety risk. Stay tuned for our upcoming work @psyonp @SimkoSamuel @ZhijingJin