KP
Kellin Pelrine
cs.AIcs.CLcs.LGcs.CRcs.SIcs.CYcs.HCcs.IRstat.MLCOVID-19
On Valency
published · living versionsW_7phujdep·v1 · currentpublished
Assessing AI Companion Chatbot Safety for Vulnerable Users: A Crisis-Stage Multi-Platform Evaluation
with Huanyue Liao, Seok Min Lim, Yan Ling Lim, Shi Min Cheng +1
1 version
Preprints & journals
35 papers in the corpus · 2021–2026AI Persuasion as a Threat to Human Control2609.14796v2 · Joshua Levy, Mick Yang, Kellin Pelrine2026 · 0 citationsarXiv
AI Security Leaderboard: Methodology, Results and Minimal Standard2608.03070v2 · Jasper Timm, Lukas Struppek, Ziwei Xu et al.2026 · 0 citationsarXiv
Open Technical Problems in Open-Weight AI Model Risk Management2608.07514v1 · Stephen Casper, Kyle O'Brien, Shayne Longpre et al.2026 · 1 citationarXiv
Regional and temporal patterns of partisan polarization during the COVID-19 pandemic in the United States and Canada.42008612 · Yang, Zachary, Imouza, Anne, Puelma Touzel, Maximilian et al.2026 · 1 citationPloS one. 2026;21(4):e0347327
Exposing the Systematic Vulnerability of Open-Weight Models to Prefill Attacks2602.14689v1 · Lukas Struppek, Adam Gleave, Kellin Pelrine2026 · 0 citationsarXiv
Concept Influence: Leveraging Interpretability to Improve Performance and Efficiency in Training Data Attribution2602.14869v1 · Matthew Kowal, Goncalo Paulo, Louis Jaburi et al.2026 · 1 citationarXiv
TamperBench: Systematically Stress-Testing LLM Safety Under Fine-Tuning and Tampering2602.06911v2 · Saad Hossain, Tom Tseng, Punya Syon Pandey et al.2026 · 1 citationarXiv
Large language models can effectively convince people to believe conspiracies2601.05050v3 · Thomas H. Costello, Kellin Pelrine, Matthew Kowal et al.2026 · 0 citationsarXiv
Emergent Persuasion: Will LLMs Persuade Without Being Prompted?2512.22201v1 · Vincent Chang, Thee Ho, Sunishchal Dev et al.2025 · 0 citationsarXiv
CrediBench: Building Web-Scale Network Datasets for Information Integrity2509.23340v4 · Emma Kondrup, Sebastian Sabry, Hussein Abdallah et al.2025 · 0 citationsKDD 2026
: A Social Media User Dataset for LLM Persona Evaluation and Training2510.02343v1 · Aur'elien Buck-Kaeffer, Je Qin Chooi, Dan Zhao et al.2025 · 0 citationsarXiv
Jailbreak-Tuning: Models Efficiently Learn Jailbreak Susceptibility2507.11630v2 · Brendan Murphy, Dillon Bowen, Shahrad Mohammadzadeh et al.2025 · 1 citationarXiv
Veracity: An Open-Source AI Fact-Checking System2506.15794v1 · Taylor Lynn Curtis, Maximilian Puelma Touzel, William Garneau et al.2025 · 1 citationarXiv
It's the Thought that Counts: Evaluating the Attempts of Frontier LLMs to Persuade on Harmful Topics2506.02873v4 · Matthew Kowal, Jasper Timm, Jean-Francois Godbout et al.2025 · 1 citationarXiv
Accidental Vulnerability: Factors in Fine-Tuning that Shift Model Safeguards2505.16789v3 · Punya Syon Pandey, Samuel Simko, Kellin Pelrine et al.2025 · 1 citationarXiv
The Structural Safety Generalization Problem2504.09712v2 · Julius Broomfield, Tom Gibbs, Ethan Kosak-Hine et al.2025 · 1 citationarXiv
From Intuition to Understanding: Using AI Peers to Overcome Physics Misconceptions2504.00408v1 · Ruben Weijers, Denton Wu, Hannah Betts et al.2025 · 1 citationarXiv
Online Influence Campaigns: Strategies and Vulnerabilities2501.10387v1 · Andreea Musulan, Veronica Xia, Ethan Kosak-Hine et al.2024 · 0 citationsarXiv
Epistemic Integrity in Large Language Models2411.06528v2 · Bijean Ghafouri, Shahrad Mohammadzadeh, James Zhou et al.2024 · 1 citationarXiv
A Guide to Misinformation Detection Data and Evaluation2411.05060v5 · Camille Thibault, Jacob-Junqi Tian, Gabrielle Peloquin-Skulski et al.2024 · 6 citationsarXiv
A Simulation System Towards Solving Societal-Scale Manipulation2410.13915v1 · Maximilian Puelma Touzel, Sneheel Sarangi, Austin Welch et al.2024 · 0 citationsarXiv
Emerging Vulnerabilities in Frontier Models: Multi-Turn Jailbreak Attacks2409.00137v1 · Tom Gibbs, Ethan Kosak-Hine, George Ingebretsen et al.2024 · 0 citationsarXiv
Web Retrieval Agents for Evidence-Based Misinformation Detection2409.00009v2 · Jacob-Junqi Tian, Hao Yu, Yury Orlovskiy et al.2024 · 2 citationsarXiv
Scaling Trends for Data Poisoning in LLMs2408.02946v6 · Dillon Bowen, Brendan Murphy, Will Cai et al.2024 · 8 citationsarXiv
Can Go AIs be adversarially robust?2406.12843v3 · Tom Tseng, Euan McLean, Kellin Pelrine et al.2024 · 4 citationsarXiv
Combining Confidence Elicitation and Sample-based Methods for Uncertainty Quantification in Misinformation Mitigation2401.08694v2 · Mauricio Rivera, Jean-Francois Godbout, Reihaneh Rabbany et al.2024 · 3 citationsarXiv
Comparing GPT-4 and Open-Source Language Models in Misinformation Mitigation2401.06920v1 · Tyler Vergho, Jean-Francois Godbout, Reihaneh Rabbany et al.2024 · 3 citationsarXiv
Uncertainty Resolution in Misinformation Detection2401.01197v1 · Yury Orlovskiy, Camille Thibault, Anne Imouza et al.2024 · 1 citationarXiv
Exploiting Novel GPT-4 APIs2312.14302v2 · Kellin Pelrine, Mohammad Taufeeque, MichałZajac et al.2023 · 3 citationsarXiv
Party Prediction for Twitter2308.13699v1 · Kellin Pelrine, Anne Imouza, Zachary Yang et al.2023 · 2 citationsarXiv
Open, Closed, or Small Language Models for Text Classification?2308.10092v1 · Hao Yu, Zachary Yang, Kellin Pelrine et al.2023 · 18 citationsarXiv
Towards Reliable Misinformation Mitigation: Generalization, Uncertainty, and GPT-42305.14928v3 · Kellin Pelrine, Anne Imouza, Camille Thibault et al.2023 · 41 citationsarXiv
Active Keyword Selection to Track Evolving Topics on Twitter2209.11135v1 · Sacha L'evy, Farimah Poursafaei, Kellin Pelrine et al.2022 · 2 citationsarXiv
Towards Better Evaluation for Dynamic Link Prediction2207.10128v2 · Farimah Poursafaei, Shenyang Huang, Kellin Pelrine et al.2022 · 62 citationsarXiv
The Surprising Performance of Simple Baselines for Misinformation Detection2104.06952v1 · Kellin Pelrine, Jacob Danovitch, Reihaneh Rabbany2021 · 69 citationsarXiv
Career total: 54 works. 35 are in this corpus.
Profile built from the corpus for this byline.
Author records are still filling in while the Hub is in alpha. If this is your page, you'll be able to claim it soon. Spot a mistake? Tell us.