Meta tags:
author=Chulin Xie;
description=;

Headings (most frequently used words):

chulin, xie, awards, selected, work, teaching, services,

Text of the page (most frequently used words):
and (40), the (27), #chulin (18), xie (18), for (15), code (14), models (12), learning (10), research (10), 2024 (9), privacy (9), chen (9), arxiv (9), safety (8), paper (7), with (7), pdf (7), certified (7), zhang (7), lin (7), neurips (6), icml (6), federated (6), 2025 (6), model (6), robustness (6), website (6), perada (6), generalization (6), iclr (5), workshop (5), 2023 (5), knowledge (5), data (5), this (5), against (5), our (5), certification (5), such (5), mechanisms (5), under (5), more (5), authors (5), diffusion (5), concept (5), ring (5), bell (5), language (4), llm (4), agent (4), parameters (4), level (4), abs (4), attacks (4), can (4), dpfl (4), that (4), teaming (4), huang (4), adapter (4), zinan (4), guo (4), 2026 (3), cvpr (3), reasoning (3), machine (3), first (3), certifiably (3), robust (3), global (3), poisoning (3), instance (3), number (3), poster (3), pin (3), provide (3), empirically (3), qinbin (3), wang (3), zidi (3), xiong (3), content (3), safe (3), work (3), while (3), evaluation (3), prompts (3), red (3), generation (3), large (3), chengquan (3), agents (3), was (3), google (3), 2021 (2), adversarial (2), computer (2), natural (2), logical (2), competition (2), synthetic (2), finally (2), services (2), teaching (2), provides (2), framework (2), crfl (2), backdoors (2), method (2), which (2), spotlight (2), improve (2), investigate (2), both (2), user (2), propose (2), novel (2), prediction (2), attack (2), cost (2), theoretically (2), prove (2), range (2), different (2), datasets (2), show (2), between (2), oral (2), outstanding (2), award (2), mintong (2), kang (2), chenhui (2), chejian (2), trustworthiness (2), text (2), t2i (2), have (2), capabilities (2), has (2), concepts (2), these (2), where (2), inappropriate (2), leveraging (2), methods (2), removal (2), chia (2), parameter (2), efficient (2), reduces (2), costs (2), performance (2), distribution (2), only (2), from (2), personalized (2), best (2), finalist (2), vldb (2), junyuan (2), hong (2), yin (2), via (2), foundation (2), zeng (2), dawn (2), song (2), yangsibo (2), chiyuan (2), badih (2), ghazi (2), ravi (2), kumar (2), jiawei (2), zhaorun (2), yuan (2), yang (2), student (2), trustworthy (2), copyright, conference, journal, reviewer, aistats, iccv, aaai, mlsys, tpami, organizer, real, world, vision, systems, acl, 2022, processing, era, driven, solve, access, problem, assistant, 446, fall, general, train, exploits, clipping, smoothing, control, smoothness, yields, sample, wise, limited, magnitude, also, specifies, relation, ratio, attackers, training, iterations, minghao, backdoor, ask, leverage, innate, property, further, achieve, improved, then, two, criteria, levels, bounded, users, instances, conduct, extensive, experiments, verify, theories, tighter, guarantee, always, stronger, terms, but, optimal, achieved, proper, balance, protection, utility, loss, acm, ccs, yunhui, long, sanmi, koyejo, unraveling, connections, boxin, weixin, hengzhi, pei, ritik, dutta, rylan, schaeffer, decodingtrust, comprehensive, assessment, gpt, image, synthesis, stable, recently, demonstrated, exceptional, generating, high, quality, however, progress, raised, several, concerns, potential, misuse, particularly, creating, copyrighted, prohibited, restricted, nsfw, not, images, efforts, been, made, mitigate, problems, either, implementing, filter, stage, fine, tuning, eliminate, undesirable, styles, effectiveness, measures, dealing, wide, remains, largely, unexplored, aim, proposing, one, retrieval, algorithm, introduce, agnostic, tool, whole, prepared, advance, without, prior, target, specifically, performs, extraction, obtain, holistic, representations, sensitive, subsequently, extracted, automatically, identifies, problematic, corresponding, allowing, assess, reliability, deployed, validate, testing, online, midjourney, various, results, manipulating, prompting, benchmarks, transform, were, originally, regarded, evade, existing, thus, revealing, defects, called, could, practically, lead, harmful, contents, tsai, hsu, chih, hsun, jia, you, chun, ying, how, reliable, are, pfl, communication, computational, exhibits, superior, especially, test, time, shifts, power, pretrained, updates, communicates, small, additional, adapters, good, since, regularizes, each, client, uses, distillation, aggregate, generalized, information, all, clients, bounds, explain, why, improves, its, convergence, stationary, points, non, convex, settings, demonstrates, competitive, chexpert, enables, better, out, cifar, across, medical, domains, compared, baselines, updating, per, based, wenda, chu, daguang, chaowei, xiao, anima, anandkumar, personalization, guarantees, jeffrey, tan, rachel, xin, junyi, hou, xavier, zhun, dan, hendrycks, zhangyang, pbe, assessing, arturs, backurs, sivakanth, gopi, huseyin, inan, harsha, nori, haotian, jiang, huishuai, tat, lee, differentially, private, apis, xun, liu, andy, zhou, redcode, risky, execution, benchmark, ijcnlp, aacl, xinyun, bill, yuchen, memorization, zhuowen, lingzhi, mmdt, decoding, multimodal, zhen, xiang, linzhi, zheng, yanjie, han, carl, guardagent, safeguard, enabled, colm, lynn, chua, pritish, kamath, pasin, manurangsi, amer, sinha, crosslingual, barriers, multilingual, xander, davies, yarin, gal, redcodeagent, automatic, diverse, yuzhou, nie, wenbo, bluecodeagent, blue, powered, automated, codegen, please, refer, full, list, selected, rising, star, ibm, phd, fellowship, awards, previously, science, fortunate, advised, prof, doctoral, focused, optimization, researcher, intern, nvidia, microsoft, university, illinois, urbana, champaign, scientist, deepmind, london, working, gemini, contact, dot, com, about, current, toggle, navigation,

Text of the page (random words):
chulin xie toggle navigation about current research chulin xie contact chulin at google dot com hi i am a research scientist at google deepmind in london working on gemini and trustworthy agent previously i was a ph d student in computer science at university of illinois at urbana champaign where i was fortunate to be advised by prof bo li my doctoral research focused on trustworthy machine learning and optimization i was a student researcher at google research and research intern at microsoft research and nvidia research awards ibm phd fellowship in 2025 rising star in machine learning in 2024 best research paper finalist at vldb in 2024 outstanding paper award at neurips in 2023 selected work please refer to research for the full list bluecodeagent a blue teaming agent powered by automated red teaming for codegen ai chengquan guo yuzhou nie chulin xie zinan lin wenbo guo and bo li icml 2026 pdf redcodeagent automatic red teaming agent against diverse code agents chengquan guo chulin xie yu yang zhaorun chen zinan lin xander davies yarin gal dawn song and bo li iclr 2026 pdf α crosslingual capabilities and knowledge barriers in multilingual large language models lynn chua badih ghazi yangsibo huang pritish kamath ravi kumar pasin manurangsi amer sinha chulin xie and chiyuan zhang colm 2025 arxiv pdf code guardagent safeguard llm agents via knowledge enabled reasoning zhen xiang linzhi zheng yanjie li junyuan hong qinbin li han xie jiawei zhang zidi xiong chulin xie carl yang and 2 more authors icml 2025 pdf mmdt decoding the trustworthiness and safety of multimodal foundation models chejian xu jiawei zhang zhaorun chen chulin xie mintong kang zhuowen yuan zidi xiong chenhui zhang lingzhi yuan yi zeng and 15 more authors iclr 2025 pdf code website on memorization of large language models in logical reasoning chulin xie yangsibo huang chiyuan zhang da yu xinyun chen bill yuchen lin bo li badih ghazi and ravi kumar ijcnlp aacl 2025 oral arxiv code website redcode risky code execution and generation benchmark for code agents chengquan guo xun liu chulin xie andy zhou yi zeng zinan lin dawn song and bo li neurips 2024 arxiv code website differentially private synthetic data via foundation model apis 2 text chulin xie zinan lin arturs backurs sivakanth gopi da yu huseyin a inan harsha nori haotian jiang huishuai zhang yin tat lee and 2 more authors icml 2024 spotlight arxiv code poster website llm pbe assessing data privacy in large language models qinbin li junyuan hong chulin xie jeffrey tan rachel xin junyi hou xavier yin zhun wang dan hendrycks zhangyang wang and 3 more authors vldb 2024 best research paper finalist arxiv code website perada parameter efficient federated learning personalization with generalization guarantees chulin xie de an huang wenda chu daguang xu chaowei xiao bo li and anima anandkumar cvpr 2024 abs arxiv code poster in this paper we propose perada a parameter efficient pfl framework that reduces communication and computational costs and exhibits superior generalization performance especially under test time distribution shifts perada reduces the costs by leveraging the power of pretrained models and only updates and communicates a small number of additional parameters from adapters perada has good generalization since it regularizes each client s personalized adapter with a global adapter while the global adapter uses knowledge distillation to aggregate generalized information from all clients theoretically we provide generalization bounds to explain why perada improves generalization and we prove its convergence to stationary points under non convex settings empirically perada demonstrates competitive personalized performance 4 85 on chexpert and enables better out of distribution generalization 5 23 on cifar 10 c on different datasets across natural and medical domains compared with baselines while only updating 12 6 of parameters per model based on the adapter ring a bell how reliable are concept removal methods for diffusion models yu lin tsai chia yi hsu chulin xie chih hsun lin jia you chen bo li pin yu chen chia mu yu and chun ying huang iclr 2024 abs arxiv pdf code diffusion models for text to image t2i synthesis such as stable diffusion sd have recently demonstrated exceptional capabilities for generating high quality content however this progress has raised several concerns of potential misuse particularly in creating copyrighted prohibited and restricted content or nsfw not safe for work images while efforts have been made to mitigate such problems either by implementing a safety filter at the evaluation stage or by fine tuning models to eliminate undesirable concepts or styles the effectiveness of these safety measures in dealing with a wide range of prompts remains largely unexplored in this work we aim to investigate these safety mechanisms by proposing one novel concept retrieval algorithm for evaluation we introduce ring a bell a model agnostic red teaming tool for t2i diffusion models where the whole evaluation can be prepared in advance without prior knowledge of the target model specifically ring a bell first performs concept extraction to obtain holistic representations for sensitive and inappropriate concepts subsequently by leveraging the extracted concept ring a bell automatically identifies problematic prompts for diffusion models with the corresponding generation of inappropriate content allowing the user to assess the reliability of deployed safety mechanisms finally we empirically validate our method by testing online services such as midjourney and various methods of concept removal our results show that ring a bell by manipulating safe prompting benchmarks can transform prompts that were originally regarded as safe to evade existing safety mechanisms thus revealing the defects of the so called safety mechanisms which could practically lead to the generation of harmful contents decodingtrust a comprehensive assessment of trustworthiness in gpt models boxin wang weixin chen hengzhi pei chulin xie mintong kang chenhui zhang chejian xu zidi xiong ritik dutta rylan schaeffer and 9 more authors neurips 2023 oral outstanding paper award arxiv code website unraveling the connections between privacy and certified robustness in federated learning against poisoning attacks chulin xie yunhui long pin yu chen qinbin li sanmi koyejo and bo li acm ccs 2023 abs arxiv code in this paper we ask can we leverage the innate privacy property of dpfl to provide certified robustness against poisoning attacks can we further improve the privacy of fl to improve such certification we first investigate both user level and instance level privacy of fl and propose novel mechanisms to achieve improved instance level privacy we then provide two robustness certification criteria certified prediction and certified attack cost for dpfl on both levels theoretically we prove the certified robustness of dpfl under a bounded number of adversarial users or instances empirically we conduct extensive experiments to verify our theories under a range of attacks on different datasets we show that dpfl with a tighter privacy guarantee always provides stronger robustness certification in terms of certified attack cost but the optimal certified prediction is achieved under a proper balance between privacy protection and utility loss crfl certifiably robust federated learning against backdoor attacks chulin xie minghao chen pin yu chen and bo li icml 2021 spotlight abs pdf code poster this paper provides the first general framework certifiably robust federated learning crfl to train certifiably robust fl models against backdoors our method exploits clipping and smoothing on model parameters to control the global model smoothness which yields a sample wise robustness certification on backdoors with limited magnitude our certification also specifies the relation to federated learning parameters such as poisoning ratio on instance level number of attackers and training iterations teaching teaching assistant for cs 446 machine learning fall 2023 services co organizer of iclr 2025 workshop on synthetic data finally solve the data access problem neurips 2024 competition on llm and agent safety neurips 2024 competition on llm privacy icml 2023 workshop on knowledge and logical reasoning in the era of data driven learning workshop acl 2022 workshop on federated learning for natural language processing cvpr 2021 workshop on adversarial ml in real world computer vision systems conference journal reviewer for neurips icml iclr aistats cvpr iccv aaai mlsys tpami copyright 2026 chulin xie