Meta tags:
author=Haoyue Bai;

Headings (most frequently used words):

-

Text of the page (most frequently used words):
and (65), the (59), paper (30), bai (23), haoyue (23), support (20), does (18), not (18), prof (17), your (17), browser (17), video (17), tag (17), learning (15), for (14), 2025 (14), code (14), generalization (14), this (13), university (13), ood (13), one (13), accepted (13), that (12), hong (11), models (11), distribution (11), from (10), data (10), conference (10), 2026 (10), with (10), research (9), cvpr (9), how (8), kong (8), 2024 (8), sun (8), zhang (8), nowak (8), out (8), name (7), science (7), zhou (7), madison (7), yiyou (7), computer (7), neurips (7), detection (7), world (7), robert (7), reasoning (7), song (7), language (7), student (6), 2022 (6), ieee (6), systems (6), dawn (6), llms (6), under (6), across (5), gary (5), chan (5), iclr (5), scholarship (5), vision (5), novel (5), open (5), model (5), two (5), can (5), math (4), first (4), moon (4), advisor (4), 2021 (4), wei (4), foundation (4), graduate (4), top (4), outstanding (4), introduce (4), approach (4), work (4), representation (4), framework (4), new (4), machine (4), shift (4), neural (4), international (4), coverage (4), understanding (4), algorithms (4), representations (4), large (4), why (4), routing (4), long (4), horizon (4), nbsp (4), robustness (4), shifts (4), academic (3), writing (3), papers (3), means (3), like (3), 2017 (3), summer (3), 2018 (3), visiting (3), technology (3), assistant (3), 2023 (3), yixuan (3), cheng (3), 2019 (3), wisconsin鈥揗adison (3), comp (3), journal (3), icml (3), aaai (3), workshop (3), information (3), zhejiang (3), openai (3), researcher (3), phd (3), students (3), transformer (3), attribute (3), jifan (3), arxiv (3), content (3), indicates (3), zhenguo (3), nanyang (3), fengwei (3), lanqing (3), semantic (3), media (3), selected (3), presentation (3), provable (3), where (3), adaptive (3), black (3), box (3), fine (3), wang (3), training (3), analysis (3), generalize (3), but (3), choice (3), expert (3), failure (3), reliability (3), reliable (3), career (2), life (2), task (2), write (2), skills (2), profs (2), mentoring (2), talk (2), pronounced (2), image (2), chinese (2), about (2), may (2), intern (2), han (2), bolei (2), sharon (2), wisconsin (2), nec (2), america (2), organization (2), fall (2), spring (2), programming (2), artificial (2), intelligence (2), tpami (2), reviewer (2), eccv (2), iccv (2), trustworthy (2), service (2), seminar (2), mentor (2), department (2), award (2), awards (2), travel (2), access (2), recipient (2), superalignment (2), fellowship (2), countr (2), end (2), crowd (2), counting (2), enables (2), every (2), peng (2), hao (2), based (2), core (2), current (2), multimodal (2), embeddings (2), propose (2), specific (2), unknowns (2), generated (2), aloe (2), active (2), via (2), stage (2), tmlr (2), xie (2), equal (2), contribution (2), publications (2), correlation (2), diversity (2), real (2), jia (2), systematically (2), oral (2), pattern (2), recognition (2), datasets (2), into (2), scone (2), wild (2), ming (2), within (2), region (2), covariate (2), labeling (2), processing (2), aha (2), human (2), only (2), need (2), chen (2), supervised (2), improves (2), ladder (2), nouha (2), dziri (2), wenjie (2), behaviors (2), post (2), grained (2), transfer (2), llm (2), low (2), patterns (2), grokking (2), recipe (2), performance (2), computation (2), colm (2), yang (2), fred (2), shuibai (2), hallucinate (2), agents (2), tasks (2), agentic (2), google (2), scholar (2), grateful (2), their (2), berkeley (2), will (2), working (2), labs (2), email (2), master (2), improve (2), efficient (2), such (2), developing (2), algorithmic (2), guarantees (2), ensure (2), deployment (2), detect (2), degree (2), thanks, website, template, jon, barron, site, professor, david, patterson, advice, good, know, before, referee, jennifer, widom, kleiman, tesled, technical, give, jones, hughes, launchbury, vaidya, uiuc, read, present, useful, links, visit, 鐨撴湀, 鈥渂right, moon鈥, conveys, brightness, whiteness, last, 鈥渨hite鈥, together, evokes, white, luminous, shining, night, sky, culture, symbolizes, connection, shared, emotions, reunion, distance, well, known, line, tang, poetry, says, bright, shines, thousand, miles, share, very, moment, think, carrying, same, spirit, gentle, light, warmth, bring, best, wishes, those, care, matter, far, apart, 鈥渂ye鈥, 鈥淗ow, yweh鈥, pronounce, lucas, hui, shueng, princeton, scientist, mentors, laboratories, other, experience, comp2611, cs220, cs400, iii, cs540, introduction, teaching, tnnls, tip, ijcai, wacv, socially, responsible, responsiblefm, organizer, professional, coordinator, optimization, silo, acm鈥檚, women, computing, wacm, community, prize, class, merits, social, worker, leader, 2016鈥2017, alibaba鈥揨hejiang, news, graduates, grants, program, 2024鈥損resent, faculty, worldwide, recipients, honors, european, density, estimation, which, capture, global, context, layer, tianyuan, dai, zhuoxuan, sudeep, katakol, manuel, brack, jinhong, lin, teng, richard, eli, shechtman, hareesh, ravi, ajinkya, kale, findings, identify, limitation, formulate, queryable, extraction, qare, explicitly, evaluate, query, sensitivity, invariance, yibing, towards, text, guided, disentangled, generator, attribution, remains, robust, presence, newly, released, generators, ellie, thieu, unknown, aware, algorithm, environments, designed, enhance, adaptation, incorporating, classes, transactions, tian, deep, ast, role, mentored, collaborations, rui, designs, general, tackle, possible, decaug, decomposed, feature, augmentation, takes, step, understand, network, architectures, nas, architecture, search, kaican, runpeng, benchmark, serve, strong, foothold, resorted, future, jun, zhu, bench, benchmarking, bridges, gap, between, coherent, cs762, link, jeongyeol, kwon, xuefeng, gregory, canal, feed, birds, exploiting, both, provably, learns, domain, invariant, hyperspherical, space, julian, katz, samuels, yifei, hypersphere, our, strategically, labels, examples, maximum, disambiguation, number, roughly, equalizes, maximally, disambiguate, types, thereby, maximizing, utility, fixed, budget, assisted, requires, api, sidestepping, weights, auxiliary, cvf, haifeng, where鈥檚, liability, generative, era, recovery, analyze, tuning, mathematical, uncover, structure, problem, difficulty, easy, extremely, hard, dacheng, georgia, climbing, what, can鈥攁nd, still, can鈥檛鈥擲olve, after, sft, maggie, huan, show, markedly, different, decompose, atomic, enabling, abilities, emerge, collapse, stages, acl, rolling, review, peiyang, shi, qiu, cognitive, level, bohao, huang, reinforcement, actually, teach, just, 鈥渟harpen鈥, what鈥檚, already, latent, base, set, test, directly, finding, clear, discover, capabilities, when, trained, wisely, hannaneh, hajishirzi, yuhan, cao, unlock, project, shows, efficiency, over, token, introduces, timestep, dependent, scheduling, further, boosts, allocating, more, mask, ratio, denoising, steps, modeling, liu, guanhua, zack, yanxin, zhangzhi, zhihan, chihan, cui, caspian, zhuang, diffusion, shashank, muralidhar, bharadwaj, siyang, cai, investigates, arguing, many, errors, arise, missing, knowledge, inference, misalignment, follow, statistically, dominant, shortcut, associations, instead, prompt, decisive, constraints, dataset, jiawei, ding, xuhan, tong, yangfan, testing, against, priors, succeed, short, degrade, ones, failures, are, poorly, understood, diagnostic, suite, probes, domains, reveals, consistent, sota, bilge, mutlu, mya, schroder, haorui, jessica, mirage, diagnosing, break, see, full, list, page, quantifying, neurocomputing, survey, hypo, einstein, visa, conferred, government, extraordinary, acclaim, specialized, field, olhf, passed, preliminary, exam, enter, dissertation, third, year, uw鈥揗adison, committee, guidance, tengyang, sala, yong, jae, lee, rob, poster, lightning, summit, group, oodbench, attending, san, diego, dec, 2rd, 7th, spending, time, andrey, zhmoginov, meta, superintelligence, multimodel, participate, doctoral, consortium, national, nsf, dlms, recent, updates, linkedin, twitter, wisc, dot, edu, junior, undergraduate, you, would, chat, plan, school, applications, ideas, related, feel, free, schedule, meeting, dedicate, mins, week, especially, underrepresented, groups, whoever, selecting, most, informative, signals, feedback, calibration, tight, annotation, compute, budgets, modes, boundaries, vlms, through, systematic, diagnostics, methods, strengthen, quality, learned, safety, techniques, statistical, designing, interpretable, principles, help, develops, theoretical, foundations, focus, safe, central, theme, leveraging, principled, uncertainty, objectives, reason, safely, candidate, sciences, fortunately, advised, honored, supported, division, have, worked, ucla, during, studies, obtained, bachelor, hkust, cse, homepage,

Text of the page (random words):
haoyue bai s homepage haoyue bai hi my name is haoyue bai i am a ph d candidate at the computer sciences department university of wisconsin madison fortunately advised by prof robert nowak i am honored to be supported by the openai superalignment fellowship i am a graduate visiting researcher in the computer science division at uc berkeley working with prof dawn song and dr yiyou sun i am grateful to have worked with prof sharon li uw madison dr wei cheng nec labs america and prof bolei zhou ucla during my graduate studies i obtained my bachelor s degree at zhejiang university and my master s degree at hkust cse supervised by prof s h gary chan my current research develops the theoretical and algorithmic foundations of reliable and trustworthy ai with a focus on open world robustness data efficient reliability and safe reasoning in foundation models a central theme of my work is leveraging post deployment wild data principled uncertainty and provable objectives to ensure that ai systems can generalize detect unknowns and reason safely under distribution shift out of distribution ood learning and open world robustness designing adaptive and interpretable learning principles that help ml models detect and generalize under distribution shifts such as semantic and covariate shifts correlation and diversity shifts reliable algorithms with provable guarantees developing machine learning techniques with statistical and algorithmic guarantees to ensure reliable deployment of ai systems under real world distribution shifts safety and reliability of foundation models understanding the failure modes and boundaries of large language models llms and vision language models vlms through systematic diagnostics such as failure analysis in open world reasoning robustness under distribution shift and generalization developing methods to strengthen their reliability and improve quality of learned representations and embeddings data efficient machine learning selecting the most informative data and signals e g human or model feedback to improve robustness calibration and coverage under tight annotation and compute budgets junior phd master undergraduate students if you would like to chat about life career plan graduate school applications or research ideas related to ai ml feel free to email me to schedule a meeting i will dedicate 30 mins every week especially for students from underrepresented groups or whoever is in need email haoyue bai at wisc dot edu google scholar nbsp nbsp twitter nbsp nbsp linkedin recent updates one paper on expert choice routing in dlms accepted to colm 2026 selected to participate in the cvpr doctoral consortium with support from the national science foundation nsf one paper on multimodel representation learning accepted to cvpr 2026 i will be spending time at meta superintelligence labs working with dr andrey zhmoginov this 2026 summer one paper on rl grokking recipe accepted to iclr 2026 attending neurips 2025 in san diego from dec 2rd to 7th one journal paper oodbench accepted to tpami 2025 two papers accepted for presentation at neurips 2025 math ai visiting researcher at uc berkeley with prof dawn song 鈥檚 group this 2025 summer one poster and a lightning talk accepted to agentic ai summit 2025 passed my phd preliminary exam and enter the dissertation stage in my third year at uw鈥揗adison grateful to my committee prof rob nowak prof yong jae lee prof fred sala and prof tengyang xie for their guidance and support one paper on black box ai content detection accepted to cvpr 2025 two journal paper olhf and aloe accepted to tmlr 2025 one paper aha accepted to neurips 2024 i am a recipient of the eb 1a einstein visa conferred by the us government for extraordinary acclaim in a specialized field one paper hypo accepted to iclr 2024 one paper scone accepted to icml 2023 one long survey paper accepted to neurocomputing 2022 one paper quantifying and understanding ood generalization accepted to cvpr 2022 as oral presentation publications see the full list on my google scholar page 鈭 indicates equal contribution your browser does not support the video tag the long horizon task mirage diagnosing where and why agentic systems break jessica wang haoyue bai 鈥 yiyou sun haorui wang shuibai zhang wenjie hu mya schroder bilge mutlu dawn song robert nowak arxiv 2026 paper code data llm agents succeed on short tasks but degrade on long horizon ones with failures that are poorly understood we introduce horizon a diagnostic suite that systematically probes long horizon tasks across domains and reveals consistent failure patterns in sota agents your browser does not support the video tag understanding why language models hallucinate testing reasoning against priors yangfan hu xuhan tong haoyue bai 鈥 xi ding shashank muralidhar bharadwaj siyang cai robert nowak jiawei zhang arxiv 2026 paper code dataset this paper investigates why large language models llms hallucinate by arguing that many errors arise not from missing knowledge but from inference misalignment where models follow statistically dominant shortcut associations instead of the prompt s decisive constraints your browser does not support the video tag expert choice routing enables adaptive computation in diffusion language models shuibai zhang caspian zhuang chihan cui zhihan yang fred zhangzhi peng yanxin zhang haoyue bai zack jia yang zhou guanhua chen ming liu conference on language modeling colm 2026 paper code this project shows that ec routing improves training efficiency and model performance over token choice routing and introduces timestep dependent expert scheduling that further boosts performance by allocating more computation to low mask ratio denoising steps your browser does not support the video tag rl grokking recipe how does rl unlock and transfer new algorithms in llms yiyou sun yuhan cao bohao huang haoyue bai hannaneh hajishirzi nouha dziri dawn song international conference on learning representations iclr 2026 paper code media coverage can reinforcement learning rl actually teach large language models new algorithms or just 鈥渟harpen鈥 what鈥檚 already latent in the base model we set out to test this directly and the finding is clear rl can discover new capabilities but only when trained wisely your browser does not support the video tag how and why llms generalize a fine grained analysis of llm reasoning from cognitive behaviors to low level patterns haoyue bai yiyou sun wenjie hu shi qiu maggie huan peiyang song robert nowak dawn song acl rolling review 2025 llms show markedly different generalization behaviors under post training we propose to decompose reasoning into atomic core skills enabling fine grained analysis of how specific abilities emerge transfer or collapse across training stages your browser does not support the video tag climbing the ladder of reasoning what llms can鈥攁nd still can鈥檛鈥擲olve after sft yiyou sun georgia zhou haoyue bai hao wang dacheng li nouha dziri dawn song neural information processing systems neurips math ai 2025 paper code we analyze how supervised fine tuning improves mathematical reasoning and uncover a ladder like structure of problem difficulty from easy to extremely hard your browser does not support the video tag where鈥檚 the liability in the generative era recovery based black box detection of ai generated content haoyue bai yiyou sun wei cheng haifeng chen the ieee cvf conference on computer vision and pattern recognition cvpr 2025 paper code media coverage we introduce a novel black box detection framework that requires only api access sidestepping the need for model weights or large auxiliary datasets your browser does not support the video tag aha adaptive human assisted out of distribution generalization and detection haoyue bai jifan zhang robert nowak neural information processing systems neurips 2024 paper code our approach strategically labels examples within a novel maximum disambiguation region where the number of semantic and covariate ood data roughly equalizes by labeling within this region we can maximally disambiguate the two types of ood data thereby maximizing the utility of the fixed labeling budget your browser does not support the video tag provable out of distribution generalization in hypersphere haoyue bai yifei ming julian katz samuels yixuan li international conference on learning representations iclr 2024 paper code this paper provably learns domain invariant representations in a hyperspherical space for ood generalization your browser does not support the video tag feed two birds with one scone exploiting wild data for both out of distribution generalization and detection haoyue bai gregory canal xuefeng du jeongyeol kwon robert nowak yixuan li international conference on machine learning icml 2023 selected into uw madison cs762 link paper code this paper bridges the gap between ood generalization and ood detection in one coherent framework your browser does not support the video tag ood bench benchmarking and understanding out of distribution generalization datasets and algorithms nanyang ye kaican li haoyue bai runpeng yu lanqing hong fengwei zhou zhenguo li jun zhu ieee conference on computer vision and pattern recognition cvpr 2022 selected as oral presentation paper code media coverage this benchmark may serve as a strong foothold that can be resorted to by future ood generalization research your browser does not support the video tag nas ood neural architecture search for out of distribution generalization haoyue bai fengwei zhou lanqing hong nanyang ye s h gary chan zhenguo li ieee international conference on computer vision iccv 2021 paper code this work takes the first step to understand the ood generalization of neural network architectures systematically your browser does not support the video tag decaug out of distribution generalization via decomposed feature representation and semantic augmentation haoyue bai rui sun lanqing hong fengwei zhou nanyang ye han jia ye s h gary chan zhenguo li the aaai conference on artificial intelligence aaai 2021 paper code this work designs a general ood generalization framework to tackle possible correlation shift and diversity shift in the real world mentored publications and student collaborations ast indicates equal contribution indicates the mentoring role your browser does not support the video tag deep active learning in the open world tian xie jifan zhang haoyue bai 鈥 robert nowak transactions on machine learning research tmlr 2025 paper code we introduce aloe a novel active learning algorithm for open world environments designed to enhance model adaptation by incorporating new ood classes via a two stage approach your browser does not support the video tag unknown aware ai generated content detection ellie thieu jifan zhang haoyue bai 鈥 arxiv 2025 we propose a learning framework for specific generator attribution that remains robust in the presence of unknowns or newly released generators your browser does not support the video tag towards text guided attribute disentangled multimodal representation learning yibing wei sudeep katakol manuel brack jinhong lin haoyue bai 鈥 yu teng li richard zhang eli shechtman hareesh ravi ajinkya kale cvpr findings 2026 this work identify a core limitation of current multimodal embeddings and formulate queryable attribute representation extraction qare to explicitly evaluate query sensitivity and attribute invariance your browser does not support the video tag countr a novel transformer approach for image based crowd counting haoyue bai 鈥 hao he zhuoxuan peng tianyuan dai s h gary chan european conference on computer vision eccv workshop 2022 paper code we introduce countr a novel end to end transformer approach for crowd counting and density estimation which enables capture global context in every layer of the transformer honors and awards openai superalignment fellowship worldwide 50 recipients across faculty and phd students 2024鈥損resent openai researcher access program recipient 2024 cvpr travel support award 2025 2026 research travel grants the hong kong university of science and technology 2019 outstanding graduates zhejiang university 2018 alibaba鈥揨hejiang news scholarship 2017 zhejiang university scholarship awards first prize academic scholarship top 3 first class scholarship for outstanding merits top 3 outstanding social worker scholarship top 2 outstanding student leader award top 6 2016鈥2017 service department community acm鈥檚 women in computing wacm mentor at uw madison graduate student mentor at uw madison seminar organization student coordinator systems information learning and optimization silo seminar professional service workshop co organizer neurips 2025 workshop on socially responsible and trustworthy foundation models responsiblefm conference reviewer icml neurips iclr cvpr eccv iccv aaai ijcai wacv journal reviewer ieee tpami ieee tnnls ieee tip teaching assistant university of wisconsin鈥揗adison comp cs540 introduction to artificial intelligence university of wisconsin鈥揗adison comp cs400 data science programming iii spring 2024 university of wisconsin鈥揗adison comp cs220 data science programming i fall 2022 the hong kong university of science and technology comp2611 computer organization fall 2019 spring 2019 other research experience 2024 5 2024 8 nec laboratories america princeton research scientist intern mentors dr yiyou sun dr wei cheng 2022 9 2023 12 university of wisconsin madison madison research assistant advisor prof sharon yixuan li 2021 6 2021 9 the chinese university of hong kong hong kong research assistant advisor prof bolei zhou 2018 1 2018 6 the hong kong university of science and technology hong kong visiting student advisor prof shueng han gary chan 2017 7 2017 8 the hong kong university hong kong summer intern advisor prof lucas c k hui how to pronounce my name my name is haoyue bai haoyue is pronounced 鈥淗ow yweh鈥 bai is pronounced 鈥渂ye鈥 my first name 鐨撴湀 means 鈥渂right moon鈥 鐨 conveys brightness and whiteness and 鏈 means moon my last name 鐧 means 鈥渨hite鈥 together my name evokes the image of a white luminous moon shining in the night sky in chinese culture the moon symbolizes connection shared emotions and reunion across distance a well known line from tang poetry says 鈥渁 bright moon shines across a thousand miles we share this very moment 鈥 i like to think of my name as carrying that same spirit 鈥 gentle light and warmth that bring best wishes to those we care about no matter how far apart we may be useful links i visit how to read write and present papers by prof vaidya at uiuc how to give a talk by profs jones hughes and launchbury student mentoring by prof wu writing skills writing technical papers from prof jennifer widom writing a math paper by profs kleiman and tesled task of a referee good to know before write a paper academic career advice from professor david patterson academic life a math site thanks to jon barron for website template