Meta tags:

Headings (most frequently used words):

and, research, honors, awards, teaching, services, other, activities, fereshte, khani,

Text of the page (most frequently used words):
and (19), the (17), models (11), how (8), model (8), for (7), data (6), that (6), national (5), 2011 (5), can (5), #fereshte (4), stanford (4), with (4), iranian (4), olympiad (4), informatics (4), teaching (4), from (4), discrimination (4), their (4), this (4), groups (4), 2023 (3), other (3), 2009 (3), have (3), when (3), showed (3), blog (3), post (3), where (3), performance (3), new (3), domain (3), unlabeled (3), use (3), alignment (3), openai (3), khani (3), follow (2), scientific (2), committee (2), computer (2), iran (2), 2012 (2), conference (2), 2017 (2), science (2), nlp (2), panelist (2), iclr (2), contest (2), inoi (2), summer (2), 2013 (2), assistant (2), university (2), machine (2), learning (2), 2020 (2), probability (2), review (2), grant (2), hong (2), kong (2), medal (2), lead (2), research (2), previously (2), shown (2), all (2), individuals (2), are (2), not (2), measure (2), facct2021 (2), fairness (2), cannot (2), proposed (2), only (2), predicts (2), training (2), agree (2), two (2), human (2), collaborative (2), users (2), similar (2), convey (2), before (2), under (2), supervision (2), powered, fork, minimal, mistakes, academicpages, jekyll, feed, sitemap, free, time, like, writing, running, hiking, backpacking, reading, activities, member, vision, workshop, phd, representative, grace, hopper, moderator, women, wids, breakout, application, peter, norvig, nightcap, speaker, trustml, unlimited, admission, 2018, reviewer, neurips, icml, alt, acl, designer, first, round, students, lecturer, math, camp, farzanegan, highschool, 2007, structure, artificial, intelligence, design, algorithm, sharif, technology, fall, theory, linear, algebra, services, awarded, winter, school, chinese, gold, 2008, selected, elite, recipient, elites, foundation, silver, international, ioi, plovdiv, bulgaria, ranked, 3rd, asia, regional, acm, icpc, 2010, edge, doctoral, fellowship, 2016, 2022, honors, awards, light, increasing, prevalence, necessary, address, problem, see, try, understand, investigate, mitigate, study, feedback, loops, created, unexpected, reasons, cause, exhibit, example, adding, same, amount, feature, noise, inductive, bias, overparameterized, regimes, addition, scenarios, protected, known, priori, there, exponential, number, such, what, kind, statistical, possible, loss, discrepancy, among, safeml, iclr2019, icml2020, reliable, knows, does, know, interested, selective, classification, abstain, provide, true, prediction, high, consistent, unanimously, resulting, 100, precision, however, become, more, complicated, unanimous, agreement, anymore, show, find, disjoint, those, neurips2022, acl2016, reliability, one, main, concern, about, utilizing, real, world, poor, domains, even, slightly, different, previous, work, leverage, through, pretraining, finetuning, achieve, better, make, robust, against, spurious, features, llms, large, pool, augment, has, low, acl2023, iclr2021, robustness, ensuring, conform, values, intentions, focus, multiple, engage, each, align, preference, without, interferring, envision, future, developed, fashion, open, source, software, wikipedia, benefiting, diverse, user, inputs, improved, quality, scenario, materialize, need, help, knowledge, verify, impact, changes, diffs, regression, tests, small, step, direction, pragmatics, planning, inference, dialogue, information, tacl2018, codev, neurips2023, researcher, joining, was, got, mit, martin, rinard, percy, liang, microsoft, google, scholar, twitter, san, francisco, california, posts, publications,

Text of the page (random words):
fereshte khani fereshte khani publications blog posts fereshte khani follow san francisco california openai twitter google scholar i am a researcher at openai before joining openai i was at microsoft and before that i got my ph d at stanford under supervision of percy liang and my m s at mit under supervision of martin rinard research alignment alignment is ensuring machine learning models conform to human values and intentions in neurips2023 we focus on collaborative alignment where multiple individuals engage with the model and each other to align the model to their preference without interferring other users i envision a future where nlp models are developed in a collaborative fashion similar to open source software or wikipedia benefiting from diverse user inputs for improved quality and fairness for this scenario to materialize we need to help users to convey knowledge and verify the impact of their proposed changes to models similar to diffs or regression tests codev is a small step in this direction in tacl2018 we showed how human use pragmatics planning and inference in dialogue to convey information robustness one main concern about utilizing ml models in the real world is their poor performance in new domains even when the new domain is slightly different from the training domain in my previous work we have shown how to leverage unlabeled data through pretraining and finetuning to achieve better performance in a new domain iclr2021 and how to use unlabeled data to make a model robust against spurious features facct2021 in acl2023 we showed how to use llms as a large pool of unlabeled data to augment groups that the model has low performance on reliability how can we have a reliable model that knows when it does not know i am interested in selective classification where a model can abstain if it cannot provide a true prediction with high probability previously i proposed a model that only predicts if all models consistent with training data unanimously agree resulting in 100 precision acl2016 however as models become more complicated we cannot do the unanimous agreement anymore in neurips2022 we show how to find two disjoint models and only predicts if those two model agree fairness ml models can lead to discrimination and in light of their increasing prevalence it is necessary to address this problem see this post in my research i try to understand how we can investigate and mitigate discrimination of ml models and how to study the feedback loops created by ml models previously i have shown unexpected reasons that cause ml models to exhibit discrimination for example adding the same amount of feature noise to all individuals icml2020 blog post or the inductive bias in overparameterized regimes facct2021 blog post can lead to discrimination in addition in scenarios when protected groups are not known a priori or there is an exponential number of such groups i showed that what kind of statistical measure is possible to measure loss discrepancy among groups safeml iclr2019 honors and awards edge doctoral fellowship 2016 2022 ranked 3rd in the asia regional acm icpc contest iran 2010 silver medal in international olympiad in informatics ioi 2009 plovdiv bulgaria selected as a national scientific elite and the recipient of the grant from the iranian national elites foundation 2009 2013 gold medal in iranian national olympiad in informatics inoi 2008 awarded winter school grant from chinese university of hong kong hong kong 2011 teaching and services teaching assistant for machine learning stanford summer 2020 fall 2020 review of linear algebra probability theory review teaching assistant for data structure 2011 artificial intelligence 2011 and design of algorithm 2012 sharif university of technology lecturer in summer math camp farzanegan highschool 2007 2013 teaching students for iranian national olympiad in informatics inoi 2009 2011 contest designer of iranian national olympiad in informatics first round 2011 reviewer for neurips icml iclr alt acl stanford computer science admission committee 2018 speaker and panelist in trustml unlimited iclr 2023 panelist on application of ai with peter norvig at the nightcap 2023 co moderator in women in data science wids conference nlp breakout 2017 stanford cs phd representative at grace hopper conference 2017 member of the scientific committee computer vision workshop iran 2012 other activities in my free time i like reading writing running hiking and backpacking sitemap follow feed 2023 fereshte powered by jekyll academicpages a fork of minimal mistakes