About

I did my Postdoc at MIT CSAIL, working with Prof. Kaiming He. Prior to this, I was very fortunate to be advised by Prof. Dina Katabi while completing my Ph.D. and S.M. at MIT. I received my B.E. in computer science from Yao Class, Tsinghua University.

My research interests lie in representation learning, generative models, and the synergy between these two. I aim to build intelligent visual systems that can understand and model the world beyond human perception.

News

[12/25] I co-organized the "Autoregressive Models Beyond Language" tutorial at NeurIPS 2025.

[12/25] I served as an Area Chair for WACV 2026, ICLR 2026, and ICML 2026.

[06/25] I co-organized the "Visual Generative Modeling: What’s After Diffusion?" workshop at CVPR 2025.

[01/25] I co-instructed 6.S058: Introduction to Computer Vision at MIT with Phillip Isola and Bill Freeman.

[12/24] I served as an Area Chair for ICLR 2025, ICML 2025, and ICCV 2025.

[09/24] I joined Prof. Kaiming He's group as a Postdoctoral Associate at MIT CSAIL.

[05/24] I defended my Ph.D. thesis: Towards a Unified Framework for Visual Recognition and Generation via Masked Generative Modeling.

Selected Publications [Full list]

Back to Basics: Let Denoising Generative Models Denoise
T. Li, and K. He
Computer Vision and Pattern Recognition (CVPR), 2026
[Paper] [Code]
Fractal Generative Models
T. Li, Q. Sun, L. Fan, and K. He
Transactions on Machine Learning Research (TMLR), 2025
[Paper] [Code] [Demo]
Autoregressive Image Generation without Vector Quantization
T. Li, Y. Tian, H. Li, M. Deng, and K. He
Neural Information Processing Systems (NeurIPS), 2024
[Paper] [Code] [Demo] [Hugging Face] [Talk]
Spotlight Presentation (top 2%)
Return of Unconditional Generation: A Self-supervised Representation Generation Method
T. Li, D. Katabi, and K. He
Neural Information Processing Systems (NeurIPS), 2024
[Paper] [Code] [Slides]
Oral Presentation (top 0.4%)
Leveraging Unpaired Data for Vision-Language Generative Models via Cycle Consistency
T. Li, S. Bhardwaj, Y. Tian, H. Zhang, J. Barber, D. Katabi, G. Lajoie, H. Chang, and D. Krishnan
International Conference on Learning Representations (ICLR), 2024
[Paper]
Spotlight Presentation (top 5%)
MAGE: MAsked Generative Encoder to Unify Representation Learning and Image Synthesis
T. Li, H. Chang, S. K. Mishra, H. Zhang, D. Katabi and D. Krishnan
Computer Vision and Pattern Recognition (CVPR), 2023
[Paper] [Code] [MIT News]
Targeted Supervised Contrastive Learning for Long-Tailed Recognition
T. Li*, P. Cao*, Y. Yuan, L. Fan, Y. Yang, R. Feris, P. Indyk and D. Katabi
Computer Vision and Pattern Recognition (CVPR), 2022
[Paper] [Code]
Label-Free Few Sample Knowledge Distillation for Efficient Network Compression
T. Li, J. Li, Z. Liu, C. Zhang
Computer Vision and Pattern Recognition (CVPR), 2020
[PDF] [Code] [Arxiv]
Through-Wall Human Pose Estimation Using Radio Signals
M. Zhao, T. Li, M. Alsheikh, Y. Tian, H. Zhao, A. Torralba and D. Katabi
Computer Vision and Pattern Recognition (CVPR), 2018
[PDF] [Website]
Spotlight Presentation
Multi-Scale Dense Networks for Resource Efficient Image Classification
G. Huang, D. Chen, T. Li, F. Wu, L. Maaten, K. Weinberger
International Conference on Learning Representations (ICLR), 2018
[PDF] [Github]
Oral Presentation



Recipes

I really love cooking, almost the same as doing research. Here I list some of my favorite recipes as comforts for a child from Hunan, China when he misses his hometown food.

梅菜扣肉

来自奶奶的菜谱：1，将肉洗净，皮用刀刮一下，2，将洗好的肉放一定凉水加葱姜料酒煮，水开后还要小火再煮一会儿，煮至能用筷子从猪皮扦进去就可以了，3，把肉拿出来放在碗里，牙签戳一戳，猪皮上面抹一点蜂蜜白酒；4，鍋里放大概100克植物油烧六成热将肉放入，猪皮在下面用中火煎，首先要准备好没水的鍋盖，盖好但要留一点缝隙，让水份从缝隙蒸发，盖子的作用就是热油遇到有水份的肉会爆，防止伤眼睛和皮肤；5，把火关小或完全关灭，再看是否炸成金黄色了，如果还没有再炸一会儿；6，把炸好的肉放入刚煮肉的水中开火煮一会儿，看皮软了就取出放在板子上，从猪皮处下刀切成一公分厚片；6，在肉体上抹上盐，再猪皮抹上酱油即可；7，把肉整齐放入碗中，猪皮在下面；8，先把酸菜用温水泡一下，再洗干净济干水份，放入锅中炒干水分，放点油、盐、辣椒，辣椒可以多放一点，炒均匀放入肉上面；9，高压锅蒸45分钟，取出时用另只碗扣过来，即肉在上面，撒上葱花即可。

东安鸡

木耳泡发。6只鸡腿冷水料酒（一定要）姜片下锅，焯水10-15分钟捞出（美国鸡比较腥）。鸡腿去骨切条，2根长红椒去芯去籽切丝，3-4只小米椒切段，姜片去皮切丝，花椒去籽，小葱切段。锅中热油，加花椒炒香，去掉花椒，加小米椒，姜丝炒香，加入鸡腿肉炒熟。延锅边依次淋入少量五粮液、50ml白醋，加入开始煮鸡腿的汤，加入红椒一勺盐，两勺辣椒粉，木耳，焖煮5分钟。开盖加入两勺香醋，半勺白糖，葱段，收汁即可。

手抓饼

培根、生菜不用放油煎熟备用。手抓饼（印度煎饼）两面煎熟不要焦备用。打一个鸡蛋，一面煎熟另一面半凝固放上手抓饼，再煎半分钟翻面，放上培根生菜，再加入番茄酱/蛋黄酱即可。

新奥尔良鸡翅

鸡翅洗净擦干，可以稍微划开，加入新奥尔良腌料，冰箱腌制2小时。烤箱预热180度，烤盘放入，烤10分钟，取出翻面再烤10分钟，刷蜂蜜烤6分钟，翻面刷蜂蜜再烤6分钟

土豆红烧肉

五花肉炒熟，稍煎一会，去掉些油，下土豆翻炒，加老抽、醋、少许糖、味精、辣椒粉调的水，焖至土豆半烂，收汁，加葱即可出锅。

肉丝面

猪梅肉CT Butt切丝，加入生抽白胡椒粉少量盐味精调味抓匀腌制20分钟。锅中冷水加肉，大火烧开后撇去浮沫，转中小火炖1-2小时，出锅前加入一勺盐。碗里放猪油、一勺盐、少量味精、葱花、老抽调味，加入肉汤，再加入煮好的面，盖上肉码再加个煎蛋即可。

苹果蛋糕

两个鸡蛋一包豆奶粉（精髓）20g白砂糖打匀，加入20g融化后的黄油，100ml牛奶。再筛入70g面粉，搅拌均匀。脆苹果削皮切薄片加入面糊。烤箱预热325F烤制45分钟即可。

莴笋丝

两根莴笋去皮切丝。猪油热锅，加入莴笋丝翻炒至稍软，骨头汤里加入大半勺盐和少量味精化开，倒入锅内，稍煮一会即可。

手擀面

两个鸡蛋打散加水至125ml，两杯面粉，用手在面粉里打窝，慢慢将水加入面粉。成型后揉100下，放置10分钟再揉100下。分成三份擀平，切成面条。下水加2勺盐煮开5分钟捞出。碗里放猪油老抽辣子，加面后加入辣椒碎葱花，浇上滚油即可。可以放点醋。

热干面

一勺花生酱，一勺芝麻酱，一勺香油，一勺油泼辣子，一勺老抽，一勺生抽，半勺盐，少许味精调成酱汁。面条煮熟过冷水，撒上酱汁香菜芝麻即成

牛肉粉丝

牛腩1kg，筒子骨500g，放三片姜一些料酒切块焯水。锅内放油，加入蒜姜丝青花椒干辣椒炒香，加入焯好水的肉炒2分钟左右，加入豆瓣酱炒匀，放入砂锅。砂锅加入一根肉桂，四片香叶，一颗草果，五颗八角，加水没过牛肉，两勺生抽一勺老抽，炖煮4小时即可。出锅前加两勺盐再炖10分钟，加入味精。粉丝焯熟，加入牛肉汤与盖码，撒上葱花香菜芝麻辣子即成。

茄汁土豆火腿饭

番茄去皮切块，锅入黄油，翻炒番茄，加半勺盐，加水（300ml），盛出。土豆切丁，洋葱切丁，大蒜切碎，炒香大蒜洋葱，加黄油，加土豆，加泡好的米（大半杯），加番茄汤。熬煮15-20分钟。最后加入香芹黑胡椒出锅，撒上芝麻即成。

龙虾泡饭

杀好的龙虾洗净去腮，入沸水煮1-2分钟，钳子多煮2分钟，取肉，虾骨剪碎，加入橄榄油炒出香味，盛出，再加入橄榄油，胡萝卜丁、洋葱丁、芹菜丁炒软，洋葱边缘焦黄，加入1/4个西红柿丁，加入虾骨，炒热后加入30ml白葡萄酒，加水到没过虾骨，中火半小时，过滤成高汤。煮高汤时，大半杯米煮饭（少水），取出的虾肉用黑胡椒海盐白葡萄酒腌制，大火多油，两面煎20s，加入少许白葡萄酒去腥。半杯米做炒米：米洗净，油稍稍热，加入米，不断翻炒至金黄色，盛出。将过滤后的高汤入锅，加热，加入米，黑胡椒，盐，味精调味。加入芹菜碎，龙虾肉，炒米，最后点缀葱花即成。

蟹黄豆腐

软豆腐切小块，4个咸鸭蛋黄用微波炉加热后捣碎，五花肉切末，加入少量料酒盐拌匀腌制。烧热油，下入咸蛋黄炒开后下入肉末炒熟，加入豆腐翻炒两下，加入淀粉水煮开，收一会汁撒上葱花即成。

蒜蓉粉丝蒸龙虾

龙虾一只切开，花生油烧开，加蒜蓉爆香，加入少许蒸鱼豉油、盐、味精调味成酱汁。粉丝泡发，铺底，龙虾摆盘，浇上蒜蓉酱汁和少许豉油，开锅蒸12分钟，加入葱碎、蒸鱼豉油，再蒸3分钟，浇上热油即成。

蚂蚁上树

五花肉末加入生抽料酒腌制，粉丝泡发，姜末、豆瓣酱爆香，加入肉末炒熟，加一点点盐，加入粉丝，加入老抽、味精、辣椒粉调的水，焖3分钟，收汁，加入葱花即成。

Contact

MIT Schwarzman College of Computing
51 Vassar Street, 733
Cambridge, MA 02139
tianhong[at]mit.edu