Rishabh Tiwari
@tiwarishabh16
Followers
580
Following
245
Media
8
Statuses
27
Pre-doctoral Researcher @GoogleAI | Engg. Physics @iitism_dhanbad ’22 | Research area: Robust interpretable deeplearning
Bengaluru, India
Joined May 2019
Don't wanna be here?
Send us removal request.
Explore trending content on Musk Viewer
TEMU
• 453657 Tweets
Afghanistan
• 331307 Tweets
最大5,000円報酬
• 241581 Tweets
روما
• 204815 Tweets
領空侵犯
• 160032 Tweets
Arlington
• 159190 Tweets
Sven
• 130391 Tweets
Delpino
• 62519 Tweets
Abbey Gate
• 35038 Tweets
Scout
• 34019 Tweets
Olmo
• 33890 Tweets
Finale
• 33798 Tweets
Nintendo Direct
• 29811 Tweets
Alex Sandro
• 26583 Tweets
Medina
• 24294 Tweets
Charli
• 24215 Tweets
D-1 TO CHANYEOL SOLO
• 20163 Tweets
伊藤さん
• 17954 Tweets
Nelita
• 14870 Tweets
瑞希誕生日
• 14148 Tweets
Ubisoft
• 14069 Tweets
Beşiktaşlı Oğuzhan
• 11378 Tweets
Lowe
• 10892 Tweets
Christensen
• 10151 Tweets
Pinned Tweet
Excited to share that I will be joining
@UCBerkeley
as a PhD student this Fall. I wish to continue working on the efficiency and interpretability of deep learning models, hoping to deepen our understanding of these systems.
32
13
690
🚨 Excited to share our
#ICML2023
work on the Feature Sieve, by which we automatically identify and suppress irrelevant or spurious features in deep networks, hence mitigating simplicity bias.
Paper:
1
16
57
What a way to end my
#WACV2024
. Got a chance to chat (turned to short AMA 😅) with amazing and super humble
@CSProfKGD
. Hope you get a chance to tick off visiting India from your bucket list soon.
@AditayTripathi
@sourabhgothe
1
1
24
PS cross posting my LinkedIn post in the hopes of expanding my X network (first step towards becoming a serious AI researcher😅).
0
0
15
This accomplishment would not have been possible without the guidance from my incredible mentors:
@doktorshenoy
,
@jainprateek_
,
@divy93t
,
@ManishGuptaMG1
,
@DjDvij
.
1
0
14
A very interesting paper by my good friend
@rach_it_
w/
@GoogleAI
and
@GoogleDeepMind
. Shows the strategy to compose different LLMs to unlock new capabilities. Seems like a critical step in the right direction towards practical usability of LLMs.
Extending an LLM for new knowledge sources is tedious—fine-tuning is expensive/causes forgetting, LoRA is restrictive.
Excited to share our work where we show that an LLM can be efficiently *composed* with specialized (L)LMs to enable new tasks!
🧵(1/8)
21
139
671
0
1
11
A new benchmark dataset to evaluate LLMs on indic languages!
IndicGenBench is now released on huggingface:
0
11
40
0
0
4
0
0
3
(2) Problem:
Simplicity bias: Models depend on weak, easily computed predictive features even if stronger alternatives exist.
Spurious correlations: DNNs latch onto apparent feature-label correlations present in biased training data distributions.
1
1
2
(7) Optimization: We use accuracy on validation set for setting hparams. We require no knowledge of potential spurious attribute dimensions, nor even a curated / balanced validation set. Of course, well-chosen validation data can be exploited to guide the resulting classifier.
1
0
2
(5) Hypothesis: Simple features are learned early, lower in the network, and proliferate through the rest of the network. What if we can automatically identify and suppress such features?
1
1
2
(3) Results (quant): SIFER gives around 65.7% accuracy without access to test distribution outperforming other baselines. Using a validation set drawn from test distribution further improves accuracy by 6.3%. SIFER beats SOTA by significant margins across debiasing benchmarks.
1
0
2
(6) Workflow: We alternate between identifying simple features (i.e., reading them out from intermediate network layers), and erasing them in lower layers. This simple yet powerful scheme mediates between (“sieves through”) competing features, picking only those that generalize.
1
1
1
(4) Results (qual): SIFER automatically identifies and suppresses spurious features without access to attributes or other distributional information regarding features. Examples shown below for CelebA-Hair (spurious gender feature) and BAR (spurious background features).
1
0
1
@gaganjain1582
@eccvconf
@rajatkoner
@jainprateek_
Congratulations Gagan 🎉, many more ahead.
Party when? :)
1
0
1