Nilpa jha @under_fitting

Research Engineer @Google Deepmind | Hill Climbing on Veo and Gemini Omni| Interest : Video Understanding, generation, evals, safety Seattle, WA Joined August 2021

Tweets

86
Followers

28
Following

256
Likes

187

Yabba @YabbaCric

8 months ago

Contrary to popular opinion, Bazball is more effective in bowler friendly conditions! Counter attacking and maximizing runs while at the crease is uniquely valuable in these conditions, since you are never in as a batsman, there is going to be a ball with your name on it. #Ashes

1 1 1 24 0

View Details

elvis @omarsar0

9 months ago

Lookahead Routing for LLMs Proposes Lookahead, a routing framework to enable more informed routing without full inference. Achieves an average performance gain of 7.7% over the state-of-the-art. Here is why it works: Lookahead is a new framework for routing in multi-LLM systems, deciding which model should handle each query. Key idea: Instead of routing based only on the input query, Lookahead predicts latent representations of potential responses, giving it a “peek” into what each model would say without fully generating text. Smarter decisions: This response-aware prediction makes routing more context-sensitive and accurate, especially for open-ended or preference-driven tasks. Efficient learning: It’s highly data-efficient, reaching full performance with only 16% of training data compared to baselines, and learns better semantic representations for routing. Dual design: Works with both causal and masked LM variants, generalizing across multiple architectures. Performance: Outperforms state-of-the-art routing methods across seven benchmarks, with the biggest gains in nuanced, creative tasks. Lookahead shows that adding lightweight generative foresight can make multi-model systems more adaptive and cost-efficient without needing full model inference.

19 39 236 24K 174

View Details

amit @amitisinvesting

9 months ago

the thing about some of these momentum stocks is that they might not be able to save themselves with a good earnings because they dont have any... google could have a bad few months and then have a blowout quarter and change the narrative what are some of these companies going to say? no backlog, no revenue, no EPS...some of them don't even take shareholder questions! the worst part of this drawdown for them is that if they don't have anything substantial to say for earnings...and the market already trades at a historically high multiple...what happens when they can't put up any actual results?

136 35 994 212K 72

View Details

Tobias Weyand @0xtob

9 months ago

@ahmetius @SachitMenon Here's our video about the paper: youtube.com/watch?v=gOdVhJ…

0 1 3 183 0

View Details

Tobias Weyand @0xtob

9 months ago

Attending #ICCV2025? Come chat with us about our Minerva dataset that tests if models can truly reason about videos! 🕵️‍♀️ @ahmetius and @SachitMenon will be presenting the dataset at Poster Session 5 tomorrow (Thurs, Oct 23) morning. Find them at poster #391.

Tobias Weyand @0xtob

a year ago

We're excited to release Minerva 🕵️‍♀️, a benchmark to evaluate if AI can truly reason about videos, from spotting game-changing moments in sports 🏀 to understanding character motivations in short films 🍿. We provide the "why" behind the answers! Pointers below 👇

1 1 11 1K 1

1 3 5 2K 0

View Details

Tobias Weyand @0xtob

2 years ago

Thank you @JeffDean , very much appreciate the boost! This is really a team effort with my amazing colleagues @NagraniArsha, Mingda Zhang, @raminia, Rachel Hornung, @nitesh_ai, @under_fitting, Austin Meyers, @zhouxy2017, @BoqingGo, @CordeliaSchmid, @sirotenko_m, @ZhuZhu66595.

Jeff Dean @JeffDean

2 years ago

A nice new benchmark for long video understanding by Tobias Weyand @0xtob and others. This is likely to be one of the new frontiers of capabilities for large-scale multimodal models, and it's great to have a new benchmark to assess others in this area.

4 17 169 42K 51

0 6 27 17K 4

View Details

Tobias Weyand @0xtob

2 years ago

Excited that our work on Long video understanding is being featured by @GoogleAI !

Google AI @GoogleAI

2 years ago

Can #AI truly understand long videos? Tobias Weyand & the Google Research team are testing the limits w/ Neptune, an open-source benchmark for long video understanding. Dive into the details & see how AI tackles temporal reasoning, cause & effect, & more →goo.gle/4esTTNM

11 30 165 51K 52

0 2 10 844 0

View Details

Google AI @GoogleAI

2 years ago

Datasets for evaluation of long video understanding are rare. So with this in mind, today we describe Neptune, an open-source evaluation dataset that includes multiple-choice and open-ended questions for videos of variable lengths up to 15 minutes. More →goo.gle/3B41nZV

15 70 277 47K 84

View Details

Gautam Kamath @thegautamkamath

3 years ago

You might have heard that word2vec, test of time award winner at #NeurIPS2023, was rejected from #ICLR2013 back in the day. Interestingly, one reviewer felt so strongly that they recommended "Strong Reject" four times.

33 135 1K 365K 261

View Details

Godman Chikna @Madan_Chikna

3 years ago

Naveen Ul Haq's coach after #LSGvRCB match #ViratKohli #GautamGambhir #CricketTwitter

189 1K 11K 652K 46

View Details

Nilpa jha @under_fitting

3 years ago

Lol that smiley

Jon Uleis @MovingToTheSun

3 years ago

My new favorite thing - Bing's new ChatGPT bot argues with a user, gaslights them about the current year being 2022, says their phone might have a virus, and says "You have not been a good user" Why? Because the person asked where Avatar 2 is showing nearby

999 10K 50K 8.0M 6K

0 0 0 87 0

View Details

geo @geometrieva

3 years ago

making AI art feels like some high-tech occult shit. chanting magic words into a prompt field and materializing never-before-seen people, forbidden places and impossible materials. I can't imagine anything more interesting.

18 45 280 23K 18

View Details

Nilpa jha @under_fitting

3 years ago

"Talented" enough to write one too

Elon Musk @elonmusk

3 years ago

Ordered “How to Scam” book months ago but still not arrived. Starting to think it could be … but no I’m sure it’s coming soon.

16K 15K 243K 27.2M 951

0 0 1 55 0

View Details

Roman Semenov 🌪️ @semenov_roman_

3 years ago

The new jailbreak is so fun

543 5K 35K 3.6M 3K

View Details

The Telegraph @ttindia

3 years ago

US short seller Hindenburg Research has rejected Adani Group's charge that its report was an attack on India, saying a "fraud" cannot be obfuscated by nationalism. #HindenburgReport #AdaniGroup @HindenburgRes telegraphindia.com/business/hinde…

5 26 61 4K 0

View Details

S.R.Praveen @myopiclenses

3 years ago

Adani CFO, an Australian citizen, uses the tricolour and even Jallianwala Bagh, as a shield against the allegations raised by Hindenburg Research.

Sanjay Sipahimalani @SanSip

3 years ago

The Adani Group CFO, who displays the Indian flag and cites Jallianwala Bagh, is a citizen of Australia. (Source: livemint.com/companies/news…)

29 381 1K 119K 25

1 16 38 4K 1

View Details

Mohammed Zubair @zoo_bear

4 years ago

Not a single tweet from @ndtvProfit @ndtv, @ndtvvideos, @ndtvindia @ndtvfeed on Hindenburg Research.

704 4K 13K 1.0M 106

View Details

#🟦 🟧 Rieva Lesonsky 🌻🌻🐞⚾️ @Rieva

3 years ago

Wow, @AmazonFresh is now charging a delivery fee for orders less than $150. And the store doesn't honor its promotional prices for delivery. So far Amazon has raised the price of Prime, eliminated its charity contributions & rarely delivers next-day. Why am I paying for Prime?