Aditay Tripathi

Research Scientist @GoogleDeepMind

[Harman Singh @ ICML 🇰🇷🇰🇷 @Harman26Singh]

Persistent self-improvement is model- and task-dependent. Do the gains generalize? Checkpoints are scored on held-out games.

In Hard chess, Claude Fable 5 climbs from 37.5% to 73.3% on them late in training; Claude Opus 5 from 25.0% to 66.4%. Against a stronger engine, gains are generally smaller.

This motivates evaluating plasticity across a distribution of agentic tasks during frontier agent evals.

[Rishabh Tiwari @rish2k1]

https://t.co/qCh4LaqUNh

[Darshan @thought2vec]

Hi all! Today I will present our work, SRL-CLIP, a recipe to efficiently adapt CLIP to videos at the CV4Smalls CVPR workshop in Room 102. I have an oral talk scheduled at 2 pm followed by the poster session. Please do stop by! Work done with @zeeshank95 @MakarandTapaswi #CVPR2026

[Darshan @thought2vec]

Frontier models have become excellent at understanding videos. But what happens when we test them outside the comfort zone of Western, English-centric data? In our #CVPR2026 (Highlight) work, we pushed these models to their limits to see if they can function effectively in diverse global contexts. The results? They are struggling. Work done with @NagraniArsha @skawshik11 @Harman26Singh @dinesh_tewari1 @0xtob @CordeliaSchmid Anelia Angelova @shachi_dave (1/7)

[Harman Singh @ ICML 🇰🇷🇰🇷 @Harman26Singh]

Can LLMs Self-Verify? Much better than you'd expect.

LLMs are increasingly used as parallel reasoners, sampling many solutions at once.
Choosing the right answer is the real bottleneck.

We show that pairwise self-verification is a powerful primitive.

Introducing V1, a framework that unifies generation and self-verification:

💡 Pairwise self-verification beats pointwise scoring, improving test-time scaling
💡 V1-Infer: Efficient tournament-style ranking that improves self-verification
💡 V1-PairRL: RL training where generation and verification co-evolve for developing better self-verifiers

🧵👇

[Mansi @im_mansigupta]

Saw a man forcing a woman on the highway into a car. We stopped and asked, he threatened us to leave saying it's an "internal matter". We called the local police station and shared the details including location and photo of the car. Can anything else be done? In Karnataka.

[Nithish Kannen @NithishKannen]

👀

[Partha Talukdar @partha_p_t]

@GoogleDeepMind  India 🇮🇳 & Japan 🇯🇵 are looking for strong candidates in multilinguality, multicultural, & multimodality areas.

RS Bangalore: https://t.co/6df1aak2qK

RS Tokyo: https://t.co/VCHoCZ1Ehb
RE Tokyo: https://t.co/aiap4v0uum

[Harman Singh @ ICML 🇰🇷🇰🇷 @Harman26Singh]

🚨 New @GoogleDeepMind paper

𝐑𝐨𝐛𝐮𝐬𝐭 𝐑𝐞𝐰𝐚𝐫𝐝 𝐌𝐨𝐝𝐞𝐥𝐢𝐧𝐠 𝐯𝐢𝐚 𝐂𝐚𝐮𝐬𝐚𝐥 𝐑𝐮𝐛𝐫𝐢𝐜𝐬 📑
👉 https://t.co/oCk5jGNYlj

We tackle reward hacking—when RMs latch onto spurious cues (e.g. length, style) instead of true quality.
#RLAIF #CausalInference

🧵⬇️

[Nithish Kannen @NithishKannen]

🚨 New paper alert 🚨:

T2I models are about to revolutionize industries...but do they work equally well for all cultures ? 🌏 What is the *cultural diversity* of images generated for under-specified prompts? 🤔

Our latest research delves into this!

https://t.co/ZQOVWQHHOs

[Aditay Tripathi @AditayTripathi]

Thanks Prof @CSProfKGD for the insightful conversation.

[Rishabh Tiwari @rish2k1]

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

[Aditay Tripathi @AditayTripathi]

@negar_rz @negar_rz my open review email is [email protected]

[Aditay Tripathi @AditayTripathi]

@negar_rz Hi , I would like to volunteer. My publications can be found here https://t.co/btd5z9o4V5

[Aditay Tripathi @AditayTripathi]

@CVPR @W8KIDDRtlVIH4a6 I have revised the camera-ready submission after completing the copyright form. Will this a problem ?

[Aditay Tripathi @AditayTripathi]

@CVPR I cannot find the copyright form for our camera-ready submission. What should I do ?

[Aditay Tripathi @AditayTripathi]

@RealAAAI, what is the author response deadline? On the webpage, it is the 23rd, but on the email, it is the 24th 11:49 pm UTC-12.

[Aditay Tripathi @AditayTripathi]

@ashutoshk2401 @apoorv_umang Can I join your watch party?

[Apoorv Saxena @apoorv_umang]

If you're attending #acl2022nlp in Dublin, please do check out our work on seq2seq knowledge graph completion and question answering!

Poster session: 17:00-18:00 Mon 23 May
Oral presentation: 15:15 Tue 24 May (Question Answering 2)

[Cees Snoek @cgmsnoek]

#CVPR2022 cam-ready: BoxeR: Box-Attention for 2D and 3D Transformers. w/ @kienduynguyen94, Jihong Hu, Olaf Booij & @Martin_R_Oswald; @UvA_Amsterdam @TomTom #AtlasLab https://t.co/JXZ2EoSvXO 1/n

[Divy Thakkar @divy93t]

[Announcement] Ph.D Fellowships for India and Southeast Asia are now open till May 18! We are open to applications from Ph.D students in all years (including those who are about to join in 2022!).

APPLY and SHARE!

https://t.co/fkPtG3kdUR