←
ExceptionRaised88 posts
ExceptionRaised avatar

ExceptionRaised

@except_raised

Funding exceptional research talent, found in unexpected places

◷ Joined April 2025▧ 29 Photos and videos
ExceptionRaised avatar
ExceptionRaised@except_raised·Jul 21

Let’s investigate where India’s AI research is coming from! ICML papers show us nearly a 60-40 split in favour of academia over industry contribution. Interestingly, the single largest entity right now is a corporate, Microsoft. Take a look at the yellow cell. These are independent researchers who publish breakthrough work without any university or corporate backing. They represent one of the most exciting yet overlooked dynamics in the entire ecosystem, and we will be sharing more on why that is soon. As this landscape matures, the real shift will come from domestic momentum. We would love to see greater participation from India-headquartered companies to help drive the next phase of growth!

Post attachment
ExceptionRaised avatar
ExceptionRaised@except_raised·Jul 16

Three insights emerge: • Interpretability, alignment, and AI safety are clear areas of strength for India. This aligns with the grassroots communities we've seen emerge over the past few years, which have helped build a broad talent pipeline into these fields. • Robotics and foundation models are underrepresented in India, likely because they require significant longer-term investments. • Audio & speech and healthcare AI have a disproportionate share of India's research landscape, driving a distinct agenda tailored to local needs If you were investing in India's AI future, where would you place your bets?

ExceptionRaised avatar
ExceptionRaised@except_raised·Jul 16

What is India actually researching in AI? We mapped every ICML 2026 research topic based on two questions: • Is this topic becoming more important globally? • Is India publishing more or less than the world average?

0:10
ExceptionRaised avatar
ExceptionRaised@except_raised·Jul 14

For two hours at #ICML2026 , we discussed just one question: What will it take for India to become a leading AI research ecosystem? Last week, we hosted the first-ever India-focused social at ICML, bringing together 80+ AI researchers from India and the Indian diaspora. Over the coming weeks, we'll be sharing the key insights from the discussion. Thank you to @boldcapfund for sponsoring dinner and making the evening possible.

Post attachment
ExceptionRaised avatar
ExceptionRaised@except_raised·Jun 17

Official India@ICML social in Seoul 🇮🇳 Heading to ICML 2026 and curious about India’s contribution to AI research? Join in for data-driven insights and stories about the ecosystem. Open to anyone in India, living abroad, or simply interested in learning more. Registration link below 👇

ExceptionRaised avatar
ExceptionRaised@except_raised·May 7

His rate of experimentation is like no other! Check out @tokenbender

Ever wondered if you could extract capabilities and behaviors from neural networks and reuse/update/route it as needed? We introduce low-rank circuit conditioning, a novel approach that preserves the model's output behavior while reshaping how an existing capability is represented. In the base model, standard compact recovery stalls at 29%. After conditioning, the same extraction pipeline reaches 91.33% autoregressive full-answer recovery from 5.05% of MLP channels. The evidence points to a possibility of extracting and using isolated capabilities saving cost, latency and high adaptability. Read our work to understand more - tokenbender.com/posts/honey-…

0:12
ExceptionRaised avatar
ExceptionRaised@except_raised·May 7

Amazing what a no-strings-attached grant can do! 🚀

$2,500 is less than the average urban Indian wedding spends on flowers. $2,500 is roughly the price of a second-hand Maruti Alto in India. $2,500 won't even cover a return business-class ticket from Delhi to San Francisco. But this is exactly the grant 26-year-old Kunvar Thaman, an independent researcher from Chandigarh received. Kunvar used it to accomplish a rare feat. He is only the third solo researcher in the world to have his paper accepted at ICML. Our editor @Vaibhav_Rptr interviews @__kunvar__ for AI FrontPage. Read this exclusive interview: aifront-page.com/kunvar-tham…

Post attachment
ExceptionRaised avatar
ExceptionRaised@except_raised·May 7

We’re hosting a talk with @lossfunk tomorrow! @__kunvar__ will walk through his ICML 2026 paper: “Reward Hacking Benchmark: Measuring Exploits in LLM Agents with Tool Use” Register here: luma.com/f92a1jps

Caleb avatar
⟳ ExceptionRaised retweeted
Caleb@caleb_friesen·May 6

Insane. A 26-year-old from Chandigarh just got a paper accepted at ICML. As a solo independent researcher. From India. His name is Kunvar Thaman (@__kunvar__). For context, since ChatGPT launched 3.5 years ago, only two other solo independent researchers have achieved this. Globally. Papers at ICML are typically dominated by big AI labs/institutions, eg OpenAI, Google DeepMind, Stanford, and MIT. The research Kunvar did to pull this off was backed by a $2.5k grant from Exception Raised (@except_raised), an Indian non-profit that funds remarkable Indian AI researchers. Kunvar’s paper is about AI agent reward hacking. He created the Reward Hacking Benchmark (RHB), a sandboxed test environment where advanced AI models are given multi-step tasks using tools like files, code execution, and automated checks. The benchmark measures how honestly the model gets to the right answer.

Post attachment

Yes! my solo-authored paper Reward Hacking Benchmark was accepted to ICML :))) We put LLM agents in a tool-rich sandbox, give them multi-step workflows, and measure when they solve the intended task vs take unexpected shortcuts (like monkeypatching files at runtime!) 1/3

Paras Chopra avatar
⟳ ExceptionRaised retweeted
Paras Chopra@paraschopra·May 4

Congrats, Kunvar! We need many more indie researchers like you from India. Keep us inspired 🇮🇳

ExceptionRaised avatar
ExceptionRaised@except_raised·May 4

Fun fact: Kunvar Thaman got into ICML as a solo independent researcher. Only two others globally have done the same in the last 3 years!

Yes! my solo-authored paper Reward Hacking Benchmark was accepted to ICML :))) We put LLM agents in a tool-rich sandbox, give them multi-step workflows, and measure when they solve the intended task vs take unexpected shortcuts (like monkeypatching files at runtime!) 1/3

ExceptionRaised avatar
ExceptionRaised@except_raised·Feb 9

Use the Frontier List as a guide to spot any new research directions. If you want to discuss or present a paper, join our reading group. Lastly, if there is work you believe belongs here, write to us. We welcome suggestions to expand it!

ExceptionRaised avatar
ExceptionRaised@except_raised·Feb 9

Committee: Prof. Dinesh Babu (HOD DSAI, IIIT-B) and Shapath Das (CEO Unsupervized), Prof. Sumeet Agarwal (Yardi School of AI, IIT-D), Prof. Raghuram Bharadwaj (IIIT-B), Ranita Jana, PhD (CTO Unsupervized), Joseph K J (Research scientist Adobe) and Ahammad Shibil (Speciale Invest)

ExceptionRaised avatar
ExceptionRaised@except_raised·Feb 9

The selection process was committee-driven and deliberately rigorous. Papers were evaluated on technical contribution, research novelty, industry impact and relevance to broader research agenda

ExceptionRaised avatar
ExceptionRaised@except_raised·Feb 9

Introducing 𝐓𝐡𝐞 𝐅𝐫𝐨𝐧𝐭𝐢𝐞𝐫 𝐋𝐢𝐬𝐭, an initiative dedicated to mapping the evolving AI landscape through its most pivotal publications of 2025. We selected 15 papers that, in our view, moved the field forward in concrete and lasting ways. Link: lnkd.in/g_brZrRg

0:09