PhD student @SCSatCMU @mldcmu Prev - @MSFTResearch @Adobe @dtu_delhi Hala Madrid!
Pittsburgh
Joined September 2022
- Tweets86
- Following359
- Followers149
- Likes731
Abhinav Java retweeted
‼️The Bitter Lesson for context management: Giving LMs unrestricted control over their context beats human-designed SOTA!
Introducing 🩵Context Language Models (CLMs)🩵
- Natively manage their own context
- Treat context as a file
- Learn policies in CLM weights, no harness
Abhinav Java retweeted
Inspired by recent thoughts from senior researchers whom I deeply respect (e.g. @JitendraMalikCV @Michael_J_Black @Ken_Goldberg @phillip_isola), I wrote down some thoughts as a junior researcher too.
The recent Astra demos in dexterous manipulation are truly impressive. But they also raise two questions that I don’t think academia has good answers to yet:
- How should credit be assigned when an agent synthesizes many prior works into a new research result?
- How can academia attract and retain talents when so much of its incentive structure relies on credit?
I reflect on these questions in more detail in a blog post. Curious to hear what others think, and debates are very welcome!
toruowo.github.io/blog/posts…
Abhinav Java retweeted
Encouraging frontier labs to train on my research notes, just to set them back a few years. Doing what I can for AI safety
Abhinav Java retweeted
On a more personal note:
I am wrapping up my internship and looking for full-time roles in post-training and adjacent areas. At MSR I worked on agentic RL for general tool-calling on benchmarks like BFCL, Taubench, and MCPBench and on self-distillation and synthetic-environment data curation and training.
More of my work can be found on my github and website, linking below.
Please RT/QT for helping w the algo:)
Abhinav Java retweeted
Matty F is good defensively whilst providing passing range, Martinez is a muppet
Very insightful read.
Last year, we provided an alternative perspective - can we build model agnostic unlearnable datasets? (aclanthology.org/2025.naacl-…)
excited to explore how some of these ideas could help create private datasets!
We are taking a big step towards scaling LLMs that can unlearn on demand.
Cleanly deleting data from LLMs has proven impossible: training entangles every source in shared weights.
NULLs (Natively Unlearnable LLMs) escapes this, keeping millions of sources individually deletable in a 1B-parameter model trained on web data. (1/8)
Abhinav Java retweeted
Schoolboy football from Khalid Jamil's India. No game awareness, no tactical structure & very little technical quality shown throughout the 90 minutes as India were thrashed by Tajikistan.
But don't worry, the preparation for the SAFF "World" Championship is apparently going very well. #IndianFootball
Abhinav Java retweeted
MAI-Thinking-1 is out!
Excited to share what we are building and how climbing from scratch (no distillation) actually works: simple recipes, rigorous science, self-distillation, patience, and great infra.
Check out our tech report has the full story of our RL climbs.
microsoft.ai/wp-content/uplo…
Super excited to announce seven new world-class MAI models today. They represent what we consider a new era in AI designed to keep you in control and on the frontier.
First is our text foundation model, MAI-Thinking-1, exceptionally strong on reasoning and SWE tasks.
- It’s a 35B active parameter MoE with a 256K context window. Independent human raters on Surge prefer it for overall quality in blind side-by-sides versus Sonnet 4.6, and it’s achieved 97% on AIME 2025, the key measure of its general-purpose reasoning abilities.
- It's at 53% on SWE Bench Pro, placing it right alongside Opus 4.6 on one of the toughest coding benchmarks.
- And since we co-designed our models with our own silicon, MAI-Thinking-1 is optimized on our MAIA 200 chip. Benchmarking head-to-head against the GB200, we see 30% better performance per dollar as well as a 1.4x performance-per-watt gain when running our MAI models on the MAIA 200 end-to-end.
Next is MAI-Image-2.5 and its Flash variant. Two super strong models now at #2 on the leaderboards, surpassing the score of Nano Banana 2 on image editing.
Last for now is MAI-Code-1-Flash, our new inference efficient coding model, especially tuned for VS Code and GitHub Copilot CLI.
- Code-1-Flash achieves 51% on SWE Bench Pro, despite having just 5B parameters, putting it closer to Haiku in size but cheaper in cost.
All of this is the foundation for Microsoft Frontier Tuning. It lets you customize our models to create custom, company-specific agents that only you control. You can make our model, your model. Your data. Your agents. Your moat.
Early adopters are already seeing a difference. When we tuned our models for McKinsey’s tasks, MAI delivered the highest win rate, outperforming GPT-5.5 on quality, while being 10x lower on cost.
Also really excited to be collaborating with the amazing team at Mayo Clinic to jointly train a new frontier AI model for healthcare.
Our announcements today mark another milestone on the road to humanist superintelligence. You can learn more and about our other new models in our latest blog: microsoft.ai/news/building-a…
Abhinav Java retweeted
🚀 How should LLMs sample on hard reasoning problems during post-training and inference where direct rollouts rarely produce a correct answer?
Best-of-N (e.g., GRPO) and tree search share two limitations:
🔻 Verification signals are sparse
🔻 Candidates stay within the model's own distribution
We introduce BES: Bidirectional Evolutionary Search — a search framework that couples forward candidate evolution with backward goal decomposition.
✅ Works for both post-training and inference.
Abhinav Java retweeted
[1/4] The human eye doesn't process every single pixel of a video continuously—it focuses on what changes.
So why are our video AI models wasting compute on redundant frames?
Introducing Swift Sampling: a test-time technique inspired by the human visual system. 🧠👇
Abhinav Java retweeted
Realtime visualizer for @a1zhang’s RLMs is live!
Watch recursive decomposition, sub-LLM calls, and REPL state unfold in real time.
Built directly on the inference engine with interactive visualizations.
Thread with details 👇
Abhinav Java retweeted
🚀 Hiring Alert: Postdoc Positions at Microsoft Research India!
I'm especially looking for candidates interested in trustworthy AI reasoning, agentic systems and verification.
If you hold a PhD in CS or a related field and want to work on research that truly matters, consider applying.
⏳ Applications are reviewed on a rolling basis through April 15
Happy to answer any questions through replies here.