Mandar Joshi

🚨 I’m on the 2026 Research Scientist Job Market! I am a PhD student at UNC Chapel Hill (advised by @mohitban47) and recipient of the Apple Scholars in AI/ML PhD Fellowship. My research centers around: 🔸Reasoning & RL/Post-Training: Evaluating and interpreting the reasoning process, and improving post-training and alignment through self-generated and reward-based signals (Intrinsic Dim., ReCEVAL, ScPO, LASeR). 🔸Agents & Planning: Designing adaptive agent frameworks to that use extra test-time compute & reasoning upon failure (ADaPT, System-1.x, PRInTS). 🔸Reward & Skill Discovery in Code: Leveraging execution signals to build reliable rewards, automate debugging, and discover abstractions in code (UTGen, ReGAL). Prev (Research Intern): Google DeepMind, Meta FAIR, Allen Institute for AI (AI2), and Adobe Research. Feel free to reach out via DM or email if you’re interested, have leads, or would like to connect! 🌐 archiki.github.io 📧 archiki@cs.unc.edu #NLP #AI #JobSearch

Think the GoI has done really well on its part with funding Sarvam and supporting it through initial risk phase to get Sarvam to where they are. Now with their potential clearly visible, can we see industry veterans like Adani(who eloquently spoke of need for sovereign AI being a necessary strategic capability) or Infosys/TCS to come forward and invest in likes of Sarvam in Billions? Maybe with credit hours on his GPU hyperscalers etc? Sarvam is no more “high risk” in the Indian big biz owner mindset sense. The potential and the market are clearly visible.

🚨Excited to share our new work viewing reasoning strategies as teaching tools: for fixed target model, which CoT strategies best support learning and generalization? ✨Our answer is intrinsic dimensionality (minimum effective capacity a model needs to solve the task). Somewhat counterintuitively, adding CoT – which requires generating longer and more structured outputs – can reduce learning complexity. Good reasoning compresses the task, i.e., it reduces the degrees of freedom the model needs to map inputs to correct solutions. 🧵⬇️ (1/5)

Fara-7B is our first agentic small language model for computer use. This experimental model includes robust safety measures to aid responsible deployment. Despite its size, Fara-7B holds its own against larger, more resource-intensive agentic systems: msft.it/6015tpZHF

🌻 Announcing New Agents + Economics Research from Microsoft! AI agents are starting to shop and buy for us. At the same time, agents are representing and providing customer support on behalf of businesses. We believe that these two sides will soon collide, and... 1/n

We found that BLT-style dynamic patching achieves better text --> speech transfer for speech-text LLMs. Keeping data fixed, yields much better performance + 20% compute savings + more. Huge shoutout to @Yen_Ju_Lu for all his hard work on this. Read all about it in our pre-print:

Excited to share a new paper that aims to narrow the conceptual gap between the idealized notion of Kolmogorov complexity and practical complexity measures for neural networks.

We're hiring a research scientist on the Foundational Research in Language team at GDM. The role is right here in sunny Seattle! job-boards.greenhouse.io/deepmind/jobs/…

Breaking | The largest Hindu temple in California, @BAPS_PubAffairs temple in Chino Hills, was vandalized with profanities earlier today. We ask @ChinoHills_PD, @FBI @FBIDirectorKash @DNIGabbard to investigate this latest in a string of anti-Hindu hate crimes on our sacred spaces. Video of the attack was shared by bot accounts similar to previous attacks.

Wohoo! @bansalg_

@ShikharMurty "This is like working on NLP, but knowing nothing about linguistics." Careful now, before you start a war on Twitter :)

Honest admission: I work on web-agents but know surprisingly less about the web. This is like working on NLP, but knowing nothing about linguistics. Any good resources to learn the core technologies of the WWW?

New paper! Byte-Level models are finally competitive with tokenizer-based models with better inference efficiency and robustness! Dynamic patching is the answer! Read all about it here: dl.fbaipublicfiles.com/blt/BLT__Patch… (1/n)

Super excited to share NNetnav : A new method for generating complex demonstrations to train web agents—driven entirely via exploration! Here's how we’re building useful browser agents, without expensive human supervision: 🧵👇 Code: github.com/MurtyShikhar/N… Preprint: arxiv.org/abs/2410.02907

To our collaborators & community: We’ve seen questions about AutoGen forks/clones vs. the official project. Here's a summary of the latest. Please share with others. - The official repo is github.com/microsoft/auto…. - We're actively working on AutoGen v0.2, with v0.4 innovations like AutoGen-Core, -AgentChat, -Bench, and -Studio 🚀 - Development remains open-source under MIT, and contributions are welcome. - No Microsoft employee is affiliated with AG2, a fork of Microsoft/AutoGen; We recognize varying endeavors/interests in the community and wish them well. ⚠️ Please note: The current pyautogen package isn’t affiliated with Microsoft AutoGen, and admin access is blocked for us. Discord access is also limited for us -- we do not have viewing or posting rights, but you can connect via GitHub Discussions, email (autogen@microsoft.com), or our Wed office hours (10 AM PST). Thank you for shaping this collaborative project! Stay tuned for more updates soon. 🙌 Read the article below for full details 👇 #AutoGen #OpenSourceAgents #OpenSource

Excited to finally release Magentic-One! The thing I love about this multi-agent team is that the same implementation achieves very strong performance across three challenging agentic benchmarks. If you are someone working on agentic systems, you know how challenging this can be. We had to figure out a set of capabilities and their implementations that truly generalize. Think planning, keeping track of progress, action and observation spaces, error recovery, etc Super excited to release this to open-source and allow others in academia, industry, and open-source community to build off Magentic-One! Please checkout the tech report and code in the announcement below and let us know how it goes!

Excited to share a new paper: “ALTA: Compiler-Based Analysis of Transformers” (w/ @James_Cohan, @jacobeisenstein, @kentonctlee, @JonathanBerant, @toutanova) arxiv.org/abs/2410.18077

@hubermanlab And what's the wrong time to exercise (that will result in a hard crash)?

A core truth of circadian biology is that if you exercise at the right time in your circadian (temperature) cycle, you’ll have more energy all day. Exercise at the wrong time and you’re likely to crash hard later. For most people, exercising 30-60min after waking = more energy.

As Bangladesh prepares to swear in its interim government, I urge all government officials, the new administration and police chief, and the people of Bangladesh to do all they can to end the violence that has emerged across the country, including the brutal targeting of the country’s Hindu minority, their homes, businesses, and their temples. The violence must stop and those responsible must be brought to justice to help the people of Bangladesh move forward as a nation. I will continue to closely monitor developments in Bangladesh in coordination with the U.S. State Department.

@universeinanegg So sorry for your loss, Ari.

Some sad news: My father died unexpectedly a bit over a week ago. We were close. Apologies for my delayed response times, which will continue for a while—there's so much to take care of, in addition to having just moved to a new city for a new job, and needing time to grieve.

The @SFRCdems chair released a statement today omitting any mention of Hindus being targeted & their temples ransacked. Also neglects reports that Bangladeshi Hindus are at the Indian border seeking refuge under the very same #CAA he condemned. Do better. foreign.senate.gov/press/dem/rele…