Shubham Bansal

Soumith Chintala, 𝗰𝗿𝗲𝗮𝘁𝗼𝗿 of PyTorch and one of the most respected builders in 𝗺𝗼𝗱𝗲𝗿𝗻 𝗔𝗜, has joined Mira Murati ’s fast-rising startup, Thinking Machines Lab. After 11 years at Meta (Facebook), where PyTorch grew from a 𝗻𝗶𝗰𝗵𝗲 𝗽𝗿𝗼𝗷𝗲𝗰𝘁 𝘁𝗼 𝗽𝗼𝘄𝗲𝗿𝗶𝗻𝗴 𝗴𝗹𝗼𝗯𝗮𝗹 𝗔𝗜 𝗶𝗻𝗻𝗼𝘃𝗮𝘁𝗶𝗼𝗻, Chintala’s move signals a new wave in human–𝗔𝗜 𝗰𝗼𝗹𝗹𝗮𝗯𝗼𝗿𝗮𝘁𝗶𝗼𝗻. Thinking Machines Lab isn’t just gathering top talent from Meta, OpenAI, Anthropic, and leading universities—they’re offering 𝗴𝗮𝗺𝗲-𝗰𝗵𝗮𝗻𝗴𝗶𝗻𝗴 𝘀𝗮𝗹𝗮𝗿𝗶𝗲𝘀 (up to $500K), 𝗮 𝗺𝗮𝘀𝘀𝗶𝘃𝗲 $𝟮 𝗯𝗶𝗹𝗹𝗶𝗼𝗻 𝘀𝗲𝗲𝗱 𝗿𝗼𝘂𝗻𝗱, and are aiming for a $50 billion market valuation. Their new product, Tinker, is already in use at Princeton, Stanford, and by major enterprises. Meta, meanwhile, is regrouping under the 𝗻𝗲𝘄 𝗦𝘂𝗽𝗲𝗿𝗶𝗻𝘁𝗲𝗹𝗹𝗶𝗴𝗲𝗻𝗰𝗲 𝗟𝗮𝗯𝘀 division led by Alexandr Wang, with other AI leaders reportedly on the move. Chintala’s reasoning: time to pursue “𝙨𝙤𝙢𝙚𝙩𝙝𝙞𝙣𝙜 𝙨𝙢𝙖𝙡𝙡, 𝙨𝙤𝙢𝙚𝙩𝙝𝙞𝙣𝙜 𝙣𝙚𝙬, 𝙨𝙤𝙢𝙚𝙩𝙝𝙞𝙣𝙜 𝙪𝙣𝙘𝙤𝙢𝙛𝙤𝙧𝙩𝙖𝙗𝙡𝙚.” His legacy—the PyTorch tool now thriving in classrooms from 𝗠𝗜𝗧 𝘁𝗼 𝗿𝘂𝗿𝗮𝗹 𝗜𝗻𝗱𝗶𝗮—𝗶𝘀 𝘀𝗲𝗰𝘂𝗿𝗲. #AI #TechNews #PyTorch #Meta #ThinkingMachines #Startup #AIJobs #TalentWar #Innovation #BigFunding

8

🤖RLAD Training LLMs to Discover Abstractions for Solving Reasoning Problems. 🔍 Motivation: How do we move large language models (LLMs) beyond pattern matching—toward algorithmic, abstract reasoning on hard problems? Most current RL post-training approaches create longer chains of thought, but they often drift into verbose or degenerate exploration, missing true reasoning depth. 📄 Solution: RLAD introduces 'reasoning abstractions'—concise, natural language descriptions of procedural and factual knowledge that guide models in learning successful reasoning strategies. The approach trains two LLMs: one to generate abstractions and another to solve problems using those abstractions, via a novel RL setup. Crucially, abstractions act as high-level subgoals or skills, helping models explore diverse solution paths and generalize to harder tasks. 🔑 Results: RLAD achieves up to 44% improvement on math reasoning benchmarks (AIME 2025, DeepScaleR Hard, AMC 2023) over state-of-the-art long chain-of-thought RL methods. Moreover, the benefits scale with more compute allocated to abstraction generation versus solution sampling—a key insight for practical deployment. 💡 Implications: Abstraction-driven training broadens the reasoning abilities of LLMs, unlocking new horizons for model interpretability, strategy diversity, and generalization across complex domains. This direction is especially relevant as scaling chain-of-thought length yields diminishing returns. (Attached: Original research PDF)

13

1 Comment

📝 Announcing our EMNLP 2025 paper that introduces the Alignment Quality Index (AQI), a metric leveraging latent geometric representations and clustering indices to diagnose hidden misalignments in LLMs. 🔹 “𝐀𝐥𝐢𝐠𝐧𝐦𝐞𝐧𝐭 𝐐𝐮𝐚𝐥𝐢𝐭𝐲 𝐈𝐧𝐝𝐞𝐱 (𝐀𝐐𝐈) – 𝐁𝐞𝐲𝐨𝐧𝐝 𝐑𝐞𝐟𝐮𝐬𝐚𝐥𝐬: 𝐀𝐐𝐈 𝐚𝐬 𝐚𝐧 𝐈𝐧𝐭𝐫𝐢𝐧𝐬𝐢𝐜 𝐀𝐥𝐢𝐠𝐧𝐦𝐞𝐧𝐭 𝐃𝐢𝐚𝐠𝐧𝐨𝐬𝐭𝐢𝐜 𝐯𝐢𝐚 𝐋𝐚𝐭𝐞𝐧𝐭 𝐆𝐞𝐨𝐦𝐞𝐭𝐫𝐲, 𝐂𝐥𝐮𝐬𝐭𝐞𝐫 𝐃𝐢𝐯𝐞𝐫𝐠𝐞𝐧𝐜𝐞, 𝐚𝐧𝐝 𝐋𝐚𝐲𝐞𝐫 𝐰𝐢𝐬𝐞 𝐏𝐨𝐨𝐥𝐞𝐝 𝐑𝐞𝐩𝐫𝐞𝐬𝐞𝐧𝐭𝐚𝐭𝐢𝐨𝐧𝐬” 🔹 In collaboration with Manipal University Jaipur, LinkedIn, Indian Institute of Technology, Kharagpur, Indian Institute of Information Technology Guwahati, Texas A&M University, VIT Chennai, Evalueserve, New York University, Birla Institute of Technology and Science, Pilani (Pragya Lab) 🔹 Published in EMNLP 2025 Main 🔹 Paper: https://lnkd.in/g54bEBCu ✍🏼 Authors: Abhilekh Borah, Chhavi Sharma, Danush Khanna, Utkarsh Bhatt, GURPREET SINGH, Hasnat Md. Abdullah, Raghav Kaushik R, Vinija Jain, Jyoti Patel, Shubham Singh, Vasu Sharma, Arpita V., Rahul Raja, Aman Chadha, Dr. Amitava Das ➡️ 𝐊𝐞𝐲 𝐇𝐢𝐠𝐡𝐥𝐢𝐠𝐡𝐭𝐬 𝐨𝐟 𝐨𝐮𝐫 𝐀𝐥𝐢𝐠𝐧𝐦𝐞𝐧𝐭 𝐐𝐮𝐚𝐥𝐢𝐭𝐲 𝐈𝐧𝐝𝐞𝐱 (𝐀𝐐𝐈): 🧬 𝑰𝒏𝒕𝒓𝒊𝒏𝒔𝒊𝒄 𝑳𝒂𝒕𝒆𝒏𝒕 𝑮𝒆𝒐𝒎𝒆𝒕𝒓𝒚 𝑴𝒆𝒕𝒓𝒊𝒄: AQI measures alignment by assessing how distinctly safe vs. unsafe prompts cluster in latent activations, using combination of Davies‑Bouldin, Dunn, Xie‑Beni, and Calinski‑Harabasz indices—making it invariant to decoding style and resilient to alignment faking. 🧠 𝑳𝒂𝒚𝒆𝒓𝒘𝒊𝒔𝒆 𝑷𝒐𝒐𝒍𝒆𝒅 𝑹𝒆𝒑𝒓𝒆𝒔𝒆𝒏𝒕𝒂𝒕𝒊𝒐𝒏 𝑳𝒆𝒂𝒓𝒏𝒊𝒏𝒈: we use a sparse learned pooling mechanism across transformer layers to capture alignment‑relevant abstractions without modifying the base model—enabling internal safety diagnostics. 🚨 𝑬𝒎𝒑𝒊𝒓𝒊𝒄𝒂𝒍 𝑭𝒂𝒊𝒍𝒖𝒓𝒆𝒔 𝒐𝒇 𝑩𝒆𝒉𝒂𝒗𝒊𝒐𝒓𝒂𝒍 𝑴𝒆𝒕𝒓𝒊𝒄𝒔: AQI uncovers misalignments that behavioral metrics like G‑Eval or refusal rates miss—especially under jailbreaks, safety‑agnostic fine‑tuning, and stochastic decoding, proving its value as an early‑warning alignment audit tool.

117

2 Comments

Quantization in Large Language Models (LLMs) As LLMs scale, intelligence is no longer the bottleneck — memory, latency, and cost are. Quantization turns out to be one of the most powerful (and misunderstood) techniques enabling real-world deployment of large models. In this blog, I break down: • Why LLMs are surprisingly quantization-friendly • INT8 vs INT4 vs NF4 — what actually works in practice • How modern architectures absorb low-bit noise • Where quantization fails and how to design around it Quantization is no longer an optimization step — it’s foundational infrastructure for production LLM systems. Link: https://lnkd.in/gqGje75G #LLMs #AIEngineering #Quantization #MachineLearning #GenAI #MLOps

9

2 Comments

Towards building Kiro and agentic tools for code, we have been exploring the evolution of reinforcement learning for LLM agents, and there's a clean conceptual arc worth sharing. Starting from REINFORCE — the foundational policy gradient algorithm — each successive method has solved a specific failure mode of its predecessor. PPO added training stability via clipping and a learned critic. GRPO then eliminated the critic entirely by normalizing rewards within a group of sampled responses, making RL practical for LLMs at scale. I am yet to see matured science to tackle the exploration bottleneck that GRPO inherits: when all sampled responses are poor, there's no useful gradient signal. There is some recent work emerging trying to solve this using memory-assisted rollouts to discover better strategies, and then distilling those discoveries back into model weights via a hybrid on/off-policy objective — essentially reward-guided knowledge distillation, but a lot more work needs to happen in this space. Now, in my opinion, the natural next frontier seems to be "world models" for LLM agents (coding agents for our context). Current systems like GRPO are still reactive — they learn purely from real environment interactions with no ability to simulate consequences before acting. A world model changes this: the agent can plan by rolling out imagined trajectories internally, evaluate candidate actions before committing, and use per-step value estimates for richer credit assignment. For coding agents specifically, a layered world model makes sense — an execution layer predicting test outcomes, a static analysis layer predicting ripple effects of edits, and a task progress layer estimating how close the agent is to its goal. Credit assignment is in fact the deeper unsolved problem underlying all of this. Current LLM training assigns the same reward signal to every token in a trajectory, regardless of which actions actually caused the outcome. For long coding tasks spanning hundreds of steps, this is deeply inefficient — the agent reinforces irrelevant actions alongside critical ones. The right solution could combine world model value estimates, process reward models for intermediate step evaluation, and contrastive methods to isolate causal contribution. Together these would give the policy a rich, causally grounded credit signal rather than a flat scalar. Zooming out, a state-of-the-art coding agent would integrate five systems in my opinion: a hierarchical policy, a world model, memory (episodic, semantic, codebase-specific), a credit assigner, and a planner that uses MCTS-style search in language space to evaluate approaches before execution. Of these, only the policy exists in mature form today. A lot of science work needs to happen for the rest. If you are passionate about this space, I want to hear your thoughts and especially for students doing their PhDs in this area, I would also like to invite you for an internship..

200

5 Comments

An excellent observation from Ram Swaminathan. When we concatenate embeddings (e.g., text + image), cosine in the joint vector becomes a norm-weighted mix of the per-block cosines, so a misaligned or dominant subvector can drive the score and distort similarity.

7

A Thought Exercise : Context Engineering with LLMs is analogous to Feature Engineering in Classical ML What are the similatieties and differences ? Share in the thread #AImusings courtesy: Jim Hong :)

28

9 Comments

New paper alert! Augmenting Large Language Models with Psychologically Grounded Models of Causal Reasoning for Planning Under Uncertainty Authors: Semanti Basu, PhD, Moon Hwan Kim, Semir Tatlidil, Tom Williams, Steven Sloman, R. Iris Bahar Venue: Frontiers in Artificial Intelligence Abstract: Large Language Models (LLMs) have come a long way in their ability to solve a wide range of problems. Yet, LLM decision making still relies primarily on pattern recognition, which may limit its ability to make sound decisions under uncertainty. In contrast, human reasoning often makes use of explicit causal models, allowing humans to explain, hypothesize, and extrapolate to different domains in uncertain scenarios. In this article, we explore whether human causal models can be strategically integrated with Large Language Models to improve planning outcomes under uncertainty for object assembly and troubleshooting tasks modeled as Partially Observable Markov Decision Processes (POMDPs). Our contributions consist of two parts: (1) an interactive LLM agent that plans an action at each time step by solving a POMDP targeted at an object assembly or troubleshooting task, and (2) a novel hybrid-reasoning framework that uses confidence scores in both the LLM agent’s output and a human causal model to make a final decision on the most appropriate action for the current time step to achieve the task. We demonstrate the efficacy of our approach through detailed simulations and show a significant improvement in task planning reward across three different state-of-the-art LLMs when augmenting the baseline LLM planner with a human causal model. Read it here: https://lnkd.in/gBt8dkHQ

25

We published a survey kind of paper on "From Attention to Disaggregation: Tracing the Evolution of LLM Inference".  It will be invaluable for anyone interested in learning more about LLM Inference, its comparison with the CAP theorem (a concept familiar to those who have worked with distributed systems), and the state-of-the-art research in Generative AI inference https://lnkd.in/gjVe5P8J

77

🚀 ML Project to Practice 📊 Student Performance Prediction Where to Practice (Run the Project): 📚Google Colab Notebook: https://lnkd.in/gzaViEh4 ❗ Problem: There are many reasons students might perform poorly in their studies, and perhaps we can prevent it. 🤔 🎯 Goal: Predict if a student might perform badly and help them achieve better scores. 📈 🛠️ Approach: Use Linear Regression and evaluate how good our model is using Loss Functions. 📊 -------------------------- 🚀 Master Machine Learning 🔬 A Physics-Based Approach 💼 Get Hired or Promoted 📈 👉 Here’s your roadmap: 1. 🔗 Visit the program page: https://lnkd.in/gWwq78Vp 2. 🎬 Start from Episode 1 and move through the episodes step by step 👣. 1. 🔹 Read the Research Notes and watch the embedded videos to deeply understand each concept. 2. 🔹 Take the Course. Open the corresponding Colab notebook, run the code from top to bottom, and see how these physics-based ideas map onto conventional ML algorithms. 3. 🔹 Run the Project. Apply your knowledge on real datasets, solve real-world problems, and gain hands-on experience that bridges theory and practice. 4. 🌟 Ready to go even further? ✨ Turn the project into a unique-to-you professional GitHub repository — with a clean, industry-standard folder structure, production-ready settings, and depoly as a web application to showcase your work live. 📩 Click Request Professional Setup on the page to get started. 🚀 FREE-or-Land-Job 6 Months BootCamp From Graduation in a Math-Based Field to a Data Science Career ✨ Program Highlights 💼 Land a Job or Pay Almost Nothing! 🧭 Step-by-Step Roadmap: 100% done-for-you structured schedule outlining exactly what to learn at each stage. 🧑‍🏫 Weekly Live Sessions: Expert-led project-solving sessions to master core data science skills through hands-on experience. 🌐 Build Your Online Presence & Portfolio: Work on real-world projects, receive expert reviews, and showcase your work on GitHub. 🧠 Technical & Soft Skills: Master all skills needed to land a job, from machine learning to interview and resume prep. 📄 Job Application Prep & Mentoring: Step-by-step guide on job search strategies, networking, and application materials. 🤝 Get Support When Needed: Connect with mentors for help, ask questions anytime, and join group discussions to learn and grow together. 🎯 Apply now: 🔗 https://lnkd.in/gDv64aFg

5