Building self-improving AI systems

​As large language models are applied to complex domains, the frontier is shifting from models trained once to self-improving systems.

​Amit will decompose self-improvement into three axes: environment generation, RL training, and harness evolution, including verifiers and memory. For the first two, he will present executable counterfactuals, a framework that operationalizes counterfactual reasoning as code, enabling scalable generation of RL environments with controllable difficulty and verifiable answers. RL training on these environments induces the core behaviors of counterfactual reasoning - abduction, intervention, and prediction and, unlike supervised fine-tuning, generalizes out of domain to new code structures and math word problems.

​On the harness side, Amit will present interwhen, which compiles natural-language policies into formal verifiers that certify each step of an agent's reasoning at runtime, raising the pass^4 reliability of a 30B open-weight model on τ²-bench Telecom from 32% to 87%.

​These axes compose into a single self-improvement loop, and present a central research question on how to perform RL jointly with the harness components and self-evolve both the model and its harness.

​Pre read:

• ​interwhen: A Generalizable Framework for Steering Reasoning Models with Test-time Verification

• ​Executable Counterfactuals: Improving LLMs' Causal Reasoning Through Code

​About the speakers:

​Amit Sharma is building an AI lab focused on cybersecurity and self-improving AI systems.

​He spent the past decade at Microsoft Research, where he worked on AI reasoning and causal inference. His work has led to foundational contributions fc causal reasoning, with applications for advancing AI systems' generalization, explainability, and reasoning abilities.

​He developed the DiCE algorithm for counterfactual explanation and refutation methods for evaluating causal estimates, which are widely adopted in both academia and industry.

​Amit is also the co-founder of PyWhy, an open-source ecosystem involving Carnegie Mellon University, Microsoft, Amazon, and others to advance scalable causal ML tools.

​Join online: meet.google.com/onq-amgt-pgz

​Looking forward to seeing you!