Arjun Guha
Showing 1–2 of 2 transcripts.
- Jane Street1h 0m
Arjun Guha: How Language Models Model Programming Languages & How Programmers Model Language Models
Arjun Guha presents a comprehensive analysis of large language models in programming, highlighting how traditional benchmarks are saturating while new methods like multi-PLE and language-agnostic transforms reveal significant performance gaps in low-resource languages such as OCaml. Through mechanistic interpretability techniques like activation steering, the talk demonstrates that internal model vectors can effectively correct type prediction errors and switch target languages without retraining, exposing shared representations across diverse syntaxes. These technical insights are contextualized by human studies showing that student success in prompting models hinges on providing specific semantic clues rather than syntactic fixes, while industry data reveals a surge in AI co-authorship alongside complex debates regarding actual productivity gains.
- Jane Street59 min
Arjun Guha: On Verification for System Configurations Languages
Researchers developed "Rehearsal," a verification tool designed to detect determinism and idempotency violations in Puppet configuration manifests, which are frequent causes of major system outages. The system employs a transformation pipeline that converts complex Puppet syntax into Datalog and a low-level modeling language, leveraging the Z3 SMT solver alongside optimizations like partial order reduction and state pruning to analyze real-world code efficiently. Benchmarks on GitHub-scraped manifests demonstrate that this approach successfully identifies previously unknown non-deterministic bugs and generates fixes, although current limitations exclude shell script execution and specific file permission modeling.