Varun Prashant Gangal
Member of Technical Staff, Research
Varun is a research scientist at Patronus AI who has worked on SpeedrunBench, MEMTRACK, TRAIL and HalluWorld. His current interests span LLM evaluation, model behavior and alignment. He is a coauthor of A Survey of Data Augmentation Approaches for NLP and a contributor to Humanity's Last Exam. He finished his PhD at the Language Technologies Institute at CMU, advised by Prof. Ed Hovy.
At Patronus· Newest first
Newest first
Papers
SpeedrunBench
Video game speedrunning as a test of whether agents can beat strategies humans already know.
HalluWorld
A controlled benchmark that measures hallucination against reference world models.
Unified Hallucination Framework
One definition of hallucination: inaccurate world modeling.
Generative Simulators
Environments that generate their own tasks, world dynamics and rewards.
MEMTRACK
Testing whether agents keep track of facts spread across workplace tools.
TRAIL
A benchmark for finding the step where an agent went wrong.
Posts
Introducing SpeedrunBench: A Challenging Benchmark for Frontier Agents
Getting GLM-5.2 NVFP4 Post-Training Off the Ground
Introducing Generative Simulators: Autonomously Scaling Environments for Agents
No papers or posts here yet.