Software engineer turned AI safety researcher. I study ways to reliably measure alignment and safety in frontier LLMs.
Google Scholar · Twitter · LinkedIn · GitHub · CV
ram.bharadwaj.arya [at] gmail [dot] com
Tracing Eval-Awareness Emergence Through Training of OLMo 3
Ram Bharadwaj, Robert Kirk
Alignment Forum, 2026
EvalDetectBench: A Benchmark for Measuring Evaluation Awareness in Frontier Language Models
Xinning Li*, Kemunto Ochwang'i*, Ram Bharadwaj*, Alexandra Souly, Robert Kirk
NeurIPS 2026 Evaluations & Datasets Track (Poster)
Understanding Hidden Computations in Chain-of-Thought Reasoning
Ram Bharadwaj
arXiv, 2024 · LessWrong post
Experiments with the Platonic Representation Hypothesis
Ram Bharadwaj
LessWrong, 2024
The AGI Needs to Be Honest
Ram Bharadwaj
LessWrong, 2021