Ananya Salian

ai research in interp, robustness, evals
ai solutions @ affinda
maths and stats @ unimelb
Ananya's catchief of morale
Selected updates
JUL 2026
ShiftBench has been accepted to ICML 2026 CTB & WiML, catch you all in Seoul!
JUN 2026
Joined TARA, an AI Alignment Research Accelerator organised by AI Safety Australia & New Zealand.
JAN 2026
Joined Affinda Group in AI Solutions.
JUN 2025
FaceSafe has been accepted to ICML 2025 DIG-BUG & WiML. See you in Vancouver.
JAN 2025
Joined HEX's Entrepreneurship Accelerator in Singapore as a Kliqtek Scholar - many thanks to Kliqtek.
Selected publications
ShiftBench: certify-or-abstain for cohort PPV under covariate shift
A. Salian · ICML 2026, CTB & WiML Workshops
paper
FaceSafe: privacy-compliant inpainting for scalable image datasets
A. Salian, S. Su, R. You, L. Cui, P. Cui, C. Duong, V. Sharma, S. O'Brien, K. Zhu · ICML 2025, DIG-BUG & WiML
paper