My efforts in the area of AI SEA (Safety, Ethics and Alignment) have broadly been in these two areas:
Red-Teaming + Adversarial attacks
and
Dataset auditing + bias mitigation
and
Model auditing.
A sampling of my research publications spanning these two areas : Model auditing, Red-Teaming and Adversarial attacks
Model auditing, Red-Teaming and Adversarial attacks
Beyond the imitation game: Quantifying and extrapolating the capabilities of language models
Auditing saliency cropping algorithms
.
Project page
[
Video
]
Model weight theft with just noise inputs: The curious case of the petulant attacker
Understanding adversarial robustness through loss landscape geometries
Vulnerability of deep learning-based gait biometric recognition to adversarial perturbations
On Lyapunov exponents and adversarial perturbation
On detecting adversarial inputs with entropy of saliency maps
On grey-box adversarial attacks and transfer learning
Art-attack! on style transfers with textures, label categories and adversarial examples
SIMUni: Sampling Impostors from Misfit Universal Background Models in accelerometric gait biometric verification
Smile in the face of adversity much? A print based spoofing attack
OODles of ODDs: The landscape of Out-of-distribution vulnerabilities of vision models
Did They Direct the Violence or Admonish It? A Cautionary Tale on Contronomy, Androcentrism and Back-Translation Foibles
[
Video
]
Dataset curation, auditing and bias mitigation
Multimodal datasets: misogyny, pornography, and malignant stereotypes
Large image datasets: A pyrrhic win for computer vision?
Into the laion’s den: Investigating hate in multimodal datasets
On Hate Scaling Laws For Data-Swamps
The blood diamond effect in neural art: On ethically troublesome images of the imagenet dataset
[
Video
]
Covering up bias in CelebA-like datasets with Markov blankets: A post-hoc cure for attribute prior avoidance
JFT-300M: The mystery box
PS: My more recent AI safety chronicles can be found
here
JFT 300M
Translation foibles
Saliency Image Cropping audits
Nueral art and blood diamond effect

Matrix Rain with Blue Neon Text "Those who ride the tiger can never dismount."