Abhay Puri

No, of Course I Can! Deeper Fine-Tuning Attacks That Bypass Token-Level Safety Mechanisms

VectorGym: A Multi-Task Benchmark for SVG Code Generation and Manipulation

Indirect Prompt Injections: Are Firewalls All You Need, or Stronger Benchmarks?

Malice in Agentland: Down the Rabbit Hole of Backdoors in the AI Supply Chain

Rendering-Aware Reinforcement Learning for Vector Graphics Generation

AlignVLM: Bridging Vision and Language Latent Spaces for Multimodal Document Understanding

No, of Course I Can! Deeper Fine-Tuning Attacks That Bypass Token-Level Safety Mechanisms

DoomArena: A framework for Testing AI Agents Against Evolving Security Threats

BigCharts-R1: Enhanced Chart Reasoning with Visual Reinforcement Finetuning

InsightBench: Evaluating Business Analytics Agents Through Multi-Step Insight Generation

BigDocs: An Open Dataset for Training Multimodal Models on Document and Code Tasks