5 papers
What the Guard Misses, the Robot Executes: Implied Harm in VLA Instructions
Sripad Karne, Arjun Balaji
Vision-language-action models (VLAs) act on instructions without being able to refuse, so screening harmful requests falls to monitors. We test whether these monitors catch ordinar…
Safety Monitors Mostly Catch What the Model Already Refuses
Sripad Karne
Safety monitors are evaluated by recall on harmful prompts, regardless of whether the target model would answer them. Yet a monitor matters most on the prompts the model does answe…
How Far Do Auto-Interpretation Labels Generalize: A Controlled Study Across Languages, Scripts, and Rewordings
Sripad Karne
Sparse autoencoder (SAE) features are increasingly used to interpret language models, with auto-generated natural-language labels serving as the primary interface for understanding…
AgentOpt v0.1 Technical Report: Client-Side Optimization for LLM-Based Agent
Wenyue Hua, Sripad Karne, Qian Xie +4
AI agents are increasingly deployed in real-world applications, including systems such as Manus, OpenClaw, and coding agents. Existing research has primarily focused on server-side…
One Language, Two Scripts: Probing Script-Invariance in LLM Concept Representations
Sripad Karne
Do the features learned by Sparse Autoencoders (SAEs) represent abstract meaning, or are they tied to how text is written? We investigate this question using Serbian digraphia as a…