3 papers
cs.CL2026
Gradient-Controlled Decoding: A Safety Guardrail for LLMs with Dual-Anchor Steering
Purva Chiniya, Kevin Scaria, Sagar Chaturvedi
Large language models (LLMs) remain susceptible to jailbreak and direct prompt-injection attacks, yet the strongest defensive filters frequently over-refuse benign queries and degr…
cs.CR2026
SecRepoBench: Benchmarking Code Agents for Secure Code Completion in Real-World Repositories
Chihao Shen, Connor Dilgren, Purva Chiniya +3
This paper introduces SecRepoBench, a benchmark to evaluate code agents on secure code completion in real-world repositories. SecRepoBench has 318 code completion tasks in 27 C/C++…
cs.LG2025
Open Deep Search: Democratizing Search with Open-source Reasoning Agents
Salaheddin Alzubi, Creston Brooks, Purva Chiniya +9
We introduce Open Deep Search (ODS) to close the increasing gap between the proprietary search AI solutions, such as Perplexity's Sonar Reasoning Pro and OpenAI's GPT-4o Search Pre…