6 papers
Answering the Wrong Question: Reasoning Trace Inversion for Abstention in LLMs
Abinitha Gourabathina, Inkit Padhi, Manish Nagireddy +2
For Large Language Models (LLMs) to be reliably deployed, models must effectively know when not to answer: abstain. Reasoning models, in particular, have gained attention for impre…
Final-Model-Only Data Attribution with a Unifying View of Gradient-Based Methods
Dennis Wei, Inkit Padhi, Soumya Ghosh +3
Training data attribution (TDA) is concerned with understanding model behavior in terms of the training data. This paper draws attention to the common setting where one has access…
Building a Foundational Guardrail for General Agentic Systems via Synthetic Data
Yue Huang, Hang Hua, Yujun Zhou +11
While LLM agents can plan multi-step tasks, intervening at the planning stage-before any action is executed-is often the safest way to prevent harm, since certain risks can lead to…
When in Doubt, Cascade: Towards Building Efficient and Capable Guardrails
Manish Nagireddy, Inkit Padhi, Soumya Ghosh +1
Large language models (LLMs) have convincing performance in a variety of downstream tasks. However, these systems are prone to generating undesirable outputs such as harmful and bi…
Programming Refusal with Conditional Activation Steering
Bruce W. Lee, Inkit Padhi, Karthikeyan Natesan Ramamurthy +4
LLMs have shown remarkable capabilities, but precisely controlling their response behavior remains challenging. Existing activation steering methods alter LLM behavior indiscrimina…
Granite Guardian
Inkit Padhi, Manish Nagireddy, Giandomenico Cornacchia +20
We introduce the Granite Guardian models, a suite of safeguards designed to provide risk detection for prompts and responses, enabling safe and responsible use in combination with…