6 papers
InfoSFT: Learn More and Forget Less with Information-Aware Token Weighting
Mahdi Sabbaghi, George Pappas, Adel Javanmard +1
Supervised fine-tuning (SFT) provides the standard approach for teaching LLMs new behaviors from offline expert demonstrations. However, standard SFT uniformly fits all samples --…
Robust Policy Optimization to Prevent Catastrophic Forgetting
Mahdi Sabbaghi, George Pappas, Adel Javanmard +1
Large language models are commonly trained through multi-stage post-training: first via RLHF, then fine-tuned for other downstream objectives. Yet even small downstream updates can…
Benchmarking Misuse Mitigation Against Covert Adversaries
Davis Brown, Mahdi Sabbaghi, Luze Sun +4
Existing language model safety evaluations focus on overt attacks and low-stakes tasks. In reality, an attacker can easily subvert existing safeguards by requesting help on small,…
Llama-3.1-FoundationAI-SecurityLLM-Reasoning-8B Technical Report
Zhuoran Yang, Ed Li, Jianliang He +18
We present Foundation-Sec-8B-Reasoning, the first open-source native reasoning model for cybersecurity. Built upon our previously released Foundation-Sec-8B base model (derived fro…
Adversarial Reasoning at Jailbreaking Time
Mahdi Sabbaghi, Paul Kassianik, George Pappas +3
As large language models (LLMs) are becoming more capable and widespread, the study of their failure cases is becoming increasingly important. Recent advances in standardizing, mea…
Explicitly Encoding Structural Symmetry is Key to Length Generalization in Arithmetic Tasks
Mahdi Sabbaghi, George Pappas, Hamed Hassani +1
Despite the success of Transformers on language understanding, code generation, and logical reasoning, they still fail to generalize over length on basic arithmetic tasks such as a…