Related papers: SCAFFOLD-CEGIS: Preventing Latent Security Degrada…

Large Language Model-based Multi-Agent Systems (LLM-MAS) are increasingly applied to complex collaborative scenarios. However, their collaborative mechanisms may cause minor inaccuracies to gradually solidify into system-level false…

Safety post-training can improve the harmfulness and policy compliance of Large Language Models (LLMs), but it may also reduce general utility, a phenomenon often described as the \emph{alignment tax}. We study this trade-off through the…

Drug development and pharmacovigilance are frequently bottlenecked by legacy clinical reporting pipelines. These monolithic systems encode regulatory-grade logic but resist AI integration by producing opaque output with no machine-readable…

Recent secure code generation methods, using vulnerability-aware fine-tuning, prefix-tuning, and prompt optimization, claim to prevent LLMs from producing insecure code. However, their robustness under adversarial conditions remains…

Autonomous AI agents are being deployed with filesystem access, email control, and multi-step planning. This thesis contributes to four open problems in AI safety: understanding dangerous internal computations, removing dangerous behaviors…

Large Language Models (LLMs) deployed in production environments face a fundamental safety-utility trade-off either a strict filtering mechanisms prevent harmful outputs but often block benign queries or a relaxed controls risk unsafe…

Learned classifiers deployed in agentic pipelines face a fundamental reliability problem: predictions are probabilistic inferences, not verified conclusions, and acting on them without grounding in observable evidence leads to compounding…

Background: Security regressions are vulnerabilities introduced in a previously unaffected software system. They often happen as a result of source code changes (e.g., a bug fix) and can have severe effects. Aims: To increase the…

The rapid advancement of multimodal large language models (MLLMs) has led to breakthroughs in various applications, yet their security remains a critical challenge. One pressing issue involves unsafe image-query pairs--jailbreak inputs…

Large Language Models (LLMs) have shown impressive proficiency in code generation. Unfortunately, these models share a weakness with their human counterparts: producing code that inadvertently has security vulnerabilities. These…

Achieving robust safety alignment in large language models (LLMs) while preserving their utility remains a fundamental challenge. Existing approaches often struggle to balance comprehensive safety with fine-grained controllability at the…

Automatically generating source code from natural language using large language models (LLMs) is becoming common, yet security vulnerabilities persist despite advances in fine tuning and prompting. In this work, we systematically evaluate…

Iterative self-correction is increasingly deployed in agentic LLM systems, yet whether repeated refinement improves or degrades performance remains inconsistent across models. We recast self-correction as a closed-loop feedback-control…

LLM-based code interpreter agents are increasingly deployed in critical workflows, yet their robustness against risks introduced by their code execution capabilities remains underexplored. Existing benchmarks are limited to static datasets…

Maintaining the safety of large language models (LLMs) is crucial as they are increasingly deployed in real-world applications. Existing safety guardrails typically rely on single-pass classification or, more recently, distilled reasoning.…

Synchronous federated learning scales poorly due to the straggler effect. Asynchronous algorithms increase the update throughput by processing updates upon arrival, but they introduce two fundamental challenges: gradient staleness, which…

Large Language Models (LLMs) have shown promising performance in software vulnerability detection, particularly after domain-specific Supervised Fine-Tuning (SFT). However, it remains unclear whether these models genuinely internalize…

Looped Language Models (LoopLMs) enable efficient latent reasoning through depth recurrence, yet exhibit unreliable test-time scaling behavior: performance often peaks at a certain iteration depth and then collapses with further recurrence.…

Supervised Fine-Tuning (SFT) followed by Reinforcement Learning (RL) has emerged as the standard post-training paradigm for large language models (LLMs). However, the conventional SFT process, driven by Cross-Entropy (CE) loss, often…

Techniques for verifying or invalidating the security of computer systems have come a long way in recent years. Extremely sophisticated tools are available to specify and formally verify the behavior of a system and, at the same time,…