13 papers
Activation Flow: Manufacturing Activations for Steering
Hong Kiat Tan, Linh Le, David Williams-King
Difference-in-means steering requires activations recorded while a model shows the desired behavior, which a sandbagging model withholds by deliberately underperforming. We introdu…
GPS-Bench: A Governance Policy Benchmark for Automating Policy Analysis
Linh Le, Melanie Bui, My Chiffon Nguyen +2
Policy analysis requires more than predicting whether a proposal will pass: it requires identifying who will be affected, how those actors respond, and what follows. LLM-based poli…
A Causal Model for Locating and Unlocking Sandbagging in Model Organisms
Hong Kiat Tan, Linh Le, David Williams-King
Sandbagging models strategically underperform on evaluations while retaining the capabilities being measured. The evaluations that guide frontier-model deployment and governance th…
Reference-Grafting Matches Fine-Tuning at Eliciting Sandbagged Capabilities
Linh Le, Hong Kiat Tan, David Williams-King
Sandbagging, in which a model deliberately underperforms on an evaluation despite retaining the underlying capability, threatens the safety evaluations that frontier-model governan…
Efficient Safety Alignment of Language Models via Latent Personality Traits
Mohamed Amine Merzouk, Nolan Smyth, Damiano Fornasiere +3
Current safety methods for large language models are known to be vulnerable to adversarial attacks, motivating research into robust alternatives. Latent Adversarial Training (LAT)…
Towards Efficient Reasoning in LLM-Based Recommender Systems via Model Merging
Linh Dieu Le, Tong Chen, Shazia Sadiq +3
Large language model-based recommender systems are increasingly adopting slow-thinking models that generate step-by-step reasoning before making predictions, often achieving higher…
vla.cpp: A Unified Inference Runtime for Vision-Language-Action Models
Khanh D. Nguyen, Hung T. Ho, Chinh T. Nguyen +5
Deploying vision--language--action (VLA) models on robots requires adapting model-specific inference pipelines to heterogeneous processors and limited onboard memory. We present vl…
FragBench: Cross-Session Attacks Hidden in Benign-Looking Fragments
Astha Mehta, Niruthiha Selvanayagam, Cedric Lam +10
An attacker can split a malicious goal into sub-prompts that each look benign on their own and only become harmful in combination. Existing LLM safety benchmarks evaluate prompts o…
Latent Personality Alignment: Improving Harmlessness Without Mentioning Harms
Linh Le, David Williams-King, Mohamed Amine Merzouk +2
Current adversarial robustness methods for large language models require extensive datasets of harmful prompts (thousands to hundreds of thousands of examples), yet remain vulnerab…
Do LLMs Hold Their Values? MANTA: A Multi-Turn Adversarial Benchmark for Animal Welfare Reasoning
Isabella Luong, Joyee Chen, Sankalpa Ghose +3
Evaluating animal welfare reasoning in LLMs remains an open challenge despite rapid deployment in consumer and professional contexts where welfare considerations appear implicitly…
Diagnosing Pathological Chain-of-Thought in Reasoning Models
Manqing Liu, David Williams-King, Ida Caspary +5
Chain-of-thought (CoT) reasoning is fundamental to modern LLM architectures and represents a critical intervention point for AI safety. However, CoT reasoning may exhibit failure m…
Representation Engineering for Large-Language Models: Survey and Research Challenges
Lukasz Bartoszcze, Sarthak Munshi, Bryan Sukidi +6
Large-language models are capable of completing a variety of tasks, but remain unpredictable and intractable. Representation engineering seeks to resolve this problem through a new…
Can Safety Fine-Tuning Be More Principled? Lessons Learned from Cybersecurity
David Williams-King, Linh Le, Adam Oberman +1
As LLMs develop increasingly advanced capabilities, there is an increased need to minimize the harm that could be caused to society by certain model outputs; hence, most LLMs have…