9 papers
TRAP: Understanding and Mitigating Privacy Memorization in Language Models
Muhammed Ustaomeroglu, Ziyue Xu, Hanshen Xiao +3
Fine-tuning a language model on sensitive records can leave it able to reproduce them. We ask when this memorization arises and how to prevent it without knowing in advance which s…
Activation-Keyed Momentum: An Anisotropic Momentum Update via the Delta Rule
Euijin Hong, Guannan Qu
Most modern optimizers form their momentum as an exponential moving average (EMA) of past gradients, forgetting every direction at one fixed rate. However, the inputs a deep networ…
Emergent and Subliminal Misalignment Through the Lens of Data-Mediated Transfer
Baris Askin, Muhammed Ustaomeroglu, Anupam Nayak +3
Fine-tuning LLMs on narrow harmful datasets can induce Emergent Misalignment (EM), where models exhibit misaligned behavior far beyond the fine-tuning distribution. We argue that e…
Revisiting Policy Gradients for Restricted Policy Classes: Escaping Myopic Local Optima with -step Policy Gradients
Alex DeWeese, Guannan Qu
This work revisits standard policy gradient methods used on restricted policy classes, which are known to get stuck in suboptimal critical points. We identify an important cause fo…
Towards Effective Theory of LLMs: A Representation Learning Approach
Muhammed Ustaomeroglu, Guannan Qu
We propose Representational Effective Theory (RET), a framework for describing large language model computation in terms of learned macrostates rather than microscopic details. RET…
BLOCK-EM: Preventing Emergent Misalignment via Latent Blocking
Muhammed Ustaomeroglu, Guannan Qu
Emergent misalignment can arise when a language model is fine-tuned on a narrowly scoped supervised objective: the model learns the target behavior, yet also develops undesirable o…
Transformer-Based Multi-Agent Reinforcement Learning for Networked Systems with Long-Range Interactions
Vidur Sinha, Muhammed Ustaomeroglu, Guannan Qu
Multi-agent reinforcement learning (MARL) has shown promise for large-scale network control, yet existing methods face two major limitations. First, they typically rely on an expon…
Internal Planning in Language Models: Characterizing Horizon and Branch Awareness
Muhammed Ustaomeroglu, Baris Askin, Gauri Joshi +2
The extent to which decoder-only language models (LMs) engage in planning, that is, organizing intermediate computations to support coherent long-range generation, remains an impor…
A Theoretical Study of (Hyper) Self-Attention through the Lens of Interactions: Representation, Training, Generalization
Muhammed Ustaomeroglu, Guannan Qu
Self-attention has emerged as a core component of modern neural architectures, yet its theoretical underpinnings remain elusive. In this paper, we study self-attention through the…