9 citations · 24 across the 24 of their papers we have counts for
10 papers · 1 filter
Corrupted but Correct: Why Vision-Language Models Lie to Themselves Internally
Arun Josephraj Arokiaraj, Zekun Wu, Adriano Koshiyama
A targeted adversarial perturbation can drive a vision-language model's (VLM's) teacher-forced training loss for a fixed target caption to near zero, yet the same model, allowed to…
Are Single-Token Sparse Autoencoder Features Causally Necessary? Layer-Depth and SAE-Family Effects
Seonglae Cho, Zekun Wu, Kleyton Da Costa +3
Sparse autoencoder (SAE) features are used to interpret and steer large language models, yet nobody has tested whether a feature's causal role is stable across SAE families. Single…
Automata from Agent Traces: Failure and Next-Step Prediction
Seonglae Cho, Franklin Cardenoso Fernandez, Umar Mohammed +4
LLM-based agents execute multi-step tasks, but their behavioral structure remains opaque: long unstructured traces resist the safety auditing and runtime monitoring that deployment…
Which Decisions Low-Bit Quantization Breaks, and How to Predict Them
Zekun Wu, Swati Dhiman, Adriano Koshiyama
Quantization saves memory by storing model weights with fewer bits. It can also change model decisions, such as whether to call a tool or which option to choose from a finite set.…
Routing Is Least Learnable Where It Is Most Valuable: Bounds on Representation Routing for Web Agents
Jiaming Wei, Zekun Wu, Adriano Koshiyama +1
Web agents observe a browser through text, pixels, or both, and the choice is usually fixed once for all tasks. We measure six observation modes across eight site-model combination…
OptimismBench: Forecasting Bias and the Alignment Effect in Language Model Judgment
Seonglae Cho, Adriano Koshiyama
Large language models are increasingly used as decision aids whose probability judgments shape downstream choices. Whether those judgments carry a systematic directional tilt has b…
Tool Calling is Linearly Readable and Steerable in Language Models
Zekun Wu, Ze Wang, Seonglae Cho +4
Language-model agents can take real actions by calling tools, so choosing the wrong tool can cause errors that are difficult to undo. Most evaluations only observe the tool choice…
Sell Me This Stock: Unsafe Recommendation Drift in LLM Agents
Zekun Wu, Adriano Koshiyama, Sahan Bulathwela +1
People increasingly use LLM agents for multi-turn financial recommendations, where the agent pulls market data through tools and tracks user preferences across turns. When tool out…
Control Reinforcement Learning: Interpretable Token-Level Steering of LLMs via Sparse Autoencoder Features
Seonglae Cho, Zekun Wu, Adriano Koshiyama
Sparse autoencoders (SAEs) decompose language model activations into interpretable features, but existing methods reveal only which features activate, not which change model output…
The Confidence Manifold: Geometric Structure of Correctness Representations in Language Models
Seonglae Cho, Zekun Wu, Kleyton Da Costa +1
When a language model asserts that "the capital of Australia is Sydney," does it know this is wrong? Models assert misconceptions with the same fluency as facts, so the question ca…