arXiv subjects

Timothy Qian

Publications and source records attributed to Timothy Qian.

3 recordsLinked to original sources

arXiv2026

Chunky Post-Training: Data Driven Failures of Generalization

LLM post-training involves many diverse datasets, each targeting a specific behavior. But these datasets encode incidental patterns alongside intended ones: correlations between formatting and content, narrow phrasings across diverse problems, and implicit associations arising from the discrete data curation process. These patterns are often invisible to developers yet salient to models, producing behaviors that surprise their creators, such as rejecting true facts presented in a particular question format. We call this chunky post-training: the model learns spurious correlations as a result of distinct chunks of post-training data. We introduce SURF, a black-box pipeline which surfaces these unintended behaviors at run time, and TURF, a tool that traces these failures back to specific post-training data. Applying these tools to frontier models (Claude 4.5, GPT-5.1, Grok 4.1, Gemini 3) and open models (T\"ulu 3), we show that chunky post-training produces miscalibrated behaviors, which often result from imbalanced or underspecified chunks of post-training data.

cs.LG↗

arXiv2025

Layered Unlearning for Adversarial Relearning

Our goal is to understand how post-training methods, such as fine-tuning, alignment, and unlearning, modify language model behavior and representations. We are particularly interested in the brittle nature of these modifications that makes them easy to bypass through prompt engineering or relearning. Recent results suggest that post-training induces shallow context-dependent ``circuits'' that suppress specific response patterns. This could be one explanation for the brittleness of post-training. To test this hypothesis, we design an unlearning algorithm, Layered Unlearning (LU), that creates distinct inhibitory mechanisms for a growing subset of the data. By unlearning the first $i$ folds while retaining the remaining $k - i$ at the $i$th of $k$ stages, LU limits the ability of relearning on a subset of data to recover the full dataset. We evaluate LU through a combination of synthetic and large language model (LLM) experiments. We find that LU improves robustness to adversarial relearning for several different unlearning methods. Our results contribute to the state-of-the-art of machine unlearning and provide insight into the effect of post-training updates.

cs.LG↗

arXiv2020

Optimal Measurement of Field Properties with Quantum Sensor Networks

We consider a quantum sensor network of qubit sensors coupled to a field $f(\vec{x};\vec{\theta})$ analytically parameterized by the vector of parameters $\vec\theta$. The qubit sensors are fixed at positions $\vec{x}_1,\dots,\vec{x}_d$. While the functional form of $f(\vec{x};\vec{\theta})$ is known, the parameters $\vec{\theta}$ are not. We derive saturable bounds on the precision of measuring an arbitrary analytic function $q(\vec{\theta})$ of these parameters and construct the optimal protocols that achieve these bounds. Our results are obtained from a combination of techniques from quantum information theory and duality theorems for linear programming. They can be applied to many problems, including optimal placement of quantum sensors, field interpolation, and the measurement of functionals of parametrized fields.

quant-ph↗