arXiv subjects

Guanzhe Hong

Publications and source records attributed to Guanzhe Hong.

4 recordsLinked to original sources

arXiv2026

Are Human-Aligned Models Models of Humans? A Turing-Test Gap in Preference Alignment

Human-feedback alignment has made language models useful assistants and is commonly described as aligning them with humans. However, the responses people prefer from an AI need not be the responses they themselves would give. We distinguish alignment with human preferences from alignment with human behavior, and show that alignment with human preferences can make model behavior less human-like even when both preferences and responses come entirely from humans. We call this the Turing-test gap. We show that preference alignment preserves the human response distribution only under a restrictive condition, and find no consistent evidence that real human preferences satisfy it. Empirically, the loss of human-response likelihood increases with the strength of preference weighting, regardless of its direction, and the gap also appears under standard DPO. These results establish human-likeness as an explicit dimension of alignment rather than something assumed to follow from preference alignment.

cs.AI↗

arXiv2026

$D^2$-Monitor: Dynamic Safety Monitoring for Diffusion LLMs via Hesitation-Aware Routing

Despite the emergence of diffusion large language models (D-LLMs) as an alternative to autoregressive large language models (AR-LLMs), safety monitoring for D-LLMs remains largely unexplored. Unlike AR-LLMs, D-LLMs generate text through a multi-step denoising process, exposing intermediate hidden representations that may contain safety-relevant information unavailable in standard single-step monitoring setups. Motivated by the suitability of lightweight probes for always-on monitoring, we analyze which trajectory-level signals best indicate when such probes are likely to struggle. We find that the most informative signal is safety hesitation: intermediate hidden states repeatedly falling within a small margin of the probe's decision boundary. The number of such hesitation steps in D-LLM's trajectory predicts probe failure effectively, providing a proxy of sample difficulty. Building on this analysis, we propose $D^2$-Monitor, a bi-level safety monitor for D-LLMs. $D^2$-Monitor adopts a lightweight probe as an always-on monitor to jointly estimate hesitation and perform base classification. When the hesitation level exceeds a threshold, a more expressive but computationally heavier probe is activated. This dynamic routing mechanism allocates monitoring resources efficiently at test time. Evaluated on 3 datasets (WildguardMix, ToxicChat, OpenAI-Moderation) across 4 D-LLMs, $D^2$-Monitor achieves state-of-the-art performance with a compact parameter footprint ($\leq$ 0.85M parameters), and exhibits the best trade-off between effectiveness and efficiency relative to 8 baselines.

cs.AI↗

arXiv2021

Student-Teacher Learning from Clean Inputs to Noisy Inputs

Feature-based student-teacher learning, a training method that encourages the student's hidden features to mimic those of the teacher network, is empirically successful in transferring the knowledge from a pre-trained teacher network to the student network. Furthermore, recent empirical results demonstrate that, the teacher's features can boost the student network's generalization even when the student's input sample is corrupted by noise. However, there is a lack of theoretical insights into why and when this method of transferring knowledge can be successful between such heterogeneous tasks. We analyze this method theoretically using deep linear networks, and experimentally using nonlinear networks. We identify three vital factors to the success of the method: (1) whether the student is trained to zero training loss; (2) how knowledgeable the teacher is on the clean-input problem; (3) how the teacher decomposes its knowledge in its hidden features. Lack of proper control in any of the three factors leads to failure of the student-teacher learning method.

cs.LG↗

arXiv2019

Rethinking Atmospheric Turbulence Mitigation

State-of-the-art atmospheric turbulence image restoration methods utilize standard image processing tools such as optical flow, lucky region and blind deconvolution to restore the images. While promising results have been reported over the past decade, many of the methods are agnostic to the physical model that generates the distortion. In this paper, we revisit the turbulence restoration problem by analyzing the reference frame generation and the blind deconvolution steps in a typical restoration pipeline. By leveraging tools in large deviation theory, we rigorously prove the minimum number of frames required to generate a reliable reference for both static and dynamic scenes. We discuss how a turbulence agnostic model can lead to potential flaws, and how to configure a simple spatial-temporal non-local weighted averaging method to generate references. For blind deconvolution, we present a new data-driven prior by analyzing the distributions of the point spread functions. We demonstrate how a simple prior can outperform state-of-the-art blind deconvolution methods.

eess.IV↗