Jiachen Zhao

Jia-Chen Zhao, Beiqi Chen, Xinyang Chen et al.

We present StructSplat, a feed-forward and generalizable 3D Gaussian reconstruction framework that operates directly on uncalibrated images without requiring camera parameters. Existing methods either rely on per-scene optimization or assume known camera poses, and often entangle geometry and appearance within a unified backbone, limiting reconstruction fidelity and generalization. Our key idea is to adopt a structured representation that organizes geometry, semantic, and texture cues with explicit roles in the reconstruction process. Specifically, we introduce a pixel-aligned feature injection mechanism to enable accurate texture modeling from 2D observations, incorporate semantic-aware priors to improve global consistency, and design a camera alignment strategy to prevent information leakage and improve generalization. Experiments show that our method significantly outperforms prior approaches on challenging benchmarks. On DL3DV, our method achieves 28.045 PSNR, surpassing AnySplat (22.377) by +5.67 dB. In cross-dataset evaluation, our method achieves +1.94 dB over AnySplat on ACID and +1.72 dB on RealEstate10K. Project page: https://structsplat.github.io Code: https://github.com/J-C-Zhao/StructSplat

Jia-Chen Zhao

Class imbalance, which is also called long-tailed distribution, is a common problem in classification tasks based on machine learning. If it happens, the minority data will be overwhelmed by the majority, which presents quite a challenge for data science. To address the class imbalance problem, researchers have proposed lots of methods: some people make the data set balanced (SMOTE), some others refine the loss function (Focal Loss), and even someone has noticed the value of labels influences class-imbalanced learning (Yang and Xu. Rethinking the value of labels for improving class-imbalanced learning. In NeurIPS 2020), but no one changes the way to encode the labels of data yet. Nowadays, the most prevailing technique to encode labels is the one-hot encoding due to its nice performance in the general situation. However, it is not a good choice for imbalanced data, because the classifier will treat majority and minority samples equally. In this paper, we innovatively propose the enhancement encoding technique, which is specially designed for the imbalanced classification. The enhancement encoding combines re-weighting and cost-sensitiveness, which can reflect the difference between hard and easy (or minority and majority) classes. To reduce the number of validation samples and the computation cost, we also replace the confusion matrix with a novel soft-confusion matrix which works better with a small validation set. In the experiments, we evaluate the enhancement encoding with three different types of loss. And the results show that enhancement encoding is very effective to improve the performance of the network trained with imbalanced data. Particularly, the performance on minority classes is much better.

Jiachen Zhao, Jing Huang, Zhengxuan Wu et al. · stanford

LLMs are trained to refuse harmful instructions, but do they truly understand harmfulness beyond just refusing? Prior work has shown that LLMs' refusal behaviors can be mediated by a one-dimensional subspace, i.e., a refusal direction. In this work, we identify a new dimension to analyze safety mechanisms in LLMs, i.e., harmfulness, which is encoded internally as a separate concept from refusal. There exists a harmfulness direction that is distinct from the refusal direction. As causal evidence, steering along the harmfulness direction can lead LLMs to interpret harmless instructions as harmful, but steering along the refusal direction tends to elicit refusal responses directly without reversing the model's judgment on harmfulness. Furthermore, using our identified harmfulness concept, we find that certain jailbreak methods work by reducing the refusal signals without reversing the model's internal belief of harmfulness. We also find that adversarially finetuning models to accept harmful instructions has minimal impact on the model's internal belief of harmfulness. These insights lead to a practical safety application: The model's latent harmfulness representation can serve as an intrinsic safeguard (Latent Guard) for detecting unsafe inputs and reducing over-refusals that is robust to finetuning attacks. For instance, our Latent Guard achieves performance comparable to or better than Llama Guard 3 8B, a dedicated finetuned safeguard model, across different jailbreak methods. Our findings suggest that LLMs' internal understanding of harmfulness is more robust than their refusal decision to diverse input instructions, offering a new perspective to study AI safety.