arXiv subjects

Daniel Zhao

Publications and source records attributed to Daniel Zhao.

9 recordsLinked to original sources

arXiv2026

Global Sketch-Based Watermarking for Diffusion Language Models

Watermarking methods for language models have been studied extensively in the autoregressive setting, where tokens are generated sequentially. These works largely focus on local-context schemes that perturb the next token's distribution as a function of its preceding tokens. In diffusion language models, distributions over many unresolved positions are jointly sampled, allowing additive statistics of the entire sequence to be tractable during generation. We propose a watermark for masked diffusion language models that controls a global, vector-valued sketch representation of the text. Compared to context-dependent watermarking, the sketch formulation decouples detection from the local contexts seen during generation, resulting in an order-agnostic statistic and a watermarking rule which does not manifest as a simple token bias. We analyze the distortion, soundness, and robustness properties of the method.

cs.CR↗

arXiv2026

Yuan3.0 Ultra: A Trillion-Parameter Enterprise-Oriented MoE LLM

We introduce Yuan3.0 Ultra, an open-source Mixture-of-Experts (MoE) large language model featuring 68.8B activated parameters and 1010B total parameters, specially designed to enhance performance on enterprise scenarios tasks while maintaining competitive capabilities on general purpose tasks. We propose Layer-Adaptive Expert Pruning (LAEP) algorithm designed for the pre-training stage of MoE LLMs. In contrast to previous expert pruning approaches that operate primarily in the post-training phase, the proposed algorithm enhances training efficiency by selectively pruning underutilized experts and reorganizing experts across computing devices according to token distribution statistics. Comprehensive experiments demonstrate that LAEP effectively reduces model size and substantially improves pre-training efficiency. When pre-training Yuan3.0 Ultra from scratch original with 1515B parameters, this algorithm delivers a 49\% boost in pre-training efficiency and a 33.3\% reduction in total parameters, while preserving the model's outstanding multi-domain performance. On enterprise scenario benchmarks including Docmatix, ChatRAG, SummEval and MMTab, Yuan3.0 Ultra achieves leading accuracy. The model and codes are publicly available at https://github.com/Yuan-lab-LLM/Yuan3.0-Ultra.

cs.LG↗

arXiv2025

Benchmarking Scientific Understanding and Reasoning for Video Generation using VideoScience-Bench

The next frontier for video generation lies in developing models capable of zero-shot reasoning, where understanding real-world scientific laws is crucial for accurate physical outcome modeling under diverse conditions. However, existing video benchmarks are physical commonsense-based, offering limited insight into video models' scientific reasoning capability. We introduce VideoScience-Bench, a benchmark designed to evaluate undergraduate-level scientific understanding in video models. Each prompt encodes a composite scientific scenario that requires understanding and reasoning across multiple scientific concepts to generate the correct phenomenon. The benchmark comprises 200 carefully curated prompts spanning 14 topics and 103 concepts in physics and chemistry. We conduct expert-annotated evaluations across seven state-of-the-art video models in T2V and I2V settings along five dimensions: Prompt Consistency, Phenomenon Congruency, Correct Dynamism, Immutability, and Spatio-Temporal Continuity. Using a VLM-as-a-Judge to assess video generations, we observe strong correlation with human assessments. To the best of our knowledge, VideoScience-Bench is the first benchmark to evaluate video models not only as generators but also as reasoners, requiring their generations to demonstrate scientific understanding consistent with expected physical and chemical phenomena. Our data and evaluation code are available at: \href{https://github.com/hao-ai-lab/VideoScience}{github.com/hao-ai-lab/VideoScience}.

cs.CV↗

arXiv2025

Steering Autoregressive Music Generation with Recursive Feature Machines

Controllable music generation remains a significant challenge, with existing methods often requiring model retraining or introducing audible artifacts. We introduce MusicRFM, a framework that adapts Recursive Feature Machines (RFMs) to enable fine-grained, interpretable control over frozen, pre-trained music models by directly steering their internal activations. RFMs analyze a model's internal gradients to produce interpretable "concept directions", or specific axes in the activation space that correspond to musical attributes like notes or chords. We first train lightweight RFM probes to discover these directions within MusicGen's hidden states; then, during inference, we inject them back into the model to guide the generation process in real-time without per-step optimization. We present advanced mechanisms for this control, including dynamic, time-varying schedules and methods for the simultaneous enforcement of multiple musical properties. Our method successfully navigates the trade-off between control and generation quality: we can increase the accuracy of generating a target musical note from 0.23 to 0.82, while text prompt adherence remains within approximately 0.02 of the unsteered baseline, demonstrating effective control with minimal impact on prompt fidelity. We release code to encourage further exploration on RFMs in the music domain.

cs.LG↗

arXiv2025

Towards Interpretable and Inference-Optimal COT Reasoning with Sparse Autoencoder-Guided Generation

We propose a novel method that leverages sparse autoencoders (SAEs) and clustering techniques to analyze the internal token representations of large language models (LLMs) and guide generations in mathematical reasoning tasks. Our approach first trains an SAE to generate sparse vector representations for training tokens, then applies k-means clustering to construct a graph where vertices represent token clusters and weighted edges capture sequential token transitions. Using this graph, we define an edge-weight based reward function to quantify adherence to established reasoning traces, thereby identifying exploitative reasoning trajectories. Additionally, we measure generation diversity from clustering to assess the extent of exploration. Our findings indicate that balancing both exploitation and exploration is crucial for achieving high accuracy in mathematical reasoning tasks. During generation, the SAE can serve as a scalable reward model to guide generations, ensuring a balanced trade-off between exploitation and exploration. This prevents extreme behaviors in either direction, ultimately fostering a higher-quality reasoning process in LLMs.

cs.AI↗

arXiv2024

Policy Gradients for Optimal Parallel Tempering MCMC

Parallel tempering is a meta-algorithm for Markov Chain Monte Carlo that uses multiple chains to sample from tempered versions of the target distribution, enhancing mixing in multi-modal distributions that are challenging for traditional methods. The effectiveness of parallel tempering is heavily influenced by the selection of chain temperatures. Here, we present an adaptive temperature selection algorithm that dynamically adjusts temperatures during sampling using a policy gradient approach. Experiments demonstrate that our method can achieve lower integrated autocorrelation times compared to traditional geometrically spaced temperatures and uniform acceptance rate schemes on benchmark distributions.

stat.CO↗

arXiv2023

Mapping "Brain Terrain" Regions on Mars using Deep Learning

One of the main objectives of the Mars Exploration Program is to search for evidence of past or current life on the planet. To achieve this, Mars exploration has been focusing on regions that may have liquid or frozen water. A set of critical areas may have seen cycles of ice thawing in the relatively recent past in response to periodic changes in the obliquity of Mars. In this work, we use convolutional neural networks to detect surface regions containing "Brain Coral" terrain, a landform on Mars whose similarity in morphology and scale to sorted stone circles on Earth suggests that it may have formed as a consequence of freeze/thaw cycles. We use large images (~100-1000 megapixels) from the Mars Reconnaissance Orbiter to search for these landforms at resolutions close to a few tens of centimeters per pixel (~25--50 cm). Over 52,000 images (~28 TB) were searched (~5% of the Martian surface) where we found detections in over 200 images. To expedite the processing we leverage a classifier network (prior to segmentation) in the Fourier domain that can take advantage of JPEG compression by leveraging blocks of coefficients from a discrete cosine transform in lieu of decoding the entire image at the full spatial resolution. The hybrid pipeline approach maintains ~93% accuracy while cutting down on ~95% of the total processing time compared to running the segmentation network at the full resolution on every image. The timely processing of big data sets helps inform mission operations, geologic surveys to prioritize candidate landing sites, avoid hazardous areas, or map the spatial extent of certain terrain. The segmentation masks and source code are available on Github for the community to explore and build upon.

astro-ph.EP↗

arXiv2022

Seasonal variations of chemical species and haze in Titan's upper atmosphere

Seasonal variation is significant in Titan's atmosphere due to the large change of solar insolation resulting from Titan's 26.7{\deg} axial tilt relative to the plane of Saturn's orbit. Here we present an investigation of hydrocarbon and nitrile species in Titan's upper atmosphere at 400-1200 km, which includes the mesosphere and the lower thermosphere, over more than one fourth of Titan's year (2006-2014, LS=318{\deg}-60{\deg}), using eighteen stellar occultation observations obtained by Cassini/UVIS. Vertical profiles of eight chemical species (CH4, C2H2, C2H4, C2H6, C4H2, C6H6, HCN, HC3N) and haze particles are retrieved from these observations using an instrument forward model, which considers the technical issue of pointing motion. The Markov-chain Monte Carlo (MCMC) algorithm is used to obtain the posterior probability distributions of parameters in the retrieval, which inherently tests the extent to which species profiles can be constrained. The results show that no change of the species profiles is noticeable before the equinox, while the decrease of atmospheric temperature and significant upwelling in the summer hemisphere are found five terrestrial years afterwards. Altitude of the detached haze layer decreases towards the vernal equinox then it disappears, and no reappearance is identified within the time range of our data, which is consistent with observations from Cassini/ISS. This study provides observational constraints on the seasonal change of Titan's upper atmosphere, and suggests further investigations of the atmospheric chemistry and dynamics therein.

astro-ph.EP↗

arXiv2021

Revisiting the Planet Mass and Stellar Metallicity Relation for Low-Mass Exoplanets Orbiting GKM Class Stars

The growing database of exoplanets have shown us the statistical characteristics of various exoplanet populations, providing insight towards their origins. Observational evidence suggests that the process by which gas giants are conceived in the stellar disk may be disparate from that of smaller planets. Using NASA's Exoplanet Archive, we analyzed a correlation between the planet mass and stellar metallicity of low-mass exoplanets (MP < 0.13 MJ) orbiting spectral class G, K, and M stars. The correlation suggests an exponential law relationship between the two that is not fully explained by observation biases alone.

astro-ph.EP↗