Ligong Han

Toward nonlinear representations with Gaussian-splat manifolds for physics-informed learning

Xu Han, Kuangxu Chen, Rui Wang, Guanzhou Wei et autres

Conventional linear discretizations, including high-order schemes, often require prohibitively many degrees of freedom to resolve sharp, localized features, and the practical advantages of nonlinear, physics-informed representations over fixed linear spaces remain unclear. Here we introduce a compact nonlinear-manifold representation, based on Gaussian …

us, cn (code pays fourni par la source)

SNLP: Layer-Parallel Inference via Structured Newton Corrections

Ligong Han, Kai Xu, Hao Wang, Akash Srivastava

Autoregressive language models execute Transformer layers sequentially, creating a latency bottleneck that is not removed by conventional tensor or pipeline parallelism. We study whether this layerwise dependency can be relaxed by treating the hidden-state trace across layers as the solution of a …

SNLP: Layer-Parallel Inference via Structured Newton Corrections

Ligong Han, Kai Xu, Hao Wang, Akash Srivastava

Autoregressive language models execute Transformer layers sequentially, creating a latency bottleneck that is not removed by conventional tensor or pipeline parallelism. We study whether this layerwise dependency can be relaxed by treating the hidden-state trace across layers as the solution of a …

us (code pays fourni par la source)

S2D2: Fast Decoding for Diffusion LLMs via Training-Free Self-Speculation

Ligong Han, Hao Wang, Han Gao, Kai Xu et autres

Block-diffusion language models offer a promising path toward faster-than-autoregressive generation by combining block-wise autoregressive decoding with within-block parallel denoising. However, in the few-step regime needed for practical acceleration, standard confidence-thresholded decoding is often brittle: aggressive thresholds hurt quality, while conservative thresholds require …

S2D2: Fast Decoding for Diffusion LLMs via Training-Free Self-Speculation

Ligong Han, Hao Wang, Han Gao, Kai Xu et autres

Block-diffusion language models offer a promising path toward faster-than-autoregressive generation by combining block-wise autoregressive decoding with within-block parallel denoising. However, in the few-step regime needed for practical acceleration, standard confidence-thresholded decoding is often brittle: aggressive thresholds hurt quality, while conservative thresholds require …

us (code pays fourni par la source)

CATP: Contextually Adaptive Token Pruning for Efficient and Enhanced Multimodal In-Context Learning

Yanshu Li, Jianjiang Yang, Zhennan Shen, Ligong Han et autres

Modern large vision-language models (LVLMs) convert each input image into a large set of tokens that far outnumber the text tokens. Although this improves visual perception, it also introduces severe image token redundancy. Because image tokens contain sparse information, many contribute little …

us, gb (code pays fourni par la source)

DICE: Discrete Inversion Enabling Controllable Editing for Masked Generative Models

Xiaoxiao He, 稲泉, Ligong Han, Song Wen et autres

Recent advances in discrete diffusion models have demonstrated strong performance in image generation and masked language modeling, yet they remain limited in their capacity for controlled content editing. We propose DICE (Discrete Inversion for Controllable Editing), a novel framework that pioneers precise …

us, gb (code pays fourni par la source)

Few-Step Diffusion Language Models via Trajectory Self-Distillation

Tunyu Zhang, Xinxi Zhang, Ligong Han, Haizhou Shi et autres

Diffusion large language models (DLLMs) have emerged as powerful generative models with the promise of fast text generation through parallel decoding. However, realizing this potential in practice remains challenging: reducing the number of decoding steps, typically causes a substantial degradation in output …

Few-Step Diffusion Language Models via Trajectory Self-Distillation

Tunyu Zhang, Xinxi Zhang, Ligong Han, Haizhou Shi et autres

Diffusion large language models (DLLMs) have emerged as powerful generative models with the promise of fast text generation through parallel decoding. However, realizing this potential in practice remains challenging: reducing the number of decoding steps, typically causes a substantial degradation in output …

nl, us, de (code pays fourni par la source)

Reasoning over Precedents Alongside Statutes: Case-Augmented Deliberative Alignment for LLM Safety

Can Jin, Rui Wu, Tong Che, Qixin Zhang et autres

Ensuring that Large Language Models (LLMs) adhere to safety principles without refusing benign requests remains a significant challenge. While OpenAI introduces deliberative alignment (DA) to enhance the safety of its o-series models through reasoning over detailed ``code-like'' safety rules, the effectiveness of …

Reasoning over Precedents Alongside Statutes: Case-Augmented Deliberative Alignment for LLM Safety

Can Jin, Rui Wu, Tong Che, Qixin Zhang et autres

Ensuring that Large Language Models (LLMs) adhere to safety principles without refusing benign requests remains a significant challenge. While OpenAI introduces deliberative alignment (DA) to enhance the safety of its o-series models through reasoning over detailed ``code-like'' safety rules, the effectiveness of …

Discrete Noise Inversion for Next-Scale Autoregressive Text-Based Image Editing

Quan Thanh Dao, Xiaoxiao He, Ligong Han, Ngan Nguyen et autres

us, hk (code pays fourni par la source)