arXiv subjects

Ziqi Gong

Publications and source records attributed to Ziqi Gong.

6 recordsLinked to original sources

arXiv2026

OLED-MoE: Accelerating MoE-Based dLLM Inference via Inter-Iteration Locality-Aware Expert Offloading

Semi-autoregressive diffusion large language models (dLLMs) improve decoding parallelism through iterative block-wise denoising, but scaling them with mixture-of-experts (MoE) layers introduces a large expert parameter footprint that exceeds memory-constrained GPU capacity. Expert offloading is a natural remedy, yet existing MoE serving systems target autoregressive decoding and rely on intra-iteration layer-wise prefetching: while computing one layer, they predict and load experts for subsequent layers. Under dLLM inference, block-wise routing expands the active expert working set within each iteration, making such prefetches difficult to complete in time and costly when mispredicted. Consequently, existing prefetch-based solutions often degenerate into on-demand expert loading with high decoding latency. We propose OLED-MoE, an expert offloading system that shifts the optimization target from intra-iteration prefetching to inter-iteration expert retention. Its key insight is that adjacent denoising iterations exhibit strong expert routing overlap, and token confidence indicates which experts are likely to be reused. OLED-MoE uses confidence-guided inter-iteration prediction to retain high-value experts in GPU memory without introducing extra prefetch traffic. It further compensates unavoidable cache misses through CPU-GPU cooperative execution, jointly considering dynamic expert computation load and predicted future reuse. Across diverse dLLM workloads, OLED-MoE reduces time per output token (TPOT) by 1.23x-7.93x and improves expert cache utilization by 1.44x-4.23x over state-of-the-art offloading systems. Notably, OLED-MoE approaches full-residency performance while using only 40% of the expert GPU memory, incurring merely 23% higher TPOT despite a 60% reduction in expert memory footprint. OLED-MoE's source code is publicly available at https://github.com/flashserve/OLED-MoE.

cs.DC↗

arXiv2026

BIRD: Distilling Decision Boundaries into Rationales for MLLM Adaptation

Adapting general-purpose multimodal large language models (MLLMs) to specialized domains requires learning domain-specific decision criteria, which often hinge on subtle visual distinctions between otherwise plausible answers. Rationale augmentation aims to expose such evidence through additional observations or inter-sample comparisons, yet a visually valid cue is not necessarily decision-relevant: it may describe how samples differ without changing the model's relative preference between competing answers. We therefore introduce BIRD, a self-improving Boundary-Informed Rationale Distillation framework that uses model-specific confusions to locate unresolved local decision boundaries and distills the evidence that resolves these confusions into rationales. For each sample, BIRD retrieves candidate neighbors from the target MLLM's own representation space and selects the most confusable one according to its answer preferences. It then generates answer-blind candidate evidence from their visual differences and functionally verifies which evidence most effectively strengthens the model's preference for the correct answer while avoiding inappropriate transfer across the pair. The verified evidence is then distilled into a single-sample rationale for standard supervised fine-tuning. Experiments on medical and chart VQA show that BIRD outperforms competing rationale-augmentation methods across two target MLLMs, while further analyses demonstrate clearer separation of confusable answers and stronger gains from model-matched supervision.

cs.AI↗

arXiv2026

Robust Biomolecular Complex Design Across Protein Conformational Landscapes

Proteins populate conformational ensembles, yet structure-based biomolecular design typically optimizes candidates against a single target conformation. Consequently, a candidate that fits one state can lose favorable interactions or develop steric clashes when the target adopts another. We introduce FlexEvo, a model-agnostic evolutionary framework that adapts candidates once at inference time from a single target conformation to improve compatibility with alternative natural conformations unseen during adaptation, without retraining the source model or requiring a conformational ensemble. FlexEvo casts cross-state adaptation as geometry-constrained bi-objective optimization, balancing preservation of input-state interactions against robustness to plausible conformational perturbations. To limit the search space and reduce invalid structural edits, geometry-derived FlexBoxes define protected anchor regions, adaptable regions for local exploration, and forbidden regions for clash avoidance. A unified all-atom representation supports topology-preserving adaptation across diverse binder categories, while Pareto selection preserves nondominated candidates across the two objectives. We evaluate FlexEvo across multiple generation baselines and nine representative binder categories spanning diverse molecular sizes and structural topologies. FlexEvo reduces the category-balanced mean relative performance degradation from 47.8% to 4.4%, while adding only 1.4--3.1 minutes of adaptation per sample. These results establish single-state inference-time adaptation as a practical route toward robust biomolecular complex design across protein conformational landscapes.

cs.AI↗

arXiv2025

DoPI: Doctor-like Proactive Interrogation LLM for Traditional Chinese Medicine

Enhancing interrogation capabilities in Traditional Chinese Medicine (TCM) diagnosis through multi-turn dialogues and knowledge graphs presents a significant challenge for modern AI systems. Current large language models (LLMs), despite their advancements, exhibit notable limitations in medical applications, particularly in conducting effective multi-turn dialogues and proactive questioning. These shortcomings hinder their practical application and effectiveness in simulating real-world diagnostic scenarios. To address these limitations, we propose DoPI, a novel LLM system specifically designed for the TCM domain. The DoPI system introduces a collaborative architecture comprising a guidance model and an expert model. The guidance model conducts multi-turn dialogues with patients and dynamically generates questions based on a knowledge graph to efficiently extract critical symptom information. Simultaneously, the expert model leverages deep TCM expertise to provide final diagnoses and treatment plans. Furthermore, this study constructs a multi-turn doctor-patient dialogue dataset to simulate realistic consultation scenarios and proposes a novel evaluation methodology that does not rely on manually collected real-world consultation data. Experimental results show that the DoPI system achieves an accuracy rate of 84.68 percent in interrogation outcomes, significantly enhancing the model's communication ability during diagnosis while maintaining professional expertise.

cs.AI↗

arXiv2024

Neural Directed Speech Enhancement with Dual Microphone Array in High Noise Scenario

In multi-speaker scenarios, leveraging spatial features is essential for enhancing target speech. While with limited microphone arrays, developing a compact multi-channel speech enhancement system remains challenging, especially in extremely low signal-to-noise ratio (SNR) conditions. To tackle this issue, we propose a triple-steering spatial selection method, a flexible framework that uses three steering vectors to guide enhancement and determine the enhancement range. Specifically, we introduce a causal-directed U-Net (CDUNet) model, which takes raw multi-channel speech and the desired enhancement width as inputs. This enables dynamic adjustment of steering vectors based on the target direction and fine-tuning of the enhancement region according to the angular separation between the target and interference signals. Our model with only a dual microphone array, excels in both speech quality and downstream task performance. It operates in real-time with minimal parameters, making it ideal for low-latency, on-device streaming applications.

eess.AS↗

arXiv2024

Harpagon: Minimizing DNN Serving Cost via Efficient Dispatching, Scheduling and Splitting

Advances in deep neural networks (DNNs) have significantly contributed to the development of real-time video processing applications. Efficient scheduling of DNN workloads in cloud-hosted inference systems is crucial to minimizing serving costs while meeting application latency constraints. However, existing systems suffer from excessive module latency during request dispatching, low execution throughput during module scheduling, and wasted latency budget during latency splitting for multi-DNN application, which undermines their capability to minimize the serving cost. In this paper, we design a DNN inference system called Harpagon, which minimizes the serving cost under latency constraints with a three-level design. It first maximizes the batch collection rate with a batch-aware request dispatch policy to minimize the module latency. It then maximizes the module throughput with multi-tuple configurations and proper amount of dummy requests. It also carefully splits the end-to-end latency into per-module latency budget to minimize the total serving cost for multi-DNN applications. Evaluation shows that Harpagon outperforms the state of the art by 1.49 to 2.37 times in serving cost while satisfying the latency objectives. Additionally, compared to the optimal solution using brute force search, Harpagon derives the lower bound of serving cost for 91.5% workloads with millisecond level runtime.

cs.DC↗