arXiv subjects

Chang Chen

Publications and source records attributed to Chang Chen.

At least 19 recordsLinked to original sources

arXiv2026

Beyond Parallel Blindness: Information Floors and Model Gaps in Block Drafting

Block drafters propose several tokens in one forward pass, before earlier target tokens are realised. Their rejection mixes two losses: missing within-block path information and imperfect modelling of observable information. Accepted length cannot distinguish them. We separate the two with an information floor, the minimum expected rejection at a specified conditioning order; rejection above this floor is the model gap. Estimating both from target rollouts across four domains, four open-weight targets, and a frontier API target yields three findings. First, the all-parallel floor reaches $0.286$ at the final slot on Qwen3-4B, limiting even the best proposal to $71\%$ per-slot acceptance. Second, one realised token removes $86$--$100\%$ of this floor, a locality also recovered by an independent mutual-information analysis. Third, current drafters remain far above their floors: the final-slot model gap accounts for $43$--$64\%$ of DFlash rejection and $85$--$92\%$ of DSpark's oracle-conditioned rejection. These findings separate the value of short-range conditioning from proposal quality.

cs.LG↗

arXiv2026

SRL-MPC: Shape-Aware Reinforcement Learned Model Predictive Control

Safe and efficient shape-aware navigation in heterogeneous crowds and robot fleets remains challenging. Traditional approaches often assume homogeneous robots, sparse workspaces, simplified geometry, offline computation, or handcrafted parameters to make the problem tractable, which limits their deployment in dense crowd scenarios. Toward this end, we propose Shape-Aware Reinforcement Learned Model Predictive Control (SRL-MPC), a method for safe, efficient, and adaptive navigation in crowds with heterogeneous shapes without geometry simplification. To encode shape-aware safety, we formulate high-order control barrier function (HOCBF) constraints from geometric separation features (GSFs) based on support function transformation. A reinforcement learning (RL) framework then learns a neural policy that reads GSFs and outputs real-time MPC parameter updates, enabling the MPC solver to adapt to neighboring crowd geometries. The key advantage of SRL-MPC is that it preserves the safety structure and generalizability of MPC while integrating the adaptability and intelligence of RL. Experiments in randomized crowd scenarios with arbitrary shaped robot fleets demonstrate the effectiveness, scalability, and robustness of SRL-MPC. The results show that SRL-MPC substantially outperforms representative baselines in safety and adaptability. Project website: https://hanruihua.github.io/srl_mpc_project/

cs.RO↗

arXiv2026

Boogu-Image-0.1: Boosting Open Agentic Multimodal Generation via Understanding under a Minimal Budget

We introduce Boogu-Image-0.1, an open-source unified multimodal understanding and generation model family, comprising Base, Turbo, Edit, and Edit-Turbo variants. It delivers competitive performance in high-quality text-to-image generation, fast inference, instruction-based editing, and bilingual (Chinese-English) text rendering. Closed-source multimodal systems like Nano-Banana-Pro and GPT-Image-2 achieve strong performance through system-level integration rather than a single model, yet their internal practices remain largely undisclosed. In this work, we demonstrate that strengthening the understanding capability of the system, through a stronger multimodal encoder, agentic prompt rewriting, and related techniques, together with improvements in data quality, training pipelines, and agentic inference-time scaling, can substantially enhance generation and editing performance even under highly constrained compute budgets. Comprehensive evaluations show that Boogu-Image-0.1 consistently matches or surpasses other open-source models across standard benchmarks, and achieves results approaching leading closed-source systems. Notably, this is accomplished with only 208.62 million unique images. The base model's theoretical training cost is only approximately \$400K. We share practical discussions that we believe are valuable to the broader research community, and release weights, code, and recipes under Apache 2.0 to advance the open ecosystem for unified multimodal understanding and generation. Our code is available here: https://github.com/Boogu-Project/Boogu-Image.

cs.CV↗

arXiv2026

SmoothAgent: Efficient Long-Horizon LLM-Based Agent Serving with Lookahead Context Engineering

LLM-based agents execute multi-turn workflows with continuously growing contexts, where LLM calls are interleaved with tool invocations and environment feedback. To maintain model quality, modern agent frameworks rely on context engineering strategies such as offloading, reduction, and isolation to control the context length. However, these strategies introduce significant context transformation overhead: each transformation invalidates existing KV caches and triggers re-prefill, leading to increased time-to-first-token (TTFT). In this paper, we identify that context transformations are segment-decomposable, where the transformation of a prefix is independent of future tokens. This property enables transformations to be executed ahead of time. Based on this insight, we propose a lookahead programming model that allows agent frameworks to express context transformations as asynchronous operations without modifying their execution logic. The runtime proactively executes these transformations and prepares transformed KV caches in advance, enabling direct context replacement without blocking. We further design a lookahead-aware scheduler in LLM serving systems to support these asynchronous requests alongside latency-critical workloads with controlled interference. We implement our approach to support representative context engineering strategies and integrate it into existing agent frameworks and LLM serving systems. Experiments show that our approach effectively eliminates transformation overhead and reduces TTFT by up to 11.9x.

cs.DC↗

arXiv2026

FlashEvolve: Accelerating Agent Self-Evolution with Asynchronous Stage Orchestration

LLM-based evolution has emerged as a promising way to improve agents by refining non-parametric artifacts, but its wall-clock cost remains a major bottleneck. We identify that this cost comes from synchronized stage execution and imbalance inside each LLM-heavy stage. We present FlashEvolve, an efficient framework that replaces synchronized execution with asynchronous workers and queues, allowing different stages and steps to overlap. To handle data staleness introduced by asynchrony, FlashEvolve tracks artifact versions and applies different policies to update, discard, or patch stale artifacts. Unlike weight-space staleness in asynchronous RL, language-space staleness is inspectable and repairable: a stale artifact is not just delayed work, but readable evidence that the LLM can reflect on, revise, and turn into useful evolution signal. FlashEvolve further improves throughput and token efficiency with speculative stage completion and adaptive workflow control. On GEPA workloads, FlashEvolve improves proposal throughput by $3.5\times$ on local vLLM and $4.9\times$ on API serving over synchronous GEPA. The same design also applies to ACE and Meta-Harness.

cs.LG↗

arXiv2026

JigsawRL: Assembling RL Pipelines for Efficient LLM Post-Training

We present JigsawRL, a cost-efficient framework that explores Pipeline Multiplexing as a new dimension of RL parallelism. JigsawRL decomposes each pipeline into a Sub-Stage Graph that exposes the intra-stage and inter-worker imbalance hidden by stage-level systems. On this abstraction, JigsawRL resolves multiplexing interference through dynamic resource allocation, eliminates fragmented utilization by migrating long-tail rollouts across workers, and formulates their coordination as a graph scheduling problem solved with a look-ahead heuristic. On 4-64 H100/A100 GPUs across different agentic RL pipelines and models, JigsawRL achieves up to 1.85x throughput over Verl on synchronous RL, 1.54x over StreamRL and AReaL on asynchronous RL, and supports heterogeneous pipelines with moderate latency trade-off.

cs.LG↗

arXiv2026

The stabilizer ground state and applications to quantum simulation

The stabilizer ground state is defined is the lowest energy stabilizer state with respect to a given Hamiltonian. In many cases it is highly degenerate and does not give a unique stabilizer state. We define the optimal stabilizer ground state as the stabilizer ground state which has the highest fidelity with the true ground state. This is useful in quantum simulation contexts as it allows for a Clifford circuit approximation of a ground state that can be further refined towards the true ground state. We show how the optimal stabilizer ground state may be evaluated. We show applications of this state in the context of measurement-based deterministic imaginary time evolution (MITE), which converges to the ground state with high efficiency. By classically selecting the optimal stabilizer generator group and employing the stabilizer tableaux formalism, the method prepares the corresponding stabilizer ground state with maximal fidelity. The identification and refinement of this generator group are performed using a genetic algorithm tailored to the structure of the target Hamiltonian. The complexity analysis further demonstrates that algorithm's quantum resource cost scales polynomially with system size, highlighting its high efficiency and potential quantum advantage.

quant-ph↗

arXiv2026

Convergence of Neural Network Policies for Risk--Reward Optimization

We develop a neural-network framework for multi-period risk--reward stochastic control problems with constrained two-step feedback policies that may be discontinuous in the state. We allow a broad class of objectives built on a finite-dimensional performance vector, including terminal and path-dependent statistics, with risk functionals admitting auxiliary-variable optimization representations (e.g.\ Conditional Value-at-Risk and buffered probability of exceedance) and optional moment dependence. Our approach parametrizes the two-step policy using two coupled feedforward networks with constraint-enforcing output layers, reducing the constrained control problem to unconstrained training over network parameters. Under mild regularity conditions, we prove that the empirical optimum of the NN-parametrized objective converges in probability to the true optimal value as network capacity and training sample size increase. The proof is modular, separating policy approximation, propagation through the controlled recursion, and preservation under the scalarized risk--reward objective. Numerical experiments confirm the predicted convergence-in-probability behavior, show close agreement between learned and reference control heat maps, and demonstrate out-of-sample robustness on a large independent scenario set.

q-fin.CP↗

arXiv2026

Probing Multimodal Large Language Models on Cognitive Biases in Chinese Short-Video Misinformation

Short-video platforms have become major channels for misinformation, where deceptive claims frequently leverage visual experiments and social cues. While Multimodal Large Language Models (MLLMs) have demonstrated impressive reasoning capabilities, their robustness against misinformation entangled with cognitive biases remains under-explored. In this paper, we introduce a comprehensive evaluation framework using a high-quality, manually annotated dataset of 200 short videos spanning four health domains. This dataset provides fine-grained annotations for three deceptive patterns-experimental errors, logical fallacies, and fabricated claims-each verified by evidence such as national standards and academic literature. We evaluate eight frontier MLLMs across five modality settings. Experimental results demonstrate that Gemini-2.5-Pro achieves the highest performance in the multimodal setting with a belief score of 71.5/100, while o3 performs the worst at 35.2. Furthermore, we investigate social cues that induce false beliefs in videos and find that models are susceptible to biases like authoritative channel IDs.

cs.CL↗

arXiv2025

Zeppelin: Balancing Variable-length Workloads in Data Parallel Large Model Training

Training large language models (LLMs) with increasingly long and varying sequence lengths introduces severe load imbalance challenges in large-scale data-parallel training. Recent frameworks attempt to mitigate these issues through data reorganization or hybrid parallel strategies. However, they often overlook how computational and communication costs scale with sequence length, resulting in suboptimal performance. We identify three critical challenges: (1) varying computation-to-communication ratios across sequences of different lengths in distributed attention, (2) mismatch between static NIC-GPU affinity and dynamic parallel workloads, and (3) distinct optimal partitioning strategies required for quadratic attention versus linear components. To address these challenges, we present Zeppelin, a novel training system that integrates three key techniques: (1) a hierarchical sequence partitioning method for the attention module that reduces communication overhead and balances computation, supported by an efficient attention engine that applies divergent parallel strategies; (2) a routing layer that orchestrates inter-node transfers to fully utilize NIC bandwidth; and (3) a remapping layer that transforms sequence layouts between attention and linear modules, ensuring high computational efficiency across both. Comprehensive evaluations across diverse configurations show that Zeppelin delivers an average 2.80x speedup over state-of-the-art methods.

cs.DC↗

arXiv2025

RhythmTA: A Visual-Aided Interactive System for ESL Rhythm Training via Dubbing Practice

English speech rhythm, the temporal patterns of stressed syllables, is essential for English as a second language (ESL) learners to produce natural-sounding and comprehensible speech. Rhythm training is generally based on imitation of native speech. However, it relies heavily on external instructor feedback, preventing ESL learners from independent practice. To address this gap, we present RhythmTA, an interactive system for ESL learners to practice speech rhythm independently via dubbing, an imitation-based approach. The system automatically extracts rhythm from any English speech and introduces novel visual designs to support three stages of dubbing practice: (1) Synchronized listening with visual aids to enhance perception, (2) Guided repeating by visual cues for self-adjustment, and (3) Comparative reflection from a parallel view for self-monitoring. Our design is informed by a formative study with nine spoken English instructors, which identified current practices and challenges. A user study with twelve ESL learners demonstrates that RhythmTA effectively enhances learners' rhythm perception and shows significant potential for improving rhythm production.

cs.HC↗

arXiv2025

Multi-period Mean-Buffered Probability of Exceedance in Defined Contribution Portfolio Optimization

We investigate multi-period mean-risk portfolio optimization for long-horizon Defined Contribution plans, focusing on buffered Probability of Exceedance (bPoE), a more intuitive, dollar-based alternative to Conditional Value-at-Risk (CVaR). We formulate both pre-commitment and time-consistent Mean-bPoE and Mean-CVaR portfolio optimization problems under realistic investment constraints (e.g., no leverage, no short selling) and jump-diffusion dynamics. These formulations are naturally framed as bilevel optimization problems, with an outer search over the shortfall threshold and an inner optimization over rebalancing decisions. We establish an equivalence between the pre-commitment formulations through a one-to-one correspondence of their scalarization optimal sets, while showing that no such equivalence holds in the time-consistent setting. We develop provably convergent numerical schemes for the value functions associated with both pre-commitment and time-consistent formulations of these mean-risk control problems. Using nearly a century of market data, we find that time-consistent Mean-bPoE strategies closely resemble their pre-commitment counterparts. In particular, they maintain alignment with investors' preferences for a minimum acceptable terminal wealth level-unlike time-consistent Mean-CVaR, which often leads to counterintuitive control behavior. We further show that bPoE, as a strictly tail-oriented measure, prioritizes guarding against catastrophic shortfalls while allowing meaningful upside exposure, making it especially appealing for long-horizon wealth security. These findings highlight bPoE's practical advantages for Defined Contribution investment planning.

q-fin.PM↗

arXiv2025

Extendable Planning via Multiscale Diffusion

Long-horizon planning is crucial in complex environments, but diffusion-based planners like Diffuser are limited by the trajectory lengths observed during training. This creates a dilemma: long trajectories are needed for effective planning, yet they degrade model performance. In this paper, we introduce this extendable long-horizon planning challenge and propose a two-phase solution. First, Progressive Trajectory Extension incrementally constructs longer trajectories through multi-round compositional stitching. Second, the Hierarchical Multiscale Diffuser enables efficient training and inference over long horizons by reasoning across temporal scales. To avoid the need for multiple separate models, we propose Adaptive Plan Pondering and the Recursive HM-Diffuser, which unify hierarchical planning within a single model. Experiments show our approach yields strong performance gains, advancing scalable and efficient decision-making over long-horizons.

cs.LG↗

arXiv2025

Identifying Sensitive Weights via Post-quantization Integral

Serving Large Language Models (LLMs) is costly. However, post-training weight quantization can address this problem by both compressing their sizes for limited memory and saving bandwidth for acceleration. As not all weight dimensions are equally important, those methods typically rely on a sensitivity metric, which indicates the element-wise influence of weights on loss function and is used to preprocess original weights for better quantization. In this work, we conduct an empirical study on the accuracy of the sensitivity metric, and find that existing gradient and Hessian based metrics are very inaccurate: they underestimate quantization's impact on the loss function by orders of magnitude, mainly due to the small convergence radius of local 2nd order approximation, \ie, gradient and Hessian term in Taylor's formula. To tackle this problem, we propose Post-quantization Integral (PQI), an accurate metric to estimate posterior sensitivity in a fine-grained manner. To leverage this accurate metric, we further propose ReQuant, a simple yet powerful framework that mainly consists of two Dense-and-Sparse detach components: self-adaptive outlier selection and step-wise significant weights detach. Results show that ReQuant boosts state-of-the-art post-training quantization methods, with a pronounced improvement of 2.66 perplexity gain on Llama 3.2 1B with QTIP.

cs.LG↗

arXiv2025

Memory Reviver: Supporting Photo-Collection Reminiscence for People with Visual Impairment via a Proactive Chatbot

Reminiscing with photo collections offers significant psychological benefits but poses challenges for people with visual impairment (PVI). Their current reliance on sighted help restricts the flexibility of this activity. In response, we explored using a chatbot in a preliminary study. We identified two primary challenges that hinder effective reminiscence with a chatbot: the scattering of information and a lack of proactive guidance. To address these limitations, we present Memory Reviver, a proactive chatbot that helps PVI reminisce with a photo collection through natural language communication. Memory Reviver incorporates two novel features: (1) a Memory Tree, which uses a hierarchical structure to organize the information in a photo collection; and (2) a Proactive Strategy, which actively delivers information to users at proper conversation rounds. Evaluation with twelve PVI demonstrated that Memory Reviver effectively facilitated engaging reminiscence, enhanced understanding of photo collections, and delivered natural conversational experiences. Based on our findings, we distill implications for supporting photo reminiscence and designing chatbots for PVI.

cs.HC↗

arXiv2024

Scalable Miniature On-chip Fourier Transform Spectrometer For Raman Spectroscopy

Miniaturized spectrometers for Raman spectroscopy have the potential to open up a new chapter in sensing. Raman spectroscopy is essential for material characterization and biomedical diagnostics, however, its weak signal and the need for sub-nanometer resolution pose challenges. Conventional spectrometers, with footprints proportional to optical throughput and resolution, are difficult to integrate into compact devices such as wearables. Waveguide-based Fourier Transform Spectrometers (FTS) enable compact spectrometers, and multi-aperture designs can achieve high throughput for applications such as Raman spectroscopy, however, experimental research in this domain remains limited. In this work, we present a multi-aperture SiN waveguide-based FTS overcoming these limitations and enabling Raman spectroscopy of isopropyl alcohol, glucose, Paracetamol, and Ibuprofen with enhanced throughput. Our spectrometer chip, fabricated on a 200 mm SiN wafer, with 160 edge-coupled waveguide apertures connected to an array of ultra-compact interferometers and a small footprint of just 1.6 mm x 4.8 mm, achieves a spectral range of 40 nm and a resolution of 0.5 nm. Experimental results demonstrate that least absolute shrinkage and selection operator (LASSO) regression significantly enhances Raman spectrum reconstruction. Our work on waveguide-based spectrometry paves the way for integrating accurate and compact Raman sensors into consumer electronics and space exploration instruments.

physics.optics↗

arXiv2024

Signage-Aware Exploration in Open World using Venue Maps

Current exploration methods struggle to search for shops or restaurants in unknown open-world environments due to the lack of prior knowledge. Humans can leverage venue maps that offer valuable scene priors to aid exploration planning by correlating the signage in the scene with landmark names on the map. However, arbitrary shapes and styles of the texts on signage, along with multi-view inconsistencies, pose significant challenges for robots to recognize them accurately. Additionally, discrepancies between real-world environments and venue maps hinder the integration of text-level information into the planners. This paper introduces a novel signage-aware exploration system to address these challenges, enabling the robots to utilize venue maps effectively. We propose a signage understanding method that accurately detects and recognizes the texts on signage using a diffusion-based text instance retrieval method combined with a 2D-to-3D semantic fusion strategy. Furthermore, we design a venue map-guided exploration-exploitation planner that balances exploration in unknown regions using directional heuristics derived from venue maps and exploitation to get close and adjust orientation for better recognition. Experiments in large-scale shopping malls demonstrate our method's superior signage recognition performance and search efficiency, surpassing state-of-the-art text spotting methods and traditional exploration approaches. Project website: https://sites.google.com/view/signage-aware-exploration.

cs.RO↗

arXiv2024

LeRF: Learning Resampling Function for Adaptive and Efficient Image Interpolation

Image resampling is a basic technique that is widely employed in daily applications, such as camera photo editing. Recent deep neural networks (DNNs) have made impressive progress in performance by introducing learned data priors. Still, these methods are not the perfect substitute for interpolation, due to the drawbacks in efficiency and versatility. In this work, we propose a novel method of Learning Resampling Function (termed LeRF), which takes advantage of both the structural priors learned by DNNs and the locally continuous assumption of interpolation. Specifically, LeRF assigns spatially varying resampling functions to input image pixels and learns to predict the hyper-parameters that determine the shapes of these resampling functions with a neural network. Based on the formulation of LeRF, we develop a family of models, including both efficiency-orientated and performance-orientated ones. To achieve interpolation-level efficiency, we adopt look-up tables (LUTs) to accelerate the inference of the learned neural network. Furthermore, we design a directional ensemble strategy and edge-sensitive indexing patterns to better capture local structures. On the other hand, to obtain DNN-level performance, we propose an extension of LeRF to enable it in cooperation with pre-trained upsampling models for cascaded resampling. Extensive experiments show that the efficiency-orientated version of LeRF runs as fast as interpolation, generalizes well to arbitrary transformations, and outperforms interpolation significantly, e.g., up to 3dB PSNR gain over Bicubic for x2 upsampling on Manga109. Besides, the performance-orientated version of LeRF reaches comparable performance with existing DNNs at much higher efficiency, e.g., less than 25% running time on a desktop GPU.

cs.CV↗