arXiv subjects

Jiang Li

Publications and source records attributed to Jiang Li.

At least 19 recordsLinked to original sources

arXiv2026

APRiL: An Open-ended Benchmark for Chinese Social Awareness in Large Language Models

Large language models are increasingly used in social interactions, where appropriate responses require recognizing implicit cues, inferring mental states, identifying conflicts, and considering potential consequences. However, existing benchmarks often focus on specific social abilities, constrained task formats, or particular interaction settings, offering limited coverage of social awareness across diverse contexts. To address this gap, we introduce APRiL, an open-ended benchmark for Chinese social awareness with instance-specific rubrics. APRiL contains 2,196 Social Awareness Evaluation Instances across six categories and two complementary tasks: Social Scenario Analysis evaluates models' understanding of social cues, mental states, and conflict structures, while Socially Aware Response assesses whether they can translate this understanding into appropriate, natural responses. Experiments on 12 models from six families reveal a consistent understanding-to-response gap. Higher-quality scenario analyses improve responses, while weak analyses may reduce response quality. APRiL provides a systematic resource for advancing socially aware language models in Chinese contexts. Our data and code are available at https://anonymous.4open.science/r/APRiL-05DE/.

cs.CL↗

arXiv2026

AIPOCH Open-Science: An Open-Source AI Research Workbench with Reproducible .science Packages

AI-assisted research leaves instructions, computations, artifact revisions and checks whose relationships matter when another researcher receives the work. We present AIPOCH Open-Science, an open-source research workbench with portable .science session packages, and specify an exchange contract for these recorded relationships. Version and assessment targets remain attributable while receiving-workspace identifiers can be remapped, eligible contents can be omitted explicitly, and imported history remains separate from a writable fork. A controlled session tested an incorrect first version, a reviewed correction, an unreviewed revision and a planned but unexecuted step. We compared complete and partial exports with source records extracted independently of the exporter, then inspected imports into separate projects within the same installation. Both conditions retained the tested version and review associations, kept the unreviewed revision distinct from earlier assessments and represented excluded content as unavailable. A documentation-rich ordinary directory also supplied evidence for all seven static inspection questions. A separate registry case connected a table to its recorded task, execution and review; offline execution of its retained script and fixed input reproduced two saved CSV tables byte for byte. The study contributes an implemented exchange contract and a controlled examination of its preservation rules, content-selection constraints and interpretation boundaries.

cs.SE↗

arXiv2026

Acmite: Mitigating Gender Bias in LLMs through Concept-Guided Mutual Information

Large language models (LLMs) can reproduce social stereotypes from their training data, motivating extensive research on model debiasing. However, existing methods often rely on explicit biased examples or predefined group-term substitutions, making them sensitive to wording and less effective at capturing stereotype concepts shared across diverse contexts. More importantly, they typically suppress biased outputs without explicitly modeling the statistical dependence between model outputs and the underlying stereotype concepts. We propose Acmite, a lightweight concept-guided framework for targeted and selective debiasing. Acmite represents stereotypes as structured semantic concepts and uses maximal marginal relevance (MMR) to select diverse concepts for debiasing. Inspired by mutual information minimization, it approximates this dependence with token-level KL divergence while preserving task semantics. A lightweight LoRA adapter is trained with the base model frozen and activated at inference time only when the input is sufficiently similar to stereotype-related concepts; otherwise, the original model is used directly. We evaluate Acmite on BBQ, CrowS-Pairs, and StereoSet, and assess general capability preservation on ARC-Challenge, GSM8K, and PIQA. Experiments across three LLMs show that Acmite effectively mitigates gender bias across complementary evaluation formats while maintaining competitive performance on bias-unrelated tasks. Anonymous code and data are available at https://anonymous.4open.science/r/Acmite-18E2/.

cs.CL↗

arXiv2026

Distance-KV: Exploiting Relative Distance for Efficient Long-Context Inference

The memory usage and decoding latency of LLM inference grow rapidly with context length. To reduce these costs, key-value (KV) cache compression methods selectively retain cached states based on token importance or differences in attention patterns across heads. However, we discover that retrieval capability varies substantially with relative distance, even within the same attention head. To exploit this structure, we introduce Distance-KV, which learns a static KV retention pattern over the joint space of layers, attention heads, and relative distances. The pattern is learned offline with the language model frozen and reused across inputs to prune and compact the KV cache without online importance scoring. Across three backbone models and four long-context benchmarks, Distance-KV consistently achieves the best overall performance among competing KV cache compression methods, exceeding the strongest compression baseline by up to 9.3 points on RULER at 128K. On Llama-3.1-8B-Instruct at 128K, Distance-KV reduces KV cache memory by 65.4% and achieves a $1.66\times$ decoding speedup relative to Dense. Together, these results identify relative distance as an important structural dimension for understanding how LLMs retrieve information over long contexts and for designing more efficient inference methods.

cs.LG↗

arXiv2026

Bourgain--Brezis--Mironescu formula for BV functions on arbitrary open sets and applications

In this paper, we establish a Bourgain--Brezis--Mironescu formula for BV functions on arbitrary open sets for a general class of radial mollifiers. Building on this formula, we identify the $Γ$-limit of the improved nonlocal energies with respect to $L^1$-convergence as a constant multiple of the total variation. As a consequence, we obtain a characterization of BV functions in terms of the finiteness of the corresponding nonlocal energies. We also establish local $L^1$-compactness on arbitrary open sets and global $L^1$-compactness on bounded open sets under a natural uniform integrability assumption. In particular, except for the global compactness result, all these results hold on arbitrary open sets, with no assumptions on boundedness, connectedness, or boundary regularity. Counterexamples show that for global compactness, both the boundedness of the domain and the uniform integrability assumption are essential. The proofs rely primarily on one-dimensional restrictions of BV functions and a concentration argument for radial mollifiers.

math.AP↗

arXiv2026

Evaluating Predicted Densities, Hamiltonians, and Density Matrices as Periodic SCF Initializers

Learned electronic states are usually evaluated by prediction error, even though their intended use is to accelerate the self-consistent-field (SCF) loop of density functional theory (DFT). We ask whether lower offline error actually yields a better SCF initializer. We construct $ρ$HD-43K, a 43,851-crystal DFT corpus with aligned charge-density, Hamiltonian, and density-matrix labels, and extend the solver workflow to inject all three predicted states. The closed-loop benchmark compares direct converged-state prediction against residual prediction from solver-native references on a frozen test set of non-magnetic crystals, using paired convergence, iteration, and SCF-loop timing measurements. Injecting the exact converged density matrix or Hamiltonian as an oracle upper bound cuts the median SCF count from 16 to one, revealing substantial acceleration headroom. Among learned inputs, direct charge-density prediction (Charge3Net-E3) accelerates about 91% of paired crystals, saves a median of three SCF iterations, and yields a $1.18\times$ SCF-loop speedup. In contrast, the tested residual-density and matrix initializers do not consistently improve over the standard superposition-of-atomic-densities baseline. These results establish that target-space accuracy alone is insufficient: a learned initializer must also be compatible with the nonlinear solver trajectory, and its utility must be measured in the loop.

cond-mat.mtrl-sci↗

arXiv2026

MMTryon: Multi-Modal Multi-Reference Control for High-Quality Fashion Generation

This paper introduces MMTryon, a multi-modal multi-reference VIrtual Try-ON (VITON) framework, which can generate high-quality compositional try-on results by taking a text instruction and multiple garment images as inputs. Our MMTryon addresses three problems overlooked in prior literature: 1) \textbf{Support of multiple try-on items.} Existing methods are commonly designed for single-item try-on tasks (e.g., upper/lower garments, dresses). 2) \textbf{Specification of dressing style}. Existing methods are unable to customize dressing styles based on instructions (e.g., zipped/unzipped, tuck-in/tuck-out, etc.) 3) \textbf{Segmentation Dependency}. They further heavily rely on category-specific segmentation models to identify the replacement regions, with segmentation errors directly leading to significant artifacts in the try-on results. To address the first two issues, our MMTryon introduces a novel multi-modality and multi-reference attention mechanism to combine the garment information from reference images and dressing-style information from text instructions. Besides, to remove the segmentation dependency, MMTryon uses a parsing-free garment encoder and leverages a novel scalable data generation pipeline to convert existing VITON datasets to a form that allows MMTryon to be trained without requiring any explicit segmentation. Extensive experiments on high-resolution benchmarks and in-the-wild test sets demonstrate MMTryon's superiority over existing SOTA methods both qualitatively and quantitatively. MMTryon's impressive performance on multi-item and style-controllable virtual try-on scenarios and its ability to try on any outfit in a large variety of scenarios from any source image, opens up a new avenue for future investigation in the fashion community.

cs.CV↗

arXiv2026

Do Large Language Models Perform Well on Comprehending Poetic Logic in Modern Chinese Poetry?

Large Language Models (LLMs) have achieved significant progress across a wide range of natural language processing (NLP) tasks, yet their ability to understand literary texts, particularly modern Chinese poetry, remains largely unexplored. The unique literary characteristics of modern Chinese poetry necessitate a distinct form of reasoning for effective comprehension. Unlike conventional texts that convey clear information, the unique "poetic logic" of modern Chinese poetry requires a holistic reasoning approach that goes beyond superficial semantic analysis to be understood. However, current evaluation paradigms largely ignore this critical dimension. To address this gap, we propose Peony, the first benchmark specifically designed for evaluating the poetic logic of modern Chinese poetry. We define poetic logic as four tasks across three levels, namely stanza, line, and imagery, and systematically evaluate and analyze six mainstream LLMs based on Peony. We evaluate these models under both non-thinking and thinking configurations. The experimental results reveal the limitations of current LLMs in understanding the poetic logic of modern Chinese poetry and validate the effectiveness and necessity of Peony. Our data and code will be available.

cs.CL↗

arXiv2026

Quantization Degradation in Large Language Models: A Signal-Noise Perspective

Post-training quantization reduces the deployment cost of large language models, yet how severely a quantized model degrades is not determined by bit-width alone. We systematically study weight-only post-training quantization across bit-widths, quantization methods, model scales and downstream tasks on multiple model families. We observe that such degradation varies substantially across these factors: 4-bit quantization usually preserves performance, 2-bit often causes broad degradation, and at 3-bit, degradation becomes apparent but varies markedly with task type, quantization method and model scale. To explain this variability, we use the signal-to-noise ratio (SNR) to measure how strongly quantization perturbs full-precision representations. We trace degradation back to two linked processes: how quantization errors arise within individual modules, and how they accumulate across layers. First, a source SNR decomposition shows that newly introduced errors depend on three factors: the magnitude of the weight error, the strength of the task-specific signal, and how strongly the quantization error aligns with task-specific activations. Different factors affect these components in distinct ways. Second, a cross-layer propagation analysis shows that these errors can be attenuated, preserved, or amplified as they pass across layers, and that larger models benefit from weaker error amplification. Together, these results establish that quantization degradation is governed by how errors are introduced at the source and how they accumulate across the network.

cs.AI↗

arXiv2026

A Heuristic Perspective on Debiasing Language Models

Language models (LMs) often acquire various biases during pre-training and may express them in interactions, potentially causing social harm. Existing methods often rely on counterfactual augmentation or representation projection. These strategies remain limited in practice due to their high computational costs and difficulty in scaling to larger models. Additionally, many of these strategies require manual data annotation, narrowing their scope to specific cultures and bias categories. To overcome these limitations, we propose HEIMAT, a HEurIstic-style autoMATic debiasing framework for LMs. HEIMAT consists of two main steps: bias disclosure and debiasing fine-tuning. In the first step, it uses simple templates to construct heuristic prompts, which are applied to reveal model biases and generate corresponding context prompts. In the second step, it fine-tunes the model by minimizing the Jensen-Shannon divergence of predictions on these context prompts to reduce bias. Extensive experiments show that HEIMAT effectively mitigates bias in different cultures while maintaining the model's natural language understanding (NLU) performance.

cs.CL↗

arXiv2026

WearWow: Native 2K Multi-Garment Virtual Try-On via Adaptive Token Packing and Preference Alignment

Synthesizing native 2K multi-garment virtual try-on is a formidable frontier in digital fashion, critically bottlenecked by two fundamental limitations: the O(N^2) memory explosion induced by 2k conditions, and the spectral bias of diffusion models that over-smooths high-frequency fabric details. We present WearWow, an end-to-end, mask-free generative framework that pioneers ultra-high-resolution multi-garment synthesis. To mitigate the memory explosion , we propose Adaptive 2D Token Packing (ATP). ATP leverages inherent garment sparsity to algorithmically pack heterogeneous items onto a unified 2D canvas and prune uninformative background tokens, minimizing the effective sequence length and subsequent memory overhead while rigorously preserving 2D spatial priors. To rectify texture degradation, we introduce the Multi-dimensional Try-on Reward (MTR) system. MTR synergizes a Semantic Guidance Reward to explicitly drive tactile restoration with a Cloth Distribution Reward to implicitly anchor the physical distribution, a joint formulation that effectively mitigates the severe reward hacking. Furthermore, we curate WearWow-2K, an extreme-quality dataset comprising native 2K triplets, providing physically correct spatial interactions that naturally empower the model's mask-free generation. Extensive experiments demonstrate that WearWow establishes a new state-of-the-art, exceeding existing commercial baselines in native 2K multi-garment synthesis.

cs.CV↗

arXiv2026

Large language models reshape the language of science

Scientific language is a central infrastructure of knowledge production, but it remains unclear whether large language models (LLMs) are altering not only how scientists write, but also how scientific knowledge is communicated and accessed. Here we analyze 21.36 million scientific abstracts published between 2020 and 2024, together with historical records from major journals, to trace recent changes in the language of science. We identify a marked turning point in 2024, when scientific writing shows a sharp increase in lexical complexity alongside a decline in syntactic complexity. This shift is pervasive across disciplines and journal tiers, and is more pronounced in texts by scholars working in non-native English contexts, especially those from language backgrounds that differ more typologically from English. Controlled polishing experiments confirm that LLMs reproduce this pattern by favoring more lexically dense and syntactically compressed expression. We further show why this linguistic shift matters: it may widen the distance between scientific discourse and public-facing language, while also helping scholars in non-native English contexts navigate English-language publishing requirements. These findings suggest that LLMs may broaden participation in scientific authorship while narrowing the accessibility of scientific communication, making them a new force in the linguistic infrastructure of science.

cs.CL↗

arXiv2026

Extreme-Scale Atomistic Simulation of Real-Temperature Magnetic Skyrmion Dynamics by Coupled Spin-Lattice Modeling

Real-temperature topological magnetic dynamics in functional materials is governed by coupled lattice and spin evolution, yet remains inaccessible to predictive simulation at device-relevant scales. As a flagship example, thermally driven helix-to-skyrmion transformation in FeGe requires atomistic resolution, explicit lattice motion, and micrometer-scale domains to resolve device-scale topological texture formation. We combine a spin-constrained density-functional-theory-trained neuro-evolution potential with a structure-preserving spin-lattice integrator within one machine-learned framework. Architecture-specific optimizations, kernel fusion, SVE2 vectorization, and NUMA-aware data layout deliver a seven orders-of-magnitude speedup over prior spin-aware methods. Deployed on LineShine exascale supercomputer, the full application scales to 12.45 million CPU cores with 89.7% weak-scaling efficiency, enabling simulations of 1.34 trillion atoms and an equal number of spins while reaching 48.5 PFLOPS in double precision. The simulations directly resolve real-temperature skyrmion nucleation and reorganization at previously inaccessible scales, establishing a new regime for predictive simulation of coupled spin-lattice topological magnetic dynamics.

cs.DC↗

arXiv2026

A Unified and Reproducible Experimentation Framework for Speech Understanding

Speech foundation models and Speech LLMs have advanced speech understanding, yet deployment-oriented model selection is hindered by non-comparable evaluations caused by mismatched post-processing, and by training results that are hard to reproduce across data scales and pipelines. We present SURE, a unified experimentation framework that standardizes prediction formats, normalization, and scoring. SURE evaluates strong systems across paradigms, from conventional pipelines to Speech LLMs, on representative tasks under realistic acoustic and linguistic stressors. Beyond evaluation, SURE introduces an agent-assisted training conversion flow that maps paper and code into versioned, runnable training pipelines under a unified protocol on matched open-data subsets. Overall, SURE improves comparability and reproducibility for deployment-oriented evaluation.

eess.AS↗

arXiv2026

AtelierEval: Agentic Evaluation of Humans & LLMs as Text-to-Image Prompters

Text-to-image (T2I) systems increasingly rely on upstream prompters, either humans or multimodal large language models (MLLMs), to translate user intent into detailed prompts. Yet current benchmarks fix the prompt and only evaluate T2I models, leaving the prompting proficiency of this upstream component entirely unmeasured. We introduce AtelierEval, the first unified benchmark that quantifies prompting proficiency across 360 expert-crafted tasks. Grounded in a cognitive view, it spans three task categories and instantiates tasks using a taxonomy of real-world challenges, with a dual interface for both humans and MLLMs. To enable scalable and reliable evaluation, we propose AtelierJudge, a skill-based, memory-augmented agentic evaluator. It produces subjective and objective scores for prompt-image pairs, achieving a Spearman correlation of 0.79 with human experts, approaching human performance. Extensive experiments benchmark 8 MLLMs against 48 human users across 4 T2I backends, validate AtelierEval as a robust diagnostic tool, and reveal the superiority of mimicry over planning, advocating for an image-augmented direction for future prompters. Our work is released to support future research.

cs.AI↗

arXiv2026

Training-Inference Consistent Segmented Execution for Long-Context LLMs

Transformer-based large language models face severe scalability challenges in long-context generation due to the computational and memory costs of full-context attention. Under practical computation and memory constraints, many inference-efficient long-context methods improve efficiency by adopting bounded-context or segment-level execution only during inference, while continuing to train models under full-context attention, resulting in a mismatch between training and inference execution and state-transition semantics. Based on this insight, we propose a training-inference consistent segment-level generation framework, in which training and inference follow the same segment-level forward execution semantics. During training, consistency with inference is enforced by restricting gradient propagation to KV states carried over from the immediately preceding segment, while permitting head-specific access to past KV states during the forward pass without involving them in gradient propagation. Across long-context benchmarks, our approach achieves performance comparable to full-context attention, while achieving competitive latency-memory trade-offs against strong inference-efficient baselines, and substantially improving scalability at very long context lengths (e.g., approximately 6x lower peak prefill memory at 128K compared to full-context attention with FlashAttention).

cs.CL↗

arXiv2026

Boxing inequalities for relative fractional perimeter and fractional Poincaré-type inequalities on John domains with the BBM factor

For $0<δ,τ<1$ and $1\le s\le \frac{n}{n-δ}$, we prove that for a given $s$-John domain $Ω\subset \mathbb{R}^n$, the following Boxing inequality holds for every Lebesgue measurable set $U\subsetΩ$ with $|U|/|Ω|\leγ<1$: \[ \mathcal{H}^{s(n-δ)}_{\infty}(U\setminus\mathcal{N}_U)\le C(1-δ)\int_Ω\int_{|x-y|<τ\operatorname{dist}(y,\partialΩ)}\frac{|χ_U(x)-χ_U(y)|}{|x-y|^{n+δ}}\,dx\,dy, \] where $\mathcal{H}^{s(n-δ)}_{\infty}(U)$ denotes the $s(n-δ)$-dimensional Hausdorff content of $U$, $\mathcal{N}_U$ is a set of Lebesgue measure zero and the constant $C$ depends only on $n,τ,s,γ$, the John constant and the diameter of $Ω$. Moreover, we establish the functional formulation of the above Boxing inequality and discuss the equivalence between these two formulations. Based on the Boxing inequality, we prove the fractional Poincaré--Wirtinger trace inequality on $s$-John domains, of which the fractional Sobolev--Poincaré inequality and fractional Hardy-type inequality are special cases. Notably, we prove all of the aforementioned inequalities with the Bourgain--Brezis--Mironescu (BBM) factor $1-δ$. Furthermore, with the aid of the Bourgain--Brezis--Mironescu formula, we recover the Poincaré--Wirtinger trace inequality. Finally, by showing that, under the separation property, any domain supporting the Boxing inequality is necessarily a John domain, we conclude that the John domain condition is essentially sharp for the above inequalities. All the above inequalities with the BBM factor are new even for Lipschitz domains.

math.FA↗

arXiv2026

Task-Stratified Knowledge Scaling Laws for Post-Training Quantized Large Language Models

Post-Training Quantization (PTQ) is a critical strategy for efficient Large Language Models (LLMs) deployment. However, existing scaling laws primarily focus on general performance, overlooking crucial fine-grained factors and how quantization differentially impacts diverse knowledge capabilities. To address this, we establish Task-Stratified Knowledge Scaling Laws. By stratifying capabilities into memorization, application, and reasoning, we develop a framework that unifies model size, bit-width, and fine-grained factors: group size and calibration set size. Validated on 293 diverse PTQ configurations, our framework demonstrates strong fit and cross-architecture consistency. It reveals distinct sensitivities across knowledge capabilities: reasoning is precision-critical, application is scale-responsive, and memorization is calibration-sensitive. We highlight that in low-bit scenarios, optimizing these fine-grained factors is essential for preventing performance collapse. These findings provide an empirically-backed foundation for designing knowledge-aware quantization strategies.

cs.CL↗