Search arXiv⌕ Search

arXiv subjects

Fan Yang

Publications and source records attributed to Fan Yang.

At least 91 records · Page 5Linked to original sources
arXiv2026

Power and sample size calculations for causal mediation analysis with a binary mediator in randomized trials

Mediation analyses are increasingly conducted in randomized trials, but a sample size adequate for the total treatment effect may leave the natural indirect effect (NIE) or natural direct effect (NDE) substantially underpowered. Randomization does not extend to the mediator, so precision depends on the conditional mediator distribution and the mediator-outcome association, neither of which enters a total-effect calculation. Planning outside linear structural equation models is largely based on simulation under a fully specified data-generating mechanism rarely available at the design stage. This paper develops analytic power and sample size formulas for the NIE and NDE with a binary mediator and a continuous or binary outcome. Under standard identification assumptions, we focus on the ratio-of-mediator-probability weighting (RMPW) estimator that does not require an outcome model for effect estimation. We decompose the oracle variances of the RMPW estimators into components capturing mediator-probability-ratio variability, outcome variation, and their association, with an additional shared-arm covariance term for the NIE. Under a probit latent-index mediator and a working outcome model, these components are determined by a small number of interpretable design inputs rather than by the full joint distribution of covariates, mediator, and outcome. Simulations show that the analytic sample sizes closely match simulation-based benchmarks, attain the target power, and maintain type I error near the nominal level. An ACTG175 illustration shows how pilot data can calibrate the inputs.

stat.ME↗
arXiv2026

LaME: Learning to Think in Latent Space for Multimodal Embedding via Information Bottleneck

Reasoning-driven universal multimodal embedding has advanced rapidly by introducing Chain-of-Thought (CoT) reasoning into the embedding pipeline. Despite the strong performance across both general and complex tasks, this paradigm suffers from two core limitations: (i) autoregressive CoT reasoning incurs high computational cost, making it impractical for low-latency retrieval; and (ii) embedding performance is heavily coupled with CoT annotation quality, making large-scale training unreliable. These raise fundamental questions: Is textual CoT the optimal form of reasoning for embedding, and can effective embedding reasoning be accomplished in latent space? To this end, we propose LaME (Latent Reasoning Multimodal Embedding), which formulates embedding-oriented latent reasoning as a weakly supervised information bottleneck. LaME employs K learnable reason tokens as a fixed-capacity bottleneck, completing all reasoning within a single forward pass. The two weak supervision signals structurally decouple contrastive from autoregressive objectives and eliminate dependence on CoT annotations, while a two-stage training pipeline ensures stable convergence. Experiments on MMEB-v2 and MRMR show that LaME achieves competitive performance, surpassing some explicit CoT-based models, while delivering 60x faster inference than explicit CoT methods and 2x faster than latent baselines with throughput comparable to discriminative embedding models. Code is available at https://github.com/PeppaWu/LaME.

cs.CV↗
arXiv2026

How Mental Health Self-Disclosure Becomes Visible: Evidence from Eight Conditions on Reddit

People share mental health diagnoses on social media, yet how such language becomes visible around their self-disclosure, and whether community engagement tracks it, remain unexamined across conditions. We analyze 89,605 Reddit posts from 739 users across eight conditions, removing each user's diagnosis disclosure and aligning their surrounding posts to that anchor. Within the pre-disclosure year, language-visible burden was highest in the month before disclosure for six conditions, earlier for post-traumatic stress disorder and furthest from it for borderline personality disorder, and remained visible afterward rather than resolving. The theme Seeking Clinical Explanations showed the largest early-to-late difference before disclosure in five conditions, yet engagement rarely tracked what users wrote: only 9 of 360 language--engagement correlations survived correction. Disclosure is therefore a waypoint in an unevenly visible process, and we offer implications for community practice and platform design where engagement metrics do not reflect clinical need.

cs.HC↗
arXiv2026

LowRankArena: A Standardized Evaluation Platform for SVD-Based LLM Compression

SVD-based low-rank compression has become a fast-growing direction for reducing the memory and computational cost of large language models (LLMs). However, meaningful comparison across existing studies remains difficult as prior evaluations use varied benchmarks, inconsistent ratios, and diverse setups, often failing to isolate low-rank effects from auxiliary techniques. As a result, it remains unclear whether reported gains reflect method-level improvements or differences in evaluation protocol. This lack of comparability highlights the need for a unified, reproducible evaluation platform. To address this problem, we present LowRankArena, a standardized evaluation platform for SVD-based LLM compression. LowRankArena unifies task versions, uniform-precision compression budgets, comparison regimes, and inference measurements, and provides a reproducible pipeline with over 3 TiB released compressed checkpoints. Using LowRankArena, our aligned audit of five representative SVD methods reveals that prior findings are highly conditional under standardized protocols: clear leaders and performance tiers shift across backbones and keep ratios, multiple-choice accuracy can hide large perplexity degradation, and nominal low-rank savings yield workload-dependent and often limited end-to-end speedups. Our code is available at: https://github.com/Zishan-Shao/lowrankarena.git.

cs.CL↗
arXiv2026

AgentSpec: Speculative Decoding for Batch Inference of LLM Agents

Large language model (LLM)-based agent applications often incur high response time. Speculative decoding is a promising solution to improve the inference efficiency of LLM agents without impacting generation quality. However, state-of-the-art speculative decoding algorithms exhibit substantial speed degradation under large batch sizes, limiting their effectiveness to deploy in real-world agent applications. In this work, we first present a systematic analysis of speculative decoding for LLM agents and identify two dominant factors of speedup degradation: high rejection rate of speculative tokens, and under-utilization of dynamic token budgets.B ased on these observations, we propose AgentSpec, a speculative decoding algorithm that addresses the limitations of existing methods for LLM agents. AgentSpec incorporates structure-isolated drafting that constrains speculation to semantically coherent segments of the agent workflow, reducing the drafts of irrelevant semantic paths and achieving an extremely low rejection rate. Moreover, AgentSpec adopts redundancy-aware budget allocation that exploits agent-level information to better utilize the dynamically-free token budget during the agent inference. We implement and evaluate AgentSpec on five different workloads and four different models from four different LLM families in vLLM. Our results demonstrate the superiority of AgentSpec over state-of-the-arts.

cs.CL↗
arXiv2026

What Should a Large Language Model See? Physical Invariants as a Data Representation for PDE Discovery

Understanding how molecular interactions govern macroscopic behaviour is a central challenge in molecular sciences. However, conventional theory building cannot keep pace with the vast datasets modern experimentation routinely produces. Large language models offer a promising route to automating theory construction, but a spatiotemporal field cannot be directly placed in a prompt. Existing models generally learn about the data only through a score measuring how well each proposal fits it. Here we introduce data interpretation, a stage that measures the field into the quantities a theorist would consult and supplies them to the model as a direct input. On a benchmark of simulated fields, interpretation nearly triples the accuracy of recovered equations relative to showing the raw data, at negligible computational cost and without any training. By allowing a language model to read field data as a theorist does, data interpretation offers a practical route to automated field theory construction that can coevolve with experimentation.

cs.LG↗
arXiv2026

When Should a Robot Think? Resource-Aware Reasoning via Reinforcement Learning for Embodied Robotic Decision-Making

Embodied robotic systems increasingly rely on large language model (LLM)-based agents to support high-level reasoning, planning, and decision-making during interactions with the environment. However, invoking LLM reasoning introduces substantial computational latency and resource overhead, which can interrupt action execution and reduce system reliability. Excessive reasoning may delay actions, while insufficient reasoning often leads to incorrect decisions and task failures. This raises a fundamental question for embodied agents: when should the agent reason, and when should it act? In this work, we propose RARRL (Resource-Aware Reasoning via Reinforcement Learning), a hierarchical framework for resource-aware orchestration of embodied agents. Rather than learning low-level control policies, RARRL learns a high-level orchestration policy that operates at the agent's decision-making layer. This policy enables the agent to adaptively determine whether to invoke reasoning, which reasoning role to employ, and how much computational budget to allocate based on current observations, execution history, and remaining resources. Extensive experiments, including evaluations with empirical latency profiles derived from the ALFRED benchmark, show that RARRL consistently improves task success rates while reducing execution latency and enhancing robustness compared with fixed or heuristic reasoning strategies. These results demonstrate that adaptive reasoning control is essential for building reliable and efficient embodied robotic agents.

cs.RO↗
arXiv2026

Introducing the CP-plot for Causal Inference with Observational Studies

Under the canonical setting of observational studies for causal inference, we derive a set of exact representations for pairwise differences among weighted average treatment effects as covariances between the conditional average treatment effect and the propensity score, up to positive scaling factors. These covariance representations bridge the two core concepts in causal inference with observational studies. They immediately imply that (i) the average treatment effect is bracketed by the average treatment effects on the treated and on the controls, with the direction determined by the sign of the covariance between the conditional average treatment effect and the propensity score, and (ii) the average treatment effect under the overlap weight, the weight that is proportional to the conditional variance of the treatment given the covariates, is bracketed by the average treatment effects on the treated and controls when the corresponding covariances have a common sign within both the treated and control groups. We further extend these results to weighted local average treatment effects in the instrumental variable framework. Building on this theory, we recommend the ``CP-plot'' of the estimated conditional average treatment effect against the estimated propensity score, and implement it in the R package \texttt{CPplot}.

stat.ME↗
arXiv2026

Generative Embedding Benchmark: How Much Information Survives in a Dense Embedding?

Embeddings have emerged as a standard representational interface linking foundation models with downstream systems. Most embedding benchmarks assess representations through discriminative tasks or geometric criteria centered on separability in embedding space. However, strong performance on such evaluations does not establish whether content compressed into an embedding remains accessible to a downstream generator. To address this gap, we introduce the Generative Embedding Benchmark (GEB), in which a decoder answers questions using only a frozen embedding and question text, without access to the original image or intermediate visual features. Answer quality under this readout measures generative information: the answer-relevant content recoverable from an embedding. GEB includes a curated visual-question-answering dataset with a 1,800-item development split and a held-out 900-item test split covering natural images, scene text, and visual documents. Using a common decoder and training recipe, we evaluate seven public embedding models in visual-only and vision-language joint modes. On the test set, visual-only scores range from 28.25 to 33.21; with image-question joint encoding, all five VLM-based embedding models score higher, and the best reaches 65.56. Matched embeddings also outperform text-only inputs, zero embeddings, and shuffled embeddings. Natural-image information is much easier to recover than scene text or visual-document information, while a Qwen3-VL-2B reference with access to the original image reaches 84.30. Together, these results show that generative readout exposes information bottlenecks that separability-based evaluation does not capture.

cs.CV↗
arXiv2026

An Ergodic Spectral Decomposition Theorem for Singular Star Flows

For Axiom A diffeomorphisms and flows, Smale's Spectral Decomposition Theorem asserts that the non-wandering set decomposes into finitely many isolated hyperbolic basic sets, each given by a homoclinic class. For singular star flows, which may be viewed as "Axiom A flows with singularities", the corresponding spectral decomposition remains open and is known as the Spectral Decomposition Conjecture. We provide a positive answer to an ergodic formulation of this conjecture: $C^1$-open and dense among singular star flows with positive topological entropy, there is a unique measure of maximal entropy. More generally, we prove the uniqueness of equilibrium states for Hölder continuous potentials under a mild and natural pressure gap condition. We further establish that $C^1$-open and dense star flows are almost expansive and that the topological pressure of continuous potentials varies continuously with respect to the vector field in the $C^1$ topology. Our approach combines ergodic and geometric arguments adapted to the multi-singular setting. In this context, classical hyperbolic tools such as uniform local product structure or invariant splittings on the tangent bundle are no longer available. To overcome this, we develop new mechanisms to control the geometry of orbit segments and to produce transversal intersections on large subsets uniformly detected by good invariant measures. These ingredients allow us to extend classical arguments to the multi-singular setting through structural properties of equilibrium states combined with refined shadowing and specification at the level of invariant measures.

math.DS↗
arXiv2026

EATR-Stereo: Embodiment-Aware Token Routing of Paired Stereo Evidence for Humanoid Vision-Language-Action Control

Long-horizon humanoid vision--language--action (VLA) control with head-mounted stereo cameras requires visual interfaces that can exploit complementary views while maintaining compatibility with pretrained representations. Existing interfaces often discard complementary stereo evidence or fuse additional observations without preserving the native primary-view pathway and adapting auxiliary information to robot embodiment. We present EATR-Stereo, an embodiment-aware token-routing framework that retains primary-view tokens and constructs primary-aligned Cross-View Auxiliary Tokens (CVATs) by querying the synchronized auxiliary-view token sequence. A body-segmented proprioceptive encoder further conditions token-wise auxiliary usage on robot configuration history, enabling selective incorporation of stereo evidence during action generation. The routed auxiliary stream augments the language and primary-visual context of a pretrained VLA while keeping its vision--language model frozen. On a 33-DoF physical humanoid with a 37-D proprioceptive state, we evaluate nine configurations in over-100-s search--approach--grasp--place--return tasks. EATR-Stereo achieves 60.0% full-task success, 100.0% grasp success, and 80.0% stage success. Under severe asymmetric occlusion, it improves recovery to 80% compared with 30% for CVAT alone. Ablation studies further show the importance of preserving primary tokens and combining cross-view auxiliary features with structured proprioceptive routing. These results demonstrate that selectively routed paired stereo evidence improves spatial grounding for reliable long-horizon humanoid VLA control.

cs.RO↗
arXiv2026

TextRefine: Improving Textual Fidelity, Spatial Placement, and Glyph Rendering for Text Editing in Product Posters

Text editing in product posters entails inserting new text or replacing existing text while preserving product appearance, background content, and global composition. Despite recent progress in instruction-based image editing, general-purpose models remain unreliable in this setting: they often omit or incorrectly render the target text, place it over salient products or pre-existing content, and produce structurally distorted or visually inconsistent glyphs. We introduce \textbf{TextRefine}, a task-aligned post-training framework that combines supervised fine-tuning with operation-specific reward optimization to address these complementary failure modes. For text insertion, our text-span-level reward jointly assesses semantic fidelity and target-span coverage, penalizes spatial conflicts with products and existing text, and employs a gated structural constraint to preserve non-text regions. For text replacement, our glyph-level reward leverages the connectionist temporal classification (CTC) posterior of the target character to provide graded supervision for fine-grained defects, including missing strokes, structural deformations, and confusion among visually similar characters. We further introduce \textbf{OpenTextEdit}, a dataset comprising 100K images for text editing in product posters, with multi-text layouts, detailed text attributes, product masks, and challenging low-frequency characters. Extensive experiments on both insertion and replacement demonstrate that TextRefine consistently outperforms the evaluated image editing baselines in textual fidelity, placement reliability, and glyph quality while better preserving source-image content.

cs.CV↗
arXiv2026

Learning latent progression states from spatial heterogeneity in uterine histopathology

Tumor progression is accompanied by changes in architecture, morphology and microenvironmental organization, yet progression-associated heterogeneity is usually compressed into static diagnostic categories in histopathology. Here we present SpaTIE, a uterus-specific computational pathology framework that learns morphology-aware representations and organizes spatial histopathological heterogeneity into progression-associated tumor states. SpaTIE was developed using 10,426 uterine hematoxylin and eosin whole-slide images and evaluated in TCGA-UCEC and TCGA-UCS cohorts. The learned representations formed morphology manifolds, supported diagnostic, molecular and survival-related prediction tasks, and localized attention to informative tumor regions. Beyond supervised prediction, SpaTIE inferred tumor-state axes from cross-sectional morphology without temporal or molecular supervision. These morphology-derived states were spatially coherent and showed associations with clinicopathological variables and survival outcomes, while not simply recapitulating staging or diagnostic labels. Integrative multi-omics analyses linked the inferred states to DNA methylation, somatic copy-number variation, mutation, RNA-seq and RPPA profiles, highlighting molecular programs related to chromatin regulation, copy-number-associated structural variation, receptor tyrosine kinase signaling, cell adhesion, extracellular-matrix remodeling and metabolic adaptation. Progression-guided virtual perturbation further prioritized molecular features coupled to the morphology-derived state organization. Together, these findings suggest that uterine histopathology contains recoverable progression-associated tumor-state information and establish SpaTIE as a framework for connecting spatial morphology with multi-omics-informed tumor-state discovery.

cs.CV↗
arXiv2026

Predict Before Replay: Joint FEC and Flight Control for Reliable Scale-Up Links

Scale-up accelerator fabrics send latency-sensitive flits over serial links at hundreds of gigabits per second. Their reliability pipeline first relies on FEC, then detects residual failures and replays unacknowledged data. At these line rates, delayed feedback lets later flits enter the replay window before a residual failure is reported, so standard replay can amplify one corrupted flit into a suffix retransmission. This paper presents PREFACE, a pre-FEC controller for temporally correlated burst errors. A two-state Bayesian filter converts corrected-symbol observations into a next-flit burst posterior and jointly selects FEC strength with an outstanding-flit cap. We implement PREFACE in ns-3 with publicly verifiable UALink 200G 1.0 replay semantics. PREFACE improves goodput by 10.52%, lowers P99 latency by 50.75%, cuts replay by 47.52%, and improves modeled ring AllReduce by 13.1--27.0%.

cs.NI↗
arXiv2026

Completion-Path Credits: Multi-Resource Control for Scale-Up Fabrics

Scale-up fabrics connecting GPUs and AI accelerators carry tensor transfers together with remote reads, writes, atomics, and notifications over shared target-side receiver resources. Byte-denominated credits protect link buffers and streaming HBM traffic, but poorly represent small operations dominated by Atomic execution or response injection. This paper presents SemaCredit, a receiver controller that admits each remote-memory operation against a vector of target-resource demands and returns each component when its corresponding HBM, Atomic, or response stage completes. In a deterministic event simulator with multipath queues, eight HBM partitions, a serialized Atomic engine, and a response engine, SemaCredit matches a strong per-resource byte baseline on HBM-hotspot traffic while reducing small-operation P99 latency by 52.4% under Atomic contention and 10.2% under response incast. Application-shaped mixes show 57.7% and 14.5% P99 latency improvements for AllReduce-shaped and remote-read-shaped traffic while matching byte credits on HBM-dominated MoE traffic.

cs.NI↗
arXiv2026

Two-flag degenerations and real circles tangent to three conics

Three general plane conics admit $184$ complex tangent circles, and it had been conjectured that at most $136$ of them could be real. We construct an explicit strongly general triple of smooth conics over $\mathbb{Q}$ with exactly $160$ real tangent circles; the same count therefore occurs on a nonempty Euclidean chamber. The construction combines a fourfold splitting theorem for two flagged double-line degenerations with exact Sturm--Tarski, elimination, and interval certificates. We also show that the Grothendieck--Witt-valued count is $92\mathbb{H}$ and express its real local signs, up to a fixed orientation convention, in terms of curvature differences, residual intersection divisors, and contact normals. The exact-arithmetic code and certificate data are archived in the accompanying repository.

math.AG↗
arXiv2026

From Reasoning Depth to Reasoning Breadth: Evaluating Multi-Point Associative Reasoning in Large Language Models

Large language models (LLMs) have made substantial progress on reasoning tasks that require increasingly long and complex inferential chains. This progress primarily reflects reasoning depth. A complementary and comparatively unexamined capability is reasoning breadth: exploring multiple semantic directions in parallel and integrating the resulting clues into one coherent answer. We introduce MPAR-Bench, a bilingual English-Chinese benchmark that isolates reasoning breadth through multi-point associative reasoning. Inspired by the cooperative game Just One, each item asks a model to recover a hidden target from several independently generated, semantically diverse clues. We construct 1,000 items using a multi-agent clue-generation pipeline, embedding-based diversity filtering, and human verification. Only the answer space is drawn from public word lists, whereas every clue set is generated from scratch. Beyond exact-match accuracy, we evaluate models using accuracy, ANLS, embedding similarity, reasoning-trace verification, and four perturbations: clue masking, order shuffling, distractor injection, and multi-step clues. Across evaluated models, perturbations reduce accuracy by 9-18 percentage points in English and 5-12 percentage points in Chinese. Thinking mode improves standard-setting accuracy, especially in English, but does not consistently reduce sensitivity to perturbations. Case-level analysis also shows that extended reasoning can overturn an initially correct hypothesis. These results indicate that greater reasoning depth does not automatically confer robust reasoning breadth, and that reasoning breadth remains largely uncovered by current benchmarks.

cs.CL↗
arXiv2026

Agent Skills Can Be Harmful: An Empirical Study of Skill-Induced Failures in LLM Agents

Agent skills are the de facto mechanism for extending LLM agents with reusable guidance. A skill can shape the agent's task execution, including planning, tool use, problem-solving, and validation. Prior work reported mixed results of agent skills: some skills improve task success rates, while others have no effect, increase token use and execution time, and even reduce success rates. This paper presents a comprehensive analysis of skill-induced agent failures by attributing task failures and cost regressions to specific loaded skills. We introduce a differential analysis framework that attributes a failure or regression to a skill by comparing a target skill-guided run against a no-skill or semantically matched skill reference run that solves the same task, or solves it more cheaply. We instantiate this framework on SkillsBench and SWE-Skills-Bench, yielding 307 skill-induced failures, including 125 functional failures and 182 efficiency regressions. We also build SkillTriage, a taxonomy-guided attribution tool that normalizes paired cases, extracts differential evidence, and produces triage reports. Our major findings include: (1) Skill induced functional failures are rarely caused by obviously irrelevant skills; instead, seemingly relevant skills often make the agent incorrectly implement or omit task-required implementation elements. (2) Skill-induced efficiency regressions are not explained by prompt length alone. (3) The largest sources within Excessive Procedure are excessive verification and heavy implementation pipelines, contributing 67 and 30 cases, respectively. This shows that skills often turn validation checklists and construction recipes into mandatory work. Based on our findings, we propose research topics and tooling improvements for safer and more cost-aware skill reuse.

cs.AI↗