Searcharxiv⌕ Search

arXiv subjects

Kevin Li

Publications and source records attributed to Kevin Li.

At least 19 recordsLinked to original sources
arXiv2026

Contractible Rips complexes of groups via metric gluings

We say that a finitely generated group equipped with a finite generating set is of type $\mathsf{R}$ if the Rips complex is contractible for sufficiently large scales. We show that the class of groups of type $\mathsf{R}$ is closed under taking graphs of groups with finite edge groups and, in particular, under taking free products. We prove that all two-dimensional right-angled Artin groups with the standard generating set are of type $\mathsf{R}$. Furthermore, we observe that products of finitely generated free abelian groups and finite groups are of type $\mathsf{R}$.

math.GR↗
arXiv2026

Checked-In Secret Detection: Strings Are All You Need

Hardcoded secrets in source code pose critical security vulnerabilities which can be easily exploited by malicious adversaries. Existing regex-based detection approaches suffer from fundamental limitations, as secrets often lack identifiable patterns, resulting in poor precision and recall. Recent studies have explored context-aware detection methods, as surrounding code can reveal the purpose of candidate strings. However, these methods confront three key challenges: (1) obfuscation robustness where models over-rely on easily obfuscated identifiers, (2) cross-language generalization difficulties due to uneven training data distribution, and (3) lengthy and noisy context that introduces excessive irrelevant tokens and slows inference. We observe that strings serve as a critical information source for code semantics, offering superior contextual density, obfuscation robustness, and language independence. Based on this insight, we propose StringGroup, a novel context extraction algorithm that mines strings surrounding potential secrets. By introducing a relatively simple modification to existing patterns that narrows the analysis specifically to string literals, the method achieves significant gains. With only 33.2% of the original context, it preserves over 80% of semantic information and significantly improves the signal-to-noise ratio for secret detection. We further design a context-aware secret detection tool, Secretron, based on StringGroup methods and Transformer model. Evaluation on the SecretBench dataset demonstrates high accuracy with 98.74% F1-score and strong robustness under obfuscation and cross-language scenarios, outperforming state-of-the-art LLM-based baselines. We deploy our tool in real-world environments and successfully detect 48 previously unknown secret keys from 26 applications, demonstrating the practical effectiveness of our approach.

cs.CR↗
arXiv2026

GazeBehavior Annotation Toolkit (GBAT): AI-powered toolkit for automatic annotation of egocentric eye-tracking and video data of child-caregiver interaction

Video recordings of child-caregiver interactions enable investigation of attentional dynamics during naturalistic behavior. Such multimodal recording also allows researchers to examine how attention interacts with action and language use in real time. However, manual annotation of such data is time-consuming. Here, we introduce GazeBehavior Annotation Toolkit, a deep-learning-based toolkit designed to facilitate three key processes in data preprocessing and feature extraction: post-hoc synchronization across multiple videos, semi-automatic annotation of gaze target categories, and categorization of participants' poses and hand actions. This toolkit improves the efficiency and scalability of feature extraction from human egocentric eye-tracking and video data. Such improvement is critical in supporting large-scale and longitudinal investigations of attentional dynamics and naturalistic behavior in human early development.

cs.CV↗
arXiv2026

Adaptive Time Series Reasoning via Segment Selection

Time series reasoning tasks often start with a natural language question and require targeted analysis of a time series. Evidence may span the full series or appear in a few short intervals, so the model must decide what to inspect. Most existing approaches encode the entire time series into a fixed representation before inference, regardless of whether or not the entire sequence is relevant. We introduce ARTIST, which formulates time-series reasoning as a sequential decision problem. ARTIST interleaves reasoning with adaptive temporal segment selection. It adopts a controller-reasoner architecture and uses reinforcement learning to train the controller role to select informative segments and the reasoner role to generate segment-conditioned reasoning traces and final answers. During inference, the model actively acquires task-relevant information instead of relying on a static summary of the full sequence. We use a novel hierarchical policy optimization approach for post-training that allows the model to excel in both segment selection and question-answering behavior. We evaluate ARTIST on six time-series reasoning benchmarks and compare it with large language models, vision-language models, and prior time-series reasoning systems. ARTIST improves average accuracy by 6.46 absolute percentage points over the strongest baseline. The largest gains appear on rare event localization and multi-segment reasoning tasks. Supervised fine-tuning improves performance, and reinforcement learning provides additional gains by optimizing question-adaptive segment selection. These results show that selective data use drives effective time-series reasoning.

cs.LG↗
arXiv2026

tLoRA: Efficient Multi-LoRA Training with Elastic Shared Super-Models

As Low-Rank Adaptation (LoRA) becomes the standard approach for efficiently fine-tuning large language models (LLMs), shared clusters increasingly execute many concurrent LoRA training jobs over the same frozen backbone. While recent advances enable batching (co-locating) multiple adapters during serving, efficient training-time co-location of heterogeneous LoRA adapters presents unique challenges. Jobs often differ in adapter rank, batch size, and resource allocation, and na\"ive batching can introduce synchronization stalls, communication overheads, and per-job slowdowns that are worse than executing independently. We introduce tLoRA, a framework that enables efficient batch training of multiple LoRA jobs. tLoRA fuses adapters that share the same base model into an elastic shared super-model, exploiting existing distributed training frameworks to derive parallelism plans that share resources effectively. At the kernel level, tLoRA employs a fused LoRA kernel that adaptively reconstructs low-rank computation tiles and schedules rank-aware nano-batches to maximize overlap between computation and communication across adapters. At the scheduling layer, tLoRA incorporates an online, residual-capacity-aware scheduler that adaptively groups jobs to maximize collective throughput. Evaluations using real-world cluster traces demonstrate that tLoRA improves training throughput by 1.2--1.8x, job training completion time by 2.3--5.4x, and GPU utilization by 37%.

cs.LG↗
arXiv2025

A note on higher coherence of graphs of groups

For $n\in \mathbb{N}$, a group is called $n$-coherent if every subgroup of type $\mathsf{F}_n$ is of type $\mathsf{F}_{n+1}$. For $n\ge 1$, we observe that graphs of groups with $n$-coherent vertex groups and virtually poly-cyclic edge groups are $n$-coherent. We deduce the $n$-coherence of certain right-angled Artin groups.

math.GR↗
arXiv2025

UniFusion: Vision-Language Model as Unified Encoder in Image Generation

Although recent advances in visual generation have been remarkable, most existing architectures still depend on distinct encoders for images and text. This separation constrains diffusion models' ability to perform cross-modal reasoning and knowledge transfer. Prior attempts to bridge this gap often use the last layer information from VLM, employ multiple visual encoders, or train large unified models jointly for text and image generation, which demands substantial computational resources and large-scale data, limiting its accessibility.We present UniFusion, a diffusion-based generative model conditioned on a frozen large vision-language model (VLM) that serves as a unified multimodal encoder. At the core of UniFusion is the Layerwise Attention Pooling (LAP) mechanism that extracts both high level semantics and low level details from text and visual tokens of a frozen VLM to condition a diffusion generative model. We demonstrate that LAP outperforms other shallow fusion architectures on text-image alignment for generation and faithful transfer of visual information from VLM to the diffusion model which is key for editing. We propose VLM-Enabled Rewriting Injection with Flexibile Inference (VERIFI), which conditions a diffusion transformer (DiT) only on the text tokens generated by the VLM during in-model prompt rewriting. VERIFI combines the alignment of the conditioning distribution with the VLM's reasoning capabilities for increased capabilities and flexibility at inference. In addition, finetuning on editing task not only improves text-image alignment for generation, indicative of cross-modality knowledge transfer, but also exhibits tremendous generalization capabilities. Our model when trained on single image editing, zero-shot generalizes to multiple image references further motivating the unified encoder design of UniFusion.

cs.CV↗
arXiv2025

PRAXA: A Grammar for What-If Analysis

What-if analysis is widely used to explore hypothetical scenarios and evaluate alternative pathways to desired results. However, current approaches are fragmented: systems implement what-if capabilities under diverse terminologies with different analytic techniques. Such fragmentation limits expressiveness, impedes flexible composition and reuse of workflows, and hinders tighter integration with AI. We present PRAXA, a compositional grammar of what-if analysis derived from recurring patterns across 141 publications in visual analytics and HCI venues. PRAXA formulates three primitives: (1) data, defining variables under analysis, (2) model, specifying predictive mechanisms, and (3) interaction operations-pairs of user actions and system responses that execute analyses. We encode PRAXA into a declarative specification language, PSL. To evaluate PRAXA, we first show expressiveness by reconstructing representative workflows from prior work as structured compositions, exposing the predominant focus on single-step rather than multi-step reasoning. Second, we demonstrate composability by revealing that capabilities described under distinct terminologies share the same grammatical structure with different parameterizations, and that new multi-step workflows emerge through composition. Third, we illustrate PSL as an intermediate representation for translating natural-language what-if queries into executable interactive interfaces, enabling inspection, validation, and more transparent AI integration. By unifying diverse what-if approaches as a grammar, PRAXA provides a foundation for analyzing, composing, and supporting workflows in next-generation what-if systems.

cs.HC↗
arXiv2025

Beyond Parameters: Exploring Virtual Logic Depth for Scaling Laws

Scaling large language models typically involves three dimensions: depth, width, and parameter count. In this work, we explore a fourth dimension, \textbf{virtual logical depth} (VLD), which increases effective algorithmic depth without changing parameter count by reusing weights. While parameter reuse is not new, its role in scaling has been underexplored. Unlike recent test-time methods that scale token-wise, VLD alters the internal computation graph during training and inference. Through controlled experiments, we obtain three key insights. (1) \textit{Knowledge capacity vs. parameters}: at fixed parameter count, VLD leaves knowledge capacity nearly unchanged, while across models capacity still scales with parameters. (2) \textit{Reasoning vs. reuse}: properly implemented VLD substantially improves reasoning ability \emph{without} more parameters, decoupling reasoning from size. This suggests a new scaling path beyond token-wise test-time methods. (3) \textit{Robustness and generality}: reasoning gains persist across architectures and reuse schedules, showing VLD captures a general scaling behavior. These results provide insight into future scaling strategies and raise a deeper question: does superintelligence require ever-larger models, or can it be achieved by reusing parameters and increasing logical depth? We argue many unknown dynamics in scaling remain to be explored. Code is available at https://anonymous.4open.science/r/virtual_logical_depth-8024/.

cs.AI↗
arXiv2025

Prompting Decision Transformers for Zero-Shot Reach-Avoid Policies

Offline goal-conditioned reinforcement learning methods have shown promise for reach-avoid tasks, where an agent must reach a target state while avoiding undesirable regions of the state space. Existing approaches typically encode avoid-region information into an augmented state space and cost function, which prevents flexible, dynamic specification of novel avoid-region information at evaluation time. They also rely heavily on well-designed reward and cost functions, limiting scalability to complex or poorly structured environments. We introduce RADT, a decision transformer model for offline, reward-free, goal-conditioned, avoid region-conditioned RL. RADT encodes goals and avoid regions directly as prompt tokens, allowing any number of avoid regions of arbitrary size to be specified at evaluation time. Using only suboptimal offline trajectories from a random policy, RADT learns reach-avoid behavior through a novel combination of goal and avoid-region hindsight relabeling. We benchmark RADT against 3 existing offline goal-conditioned RL models across 11 tasks, environments, and experimental settings. RADT generalizes in a zero-shot manner to out-of-distribution avoid region sizes and counts, outperforming baselines that require retraining. In one such zero-shot setting, RADT achieves 35.7% improvement in normalized cost over the best retrained baseline while maintaining high goal-reaching success. We apply RADT to cell reprogramming in biology, where it reduces visits to undesirable intermediate gene expression states during trajectories to desired target states, despite stochastic transitions and discrete, structured state dynamics.

cs.LG↗
arXiv2025

Semantic De-boosting in e-commerce Query Autocomplete

In ecommerce search, query autocomplete plays a critical role to help users in their shopping journey. Often times, query autocomplete presents users with semantically similar queries, which can impede the user's ability to find diverse and relevant results. This paper proposes a novel strategy to enhance this service by refining the presentation of typeahead suggestions based on their semantic similarity. Our solution uniquely demotes semantically equivalent queries using an embedding similarity of query suggestions at runtime. This strategy ensures only distinct and varied queries are prioritized, thereby promoting more diverse suggestions for users. To maintain comprehensive query coverage, we incorporate this deduplication process within the query suggestion reranking step. This approach ensures that the broad spectrum of possible queries remains available to users, while eliminating the redundancy and repetitiveness in the suggestion list. In extending this work, we propose using the distance between query embeddings to offer even more diverse suggestions to users using an algorithm similar to maximal marginal relevance. This approach will further ensure the delivery of non-redundant, unique, and pertinent suggestions to users, thus enriching their search experience. We evaluated our method through rigorous AB testing, demonstrating substantial improvements in key metrics. Notably, we observed a statistically significant rise in the search Add-to-Cart rate, signifying an enhanced user engagement and conversion rate. Furthermore, we observed a statistically significant decrease in clicks to ATC, implying that the feature improved the efficiency of the customer's product search journey. Finally, we also noticed a marked reduction in the null page view rate, indicating the increased pertinence and efficiency of user search sessions.

cs.IT↗
arXiv2025

Combination of open covers with $\pi_1$-constraints

Let~$G$ be a group and let~$\mathcal{F}$ be a family of subgroups of~$G$. The generalised Lusternik--Schnirelmann category~$\operatorname{cat}_\mathcal{F}(G)$ is the minimal cardinality of covers of~$BG$ by open subsets with fundamental group in~$\mathcal{F}$. We prove a combination theorem for~$\operatorname{cat}_\mathcal{F}(G)$ in terms of the stabilisers of contractible $G$-CW-complexes. As applications for the amenable category, we obtain vanishing results for the simplicial volume of gluings of manifolds (along not necessarily amenable boundaries) and of cyclic branched coverings. Moreover, we deduce an upper bound for Farber's topological complexity, generalising an estimate for amalgamated products of Dranishnikov--Sadykov.

math.AT↗
arXiv2025

The Serre spectral sequence in bounded cohomology

We construct the analogue of the Serre spectral sequence for the bounded cohomology of simplicial sets with seminormed local coefficients. As applications, we obtain a (non-isometric) generalization of Gromov's mapping theorem and some partial results on the simplicial volume of manifold bundles.

math.AT↗
arXiv2025

Exploiting Concavity Information in Gaussian Process Contextual Bandit Optimization

The contextual bandit framework is widely used to solve sequential optimization problems where the reward of each decision depends on auxiliary context variables. In settings such as medicine, business, and engineering, the decision maker often possesses additional structural information on the generative model that can potentially be used to improve the efficiency of bandit algorithms. We consider settings in which the mean reward is known to be a concave function of the action for each fixed context. Examples include patient-specific dose-response curves in medicine and expected profit in online advertising auctions. We propose a contextual bandit algorithm that accelerates optimization by conditioning the posterior of a Bayesian Gaussian Process model on this concavity information. We design a novel shape-constrained reward function estimator using a specially chosen regression spline basis and constrained Gaussian Process posterior. Using this model, we propose a UCB algorithm and derive corresponding regret bounds. We evaluate our algorithm on numerical examples and test functions used to study optimal dosing of Anti-Clotting medication.

stat.ML↗
arXiv2025

A note on finiteness properties of vertex stabilisers

We prove a criterion for the geometric and algebraic finiteness properties of vertex stabilisers of $G$-CW-complexes, given the finiteness properties of the group $G$ and of the stabilisers of positive dimensional cells. This generalises a result of Haglund--Wise for groups acting on trees to higher dimensions. As an application, for $n\ge 2$, we deduce the existence of uncountably many quasi-isometry classes of one-ended groups that are of type $\mathsf{FP}_n$ and not of type $\mathsf{FP}_{n+1}$.

math.GR↗
arXiv2025

Controllable Sequence Editing for Biological and Clinical Trajectories

Conditional generation models for longitudinal sequences can produce new or modified trajectories given a conditioning input. However, they often lack control over when the condition should take effect (timing) and which variables it should influence (scope). Most methods either operate only on univariate sequences or assume that the condition alters all variables and time steps. In scientific and clinical settings, interventions instead begin at a specific moment, such as the time of drug administration or surgery, and influence only a subset of measurements while the rest of the trajectory remains unchanged. CLEF learns temporal concepts that encode how and when a condition alters future sequence evolution. These concepts allow CLEF to apply targeted edits to the affected time steps and variables while preserving the rest of the sequence. We evaluate CLEF on 8 datasets spanning cellular reprogramming, patient health, and sales, comparing against 9 state-of-the-art baselines. CLEF improves immediate sequence editing accuracy by 16.28% (MAE) on average against their non-CLEF counterparts. Unlike prior models, CLEF enables one-step conditional generation at arbitrary future times, outperforming their non-CLEF counterparts in delayed sequence editing by 26.73% (MAE) on average. We test CLEF under counterfactual inference assumptions and show up to 62.84% (MAE) improvement on zero-shot conditional generation of counterfactual trajectories. In a case study of patients with type 1 diabetes mellitus, CLEF identifies clinical interventions that generate realistic counterfactual trajectories shifted toward healthier outcomes.

cs.LG↗