arXiv subjects

Heng Yao

Publications and source records attributed to Heng Yao.

9 recordsLinked to original sources

arXiv2026

Where Beam Search Loses Top-Ranked Items:First-Step Coverage in Generative Recommendation

Generative recommendation can reproduce a wider beam's ranked output with roughly half the prefix-capacity budget by reallocating retention capacity across semantic-ID positions. We identify a \emph{first-step coverage bottleneck}: across six full-test settings of the TIGER generative recommender spanning three domains and two identifier constructions, every pruned exact top-10 path is first lost at the initial decoding step. An exact-catalog reference isolates search losses and guides a controlled budget intervention. Moving width to the first step raises Search Recall@10 by 0.96--10.90 percentage points at Beam-50's unchanged budget. Across all 77,373 test users with residual-quantized variational autoencoder (RQ-VAE) tokenization, the reallocated schedule reproduces Beam-100's ordered top-10 at nominal budgets of 151 rather than 301, without retraining or changing identifiers or scores. Matched-output timing saves 6.35--6.39\% of decoder latency; profiling identifies the execution costs that limit the translation from reduced search work to runtime savings. Larger-candidate experiments reveal how output size changes retention requirements. These findings make budget placement an actionable serving decision: locate ranking loss, redirect search capacity, and preserve downstream inputs before increasing uniform beam width.

cs.IR↗

arXiv2026

Numerical Embeddings for CTR Prediction: A Taxonomy, Unified Benchmark, and Open-Source FuxiCTR Extension

Numerical embeddings map scalar features into vectors before a click-through rate (CTR) backbone combines the fields. To our knowledge, NumEmbBench is the first open benchmark dedicated to comparing numerical embeddings for CTR prediction. We classify 18 configurations into six mechanism-based families and evaluate them through one FuxiCTR interface. After equal-budget validation tuning, we compare frozen configurations on three datasets, nine backbones, and five seeds under a shared train-only z-score protocol. DEER has the best mean rank, but six methods lead at least one dataset--backbone setting. A matched native-scale test changes the winner in 11 of 27 settings, while paired latency measurements show that accuracy and inference cost need not move together. Code and per-seed results: https://github.com/YH-learning/NumEmbBench

cs.IR↗

arXiv2026

ScalarLens: Numerical Embeddings with Stable Coordinates and Contextual Responses for CTR Prediction

Numerical embeddings for click-through rate (CTR) prediction are built on a convenient but restrictive premise: a scalar has one representation. This premise conflates where a value lies with what it means for the current sample. On the Criteo validation split, the same numerical interval carries residual click evidence with opposite signs across categorical and numerical contexts, even after additive main effects are removed. Production pipelines compound this mismatch because externally normalized features require transformations and statistics to remain synchronized between training and serving. We introduce ScalarLens, a numerical embedding that preserves what a value is while adapting how it should be interpreted. A monotone local mesh constructs a stable coordinate from the focal scalar alone; bounded low-rank dynamics then produce a contextual response without moving that coordinate or replacing categorical tokens and the CTR backbone. In a 1,539-run primary evaluation covering 19 representations, three datasets, nine backbones, and three seeds, ScalarLens ranks first in 25 of 27 settings on original numerical scales and second in the remaining two. Matched ablations show that scale correction, additional local capacity, and generic conditioning do not reproduce the gain. A controlled study further recovers categorical, numerical, and mixed response mechanisms under context shift while the focal coordinate remains exactly invariant. A complete rerun under shared standardization retains significant advantages over DEER, DAES, and NaryDis, showing that the result is not explained by tolerance to raw scales alone. ScalarLens therefore recasts numerical embedding as a measurement problem: coordinates belong to values, while predictive responses belong to values in context.

cs.IR↗

arXiv2026

PRIME: Mitigating Subgroup Optimization Competition in Shared CTR Top Networks with Plug-in Residual Input-Conditioned Mixture of Expert

Click-through rate (CTR) models vary in feature-interaction design, yet their top networks usually remain a single multilayer perceptron shared by all examples. Heterogeneous user, item, and context subgroups therefore update the same parameters; weakly aligned learning signals make the aggregate gradient a compromise among competing directions. We study the competition on Avazu with 4 models and 4 semantic fields. Across all architectures, semantic subgroups show lower Top-NN gradient cosine similarity than random groups matched by sample size and label ratio, with reductions of 0.23-0.37. This competition motivates input-conditioned experts, but directly replacing an established Dense mapping changes its initial function, sharing pattern, and capacity, obscuring the source of gains. We introduce PRIME (Plug-in Residual Input-conditioned Mixture of Experts), a Dense-anchored mixture of low-rank residual experts. PRIME anchors the original prediction and uses zero-residual initialization to match the Dense baseline exactly at training onset. Input-dependent routing weights low-rank experts for example-specific logit corrections; multi-bag aggregation and EMA load biases stabilize conditional estimation. We evaluate PRIME on held-out Avazu and Criteo test sets across 13 CTR architectures and five paired seeds. Median paired AUC gains are +0.0022 and +0.0066, with LogLoss reductions of 0.0011 and 0.0081, respectively. On FiBiNET and DCNv2, PRIME outperforms APG in all ten seed-level AUC comparisons while using fewer parameters and lower inference latency on both backbones. These results show that function-preserving conditional residuals add input-dependent capacity while preserving the Dense path and its optimization stability. Code is available at https://github.com/YH-learning/PRIME.

cs.LG↗

arXiv2026

IV-CoT: Implicit Visual Chain-of-Thought for Structure-Aware Text-to-Image Generation

Unified multi-modal large language models (MLLMs) have achieved strong text-to-image generation quality, but still struggle with structure-aware prompt following, where object counts, spatial relations, attribute bindings, and coarse layouts must be preserved. We attribute this limitation in part to the entanglement of structural planning and appearance rendering within a single conditioning stream. To address this issue, we propose Implicit Visual Chain-of-Thought (IV-CoT), a latent visual reasoning framework for query-conditioned image generation. IV-CoT decomposes the visual conditioning queries into a structural-to-semantic cascade, where structural queries first form a latent visual plan and semantic queries then render appearance conditioned on this plan. To guide the structural queries, we introduce training-only sketch supervision, which encourages them to capture structure from sketches without requiring sketch extraction or intermediate decoding at inference time. IV-CoT performs implicit CoT reasoning in a single forward pass and achieves superior results on GenEval and T2I-CompBench. Visualizations and analyses demonstrate that the learned structural and semantic queries play complementary roles in structure-aware generation.

cs.CV↗

arXiv2024

Exploring the Effects of User-Agent and User-Designer Similarity in Virtual Human Design to Promote Mental Health Intentions for College Students

Virtual humans (i.e., embodied conversational agents) have the potential to support college students' mental health, particularly in Science, Technology, Engineering, and Mathematics (STEM) fields where students are at a heightened risk of mental disorders such as anxiety and depression. A comprehensive understanding of students, considering their cultural characteristics, experiences, and expectations, is crucial for creating timely and effective virtual human interventions. To this end, we conducted a user study with 481 computer science students from a major university in North America, exploring how they co-designed virtual humans to support mental health conversations for students similar to them. Our findings suggest that computer science students who engage in co-design processes of virtual humans tend to create agents that closely resemble them demographically--agent-designer demographic similarity. Key factors influencing virtual human design included age, gender, ethnicity, and the matching between appearance and voice. We also observed that the demographic characteristics of virtual human designers, especially ethnicity and gender, tend to be associated with those of the virtual humans they designed. Finally, we provide insights concerning the impact of user-designer demographic similarity in virtual humans' effectiveness in promoting mental health conversations when designers' characteristics are shared explicitly or implicitly. Understanding how virtual humans' characteristics serve users' experiences in mental wellness conversations and the similarity-attraction effects between agents, users, and designers may help tailor virtual humans' design to enhance their acceptance and increase their counseling effectiveness.

cs.HC↗

arXiv2022

In-vehicle alertness monitoring for older adults

Alertness monitoring in the context of driving improves safety and saves lives. Computer vision based alertness monitoring is an active area of research. However, the algorithms and datasets that exist for alertness monitoring are primarily aimed at younger adults (18-50 years old). We present a system for in-vehicle alertness monitoring for older adults. Through a design study, we ascertained the variables and parameters that are suitable for older adults traveling independently in Level 5 vehicles. We implemented a prototype traveler monitoring system and evaluated the alertness detection algorithm on ten older adults (70 years and older). We report on the system design and implementation at a level of detail that is suitable for the beginning researcher or practitioner. Our study suggests that dataset development is the foremost challenge for developing alertness monitoring systems targeted at older adults. This study is the first of its kind for a hitherto under-studied population and has implications for future work on algorithm development and system design through participatory methods.

cs.CV↗

arXiv2020

Perceptual Robust Hashing for Color Images with Canonical Correlation Analysis

In this paper, a novel perceptual image hashing scheme for color images is proposed based on ring-ribbon quadtree and color vector angle. First, original image is subjected to normalization and Gaussian low-pass filtering to produce a secondary image, which is divided into a series of ring-ribbons with different radii and the same number of pixels. Then, both textural and color features are extracted locally and globally. Quadtree decomposition (QD) is applied on luminance values of the ring-ribbons to extract local textural features, and the gray level co-occurrence matrix (GLCM) is used to extract global textural features. Local color features of significant corner points on outer boundaries of ring-ribbons are extracted through color vector angles (CVA), and color low-order moments (CLMs) is utilized to extract global color features. Finally, two types of feature vectors are fused via canonical correlation analysis (CCA) to prodcue the final hash after scrambling. Compared with direct concatenation, the CCA feature fusion method improves classification performance, which better reflects overall correlation between two sets of feature vectors. Receiver operating characteristic (ROC) curve shows that our scheme has satisfactory performances with respect to robustness, discrimination and security, which can be effectively used in copy detection and content authentication.

cs.CV↗

arXiv2020

Statistical Analysis of Signal-Dependent Noise: Application in Blind Localization of Image Splicing Forgery

Visual noise is often regarded as a disturbance in image quality, whereas it can also provide a crucial clue for image-based forensic tasks. Conventionally, noise is assumed to comprise an additive Gaussian model to be estimated and then used to reveal anomalies. However, for real sensor noise, it should be modeled as signal-dependent noise (SDN). In this work, we apply SDN to splicing forgery localization tasks. Through statistical analysis of the SDN model, we assume that noise can be modeled as a Gaussian approximation for a certain brightness and propose a likelihood model for a noise level function. By building a maximum a posterior Markov random field (MAP-MRF) framework, we exploit the likelihood of noise to reveal the alien region of spliced objects, with a probability combination refinement strategy. To ensure a completely blind detection, an iterative alternating method is adopted to estimate the MRF parameters. Experimental results demonstrate that our method is effective and provides a comparative localization performance.

cs.CV↗