arXiv subjects

Yuwen Tan

Publications and source records attributed to Yuwen Tan.

8 recordsLinked to original sources

arXiv2026

Dynamic Click Destinations: Personalizing Post Ad-Click Landing Experiences

Apps that monetize outbound (offsite) ad clicks have multiple post ad-click levers that influence user and business outcomes, including the landing page shown on the advertiser's site, whether the click resolves to a webpage or an app on the user's device, and whether it resolves to a native in-product surface or an external website. We refer to personalization over these levers as post ad-click destination automation. The optimal destination is heterogeneous across users, advertisers, placements, devices, and inferred intent, making static, one-size-fits-all destination policies suboptimal. We present Dynamic Click Destinations (DCD), a production personalization framework that selects a destination for each eligible offsite ad click to maximize Advertiser Value (AV), a value-weighted conversion outcome. DCD applies to any post-click destination decision where the system can choose among destination options at click time and downstream conversion feedback is available. We describe an evolution of learning objectives from post-conversion destination inference, to two-task destination-conditioned conversion prediction, to a single-task policy model that directly predicts which of two destinations is better for a given click, trained with Advantage Actor Critic (A2C)-inspired soft labels derived from a higher-capacity destination-conditioned "critic" model. In a web-vs-app case study, online A/B experiments at scale on a major advertising platform show that per-click personalization with DCD outperforms both static routing policies, with the single-task objective delivering the largest gains. These results highlight post ad-click destination as an important and under-explored lever in conversion optimization and demonstrate that personalized destination automation can outperform one-size-fits-all destination policies.

cs.IR↗

arXiv2026

DeLoop: Repeated Actions as Self-Supervision for Long-Horizon Agents

Language-model agents that run long tasks often redo work they have already done: they re-issue a tool call they made earlier and get back exactly the same observation, a step that adds no new information. We identify this failure as a loop. Loops appear natively in Qwen, Llama, and Gemini agents and become far more common once the agent's history is compressed to fit its context window. To reduce these loops, we propose DeLoop, a self-supervised training method. DeLoop builds on a simple observation: the agent's own rollout can provide the supervision needed to learn from its loops. We use a looping action as a negative example and the first later non-looping action in the same rollout as a positive example. DeLoop learns from these pairs through direct preference optimization and does not require task rewards, gold answers, or LLM judges. On CodeGym under compression, DeLoop reduces the share of Qwen3-8B tasks containing loops from 57.3% to 6.1%. It also uses about one third as many steps per task, with similar task success. On Qwen3-32B, DeLoop reduces loops and steps while raising task success from 37.7% to 44.8%.

cs.AI↗

arXiv2026

Natural Language Camera Movement Understanding

Understanding camera movement in natural language is critical for training and evaluating video generation models, among other applications. However, we demonstrate that existing vision-language models (VLMs) fail this task in surprising ways, frequently confusing translation with rotation, left with right, and object movement with camera movement. To address these limitations, we establish natural language camera movement understanding as a standalone research task. We introduce a two-level cinematographic taxonomy and an extensive, atomic benchmark featuring both real and synthetic videos. Furthermore, we curate a large-scale, multi-source training set enhanced by targeted camera movement augmentation. Our fine-tuned VLM-8B outperforms Gemini 3.1 Pro by 10% and 11% on our benchmark's real and synthetic videos, respectively. Despite these gains, a significant gap remains relative to human performance, underscoring the need to promote and facilitate future research on natural language camera movement understanding.

cs.CV↗

arXiv2026

Lifting Data-Tracing Machine Unlearning to Knowledge-Tracing for Foundation Models

Machine unlearning removes certain training data points and their influence from AI models (e.g., when a data owner revokes their consent to allow models to learn from the data). In this position paper, we propose to lift data-tracing machine unlearning to knowledge-tracing for foundation models (FMs). We support this position based on practical needs and insights from cognitive studies. Practically, tracing data cannot meet the diverse unlearning requests for FMs, which may be from regulators, enterprise users, product teams, etc., who have no access to FMs' massive training data. Instead, it is convenient for these parties to issue an unlearning request about the knowledge or capability FMs (should not) possess. Cognitively, knowledge-tracing unlearning aligns with how the human brain forgets more closely than tracing individual training data points does. We further discuss the nontrivial challenges in the knowledge-tracing machine unlearning paradigm. Finally, we provide a concrete case study about a vision-language FM to illustrate how an unlearner might instantiate the knowledge-tracing machine unlearning paradigm. Code is available at: https://1yuwen.github.io/Knowledge-Tracing-MU-Page.

cs.CV↗

arXiv2026

The LLM Bottleneck: Why Open-Source Vision LLMs Struggle with Hierarchical Visual Recognition

This paper reveals that many open-source large language models (LLMs) lack hierarchical knowledge about our visual world, unaware of even well-established biology taxonomies. This shortcoming makes LLMs a bottleneck for vision LLMs' hierarchical visual recognition (e.g., recognizing Anemone Fish but not Vertebrate). We arrive at these findings using about one million four-choice visual question answering (VQA) tasks constructed from six taxonomies and four image datasets. Interestingly, finetuning a vision LLM using our VQA tasks reaffirms LLMs' bottleneck effect because the VQA tasks improve the LLMs' hierarchical consistency more than the vision LLMs'. We conjecture that one cannot make open-source vision LLMs understand visual concepts hierarchically until LLMs possess corresponding taxonomy knowledge.

cs.CV↗

arXiv2025

Continual Adapter Tuning with Semantic Shift Compensation for Class-Incremental Learning

Class-incremental learning (CIL) aims to enable models to continuously learn new classes while overcoming catastrophic forgetting. The introduction of pre-trained models has brought new tuning paradigms to CIL. In this paper, we revisit different parameter-efficient tuning (PET) methods within the context of continual learning. We observe that adapter tuning demonstrates superiority over prompt-based methods, even without parameter expansion in each learning session. Motivated by this, we propose incrementally tuning the shared adapter without imposing parameter update constraints, enhancing the learning capacity of the backbone. Additionally, we employ feature sampling from stored prototypes to retrain a unified classifier, further improving its performance. We estimate the semantic shift of old prototypes without access to past samples and update stored prototypes session by session. Our proposed method eliminates model expansion and avoids retaining any image samples. It surpasses previous pre-trained model-based CIL methods and demonstrates remarkable continual learning capabilities. Experimental results on five CIL benchmarks validate the effectiveness of our approach, achieving state-of-the-art (SOTA) performance.

cs.CV↗

arXiv2025

Confidence Optimization for Probabilistic Encoding

Probabilistic encoding introduces Gaussian noise into neural networks, enabling a smooth transition from deterministic to uncertain states and enhancing generalization ability. However, the randomness of Gaussian noise distorts point-based distance measurements in classification tasks. To mitigate this issue, we propose a confidence optimization probabilistic encoding (CPE) method that improves distance reliability and enhances representation learning. Specifically, we refine probabilistic encoding with two key strategies: First, we introduce a confidence-aware mechanism to adjust distance calculations, ensuring consistency and reliability in probabilistic encoding classification tasks. Second, we replace the conventional KL divergence-based variance regularization, which relies on unreliable prior assumptions, with a simpler L2 regularization term to directly constrain variance. The method we proposed is model-agnostic, and extensive experiments on natural language classification tasks demonstrate that our method significantly improves performance and generalization on both the BERT and the RoBERTa model.

cs.LG↗

arXiv2021

Coarse-To-Fine Incremental Few-Shot Learning

Different from fine-tuning models pre-trained on a large-scale dataset of preset classes, class-incremental learning (CIL) aims to recognize novel classes over time without forgetting pre-trained classes. However, a given model will be challenged by test images with finer-grained classes, e.g., a basenji is at most recognized as a dog. Such images form a new training set (i.e., support set) so that the incremental model is hoped to recognize a basenji (i.e., query) as a basenji next time. This paper formulates such a hybrid natural problem of coarse-to-fine few-shot (C2FS) recognition as a CIL problem named C2FSCIL, and proposes a simple, effective, and theoretically-sound strategy Knowe: to learn, normalize, and freeze a classifier's weights from fine labels, once learning an embedding space contrastively from coarse labels. Besides, as CIL aims at a stability-plasticity balance, new overall performance metrics are proposed. In that sense, on CIFAR-100, BREEDS, and tieredImageNet, Knowe outperforms all recent relevant CIL/FSCIL methods that are tailored to the new problem setting for the first time.

cs.CV↗