arXiv subjects

Linghan Zhang

Publications and source records attributed to Linghan Zhang.

4 recordsLinked to original sources

arXiv2026

Wuying-Browser-Agent: Real-World Centric Fundamental Long-Horizon Browser Agents

Browser agents perform well on short, clean demonstrations, but real deployment is fundamentally different: agents must sustain dozens of decisions on live websites while recovering from mistakes and navigating complex UIs. We argue that closing this gap requires alignment at every level of the pipeline, including execution, supervision, optimization, and evaluation, rather than scale alone. We present Wuying-Browser-Agent, a unified framework that addresses each of these levels. A structured browser harness provides stable execution primitives and decision-oriented context management. Reflection and UI-specialized Curriculum SFT (RUIC-SFT) explicitly trains on recovery trajectories and complex-UI interactions. Divergence-Aware Online GRPO (DAO-GRPO) improves long-horizon credit assignment through potential-based reward shaping and divergence-aware step weighting. Finally, we introduce BrowserBench, a bilingual real-web benchmark of 350 tasks averaging 37.9 steps, because most existing benchmarks are too short to expose long-horizon failure modes. Wuying-Browser-Agent-27B achieves 80.6\% on WebVoyager, 66.7\% on Online-Mind2Web, and 65.1\% on BrowserBench, establishing a new open-source state of the art on browser-use benchmarks. The same pipeline also transfers beyond browser use, demonstrating strong general agentic ability and reaching an average score of 73.8 on Tau2-Bench, Claw-Eval, and BFCL-v4.

cs.AI↗

arXiv2025

CARE: Contrastive Alignment for ADL Recognition from Event-Triggered Sensor Streams

The recognition of Activities of Daily Living (ADLs) from event-triggered ambient sensors is an essential task in Ambient Assisted Living, yet existing methods remain constrained by representation-level limitations. Sequence-based approaches preserve temporal order of sensor activations but are sensitive to noise and lack spatial awareness, while image-based approaches capture global patterns and implicit spatial correlations but compress fine-grained temporal dynamics and distort sensor layouts. Naive fusion (e.g., feature concatenation) fails to enforce alignment between sequence- and image-based representation views, underutilizing their complementary strengths. We propose Contrastive Alignment for ADL Recognition from Event-Triggered Sensor Streams (CARE), an end-to-end framework that jointly optimizes representation learning via Sequence-Image Contrastive Alignment (SICA) and classification via cross-entropy, ensuring both cross-representation alignment and task-specific discriminability. CARE integrates (i) time-aware, noise-resilient sequence encoding with (ii) spatially-informed and frequency-sensitive image representations, and employs (iii) a joint contrastive-classification objective for end-to-end learning of aligned and discriminative embeddings. Evaluated on three CASAS datasets, CARE achieves state-of-the-art performance (89.8% on Milan, 88.9% on Cairo, and 73.3% on Kyoto7) and demonstrates robustness to sensor malfunctions and layout variability, highlighting its potential for reliable ADL recognition in smart homes. We release our code at https://github.com/Jhziiiig/CARE.

cs.CV↗

arXiv2021

A Continuous Liveness Detection System for Text-independent Speaker Verification

Voice authentication is drawing increasing attention and becomes an attractive alternative to passwords for mobile authentication. Recent advances in mobile technology further accelerate the adoption of voice biometrics in an array of diverse mobile applications. However, recent studies show that voice authentication is vulnerable to replay attacks, where an adversary can spoof a voice authentication system using a pre-recorded voice sample collected from the victim. In this paper, we propose VoiceLive, a liveness detection system for both text-dependent and text-independent voice authentication on smartphones. VoiceLive detects a live user by leveraging the user's unique vocal system and the stereo recording of smartphones. In particular, utilizing the built-in gyroscope, loudspeaker, and microphone, VoiceLive first measures the smartphone's distance and angle from the user, then it captures the position-specific time-difference-of-arrival (TDoA) changes in a sequence of phoneme sounds to the two microphones of the phone, and uses such unique TDoA dynamic which doesn't exist under replay attacks for liveness detection. VoiceLive is practical as it doesn't require additional hardware but two-channel stereo recording that is supported by virtually all smartphones. Our experimental evaluation with 12 participants and different types of phones shows that VoiceLive achieves over 99% detection accuracy at around 1% Equal Error Rate (EER) on the text-dependent system and around 99% accuracy and 2% EER on the text-independent one. Results also show that VoiceLive is robust to different phone positions, i.e. the user is free to hold the smartphone with distinct distances and angles.

cs.CR↗

arXiv2021

A Continuous Liveness Detection for Voice Authentication on Smart Devices

Voice biometrics is drawing increasing attention as it is a promising alternative to legacy passwords for user authentication. Recently, a growing body of work shows that voice biometrics is vulnerable to spoofing through replay attacks, where an adversary tries to spoof voice authentication systems by using a pre-recorded voice sample collected from a genuine user. To this end, we propose VoiceGesture, a liveness detection solution for voice authentication on smart devices such as smartphones and smart speakers. VoiceGesture detects a live user by leveraging both the unique articulatory gesture of the user when speaking a passphrase and the audio hardware advances on these smart devices. Specifically, our system re-uses a pair of built-in speaker and microphone on a smart device as a Doppler radar, which transmits a high-frequency acoustic sound from the speaker and listens to the reflections at the microphone when a user speaks a passphrase. Then we extract Doppler shifts resulted from the user's articulatory gestures for liveness detection. VoiceGesture is practical as it requires neither cumbersome operations nor additional hardware but a speaker and a microphone commonly available on smart devices that support voice input. Our experimental evaluation with 21 participants and different smart devices shows that VoiceGesture achieves over 99% and around 98% detection accuracy for text-dependent and text-independent liveness detection, respectively. Results also show that VoiceGesture is robust to different device placements, low audio sampling frequency, and supports medium range liveness detection on smart speakers in various use scenarios like smart homes and smart vehicles.

cs.CR↗