Searcharxiv⌕ Search

arXiv subjects

Ben Wu

Publications and source records attributed to Ben Wu.

17 recordsLinked to original sources
arXiv2026

Reflections over the Sea: Reconfigurable Intelligent Surface for Maritime Self-Powered Communications

Maritime communication is becoming a vital component of 6G networks, driven by the rapid expansion of the maritime economy. However, existing technologies face critical challenges in signal coverage, availability, and robustness, especially under harsh sea conditions. This paper proposes a novel framework for the maritime Internet-of-Things (IoT) communications that leverages the reconfigurable intelligent surface (RIS) mounted on offshore infrastructures, such as wind turbines, to enhance coverage and reliability. To capture dynamic maritime environment, a near-ocean-surface channel model is developed considering the impact of sea waves. In addition, a wave energy harvesting (EH) system is designed to self-power IoT sensors for data acquisition, processing, and transmission. To support real-time adaptation, channel state information is continuously measured to optimize RIS reflection parameters and maximize multi-user communication rates. Simulation results show that the proposed system significantly improves IoT communication performance by over 20%, under harsh sea conditions.

cs.IT↗
arXiv2025

Dense SAE Latents Are Features, Not Bugs

Sparse autoencoders (SAEs) are designed to extract interpretable features from language models by enforcing a sparsity constraint. Ideally, training an SAE would yield latents that are both sparse and semantically meaningful. However, many SAE latents activate frequently (i.e., are \emph{dense}), raising concerns that they may be undesirable artifacts of the training procedure. In this work, we systematically investigate the geometry, function, and origin of dense latents and show that they are not only persistent but often reflect meaningful model representations. We first demonstrate that dense latents tend to form antipodal pairs that reconstruct specific directions in the residual stream, and that ablating their subspace suppresses the emergence of new dense features in retrained SAEs -- suggesting that high density features are an intrinsic property of the residual space. We then introduce a taxonomy of dense latents, identifying classes tied to position tracking, context binding, entropy regulation, letter-specific output signals, part-of-speech, and principal component reconstruction. Finally, we analyze how these features evolve across layers, revealing a shift from structural features in early layers, to semantic features in mid layers, and finally to output-oriented signals in the last layers of the model. Our findings indicate that dense latents serve functional roles in language model computation and should not be dismissed as training noise.

cs.LG↗
arXiv2025

GOD model: Privacy Preserved AI School for Personal Assistant

Personal AI assistants (e.g., Apple Intelligence, Meta AI) offer proactive recommendations that simplify everyday tasks, but their reliance on sensitive user data raises concerns about privacy and trust. To address these challenges, we introduce the Guardian of Data (GOD), a secure, privacy-preserving framework for training and evaluating AI assistants directly on-device. Unlike traditional benchmarks, the GOD model measures how well assistants can anticipate user needs-such as suggesting gifts-while protecting user data and autonomy. Functioning like an AI school, it addresses the cold start problem by simulating user queries and employing a curriculum-based approach to refine the performance of each assistant. Running within a Trusted Execution Environment (TEE), it safeguards user data while applying reinforcement and imitation learning to refine AI recommendations. A token-based incentive system encourages users to share data securely, creating a data flywheel that drives continuous improvement. Specifically, users mine with their data, and the mining rate is determined by GOD's evaluation of how well their AI assistant understands them across categories such as shopping, social interactions, productivity, trading, and Web3. By integrating privacy, personalization, and trust, the GOD model provides a scalable, responsible path for advancing personal AI assistants. For community collaboration, part of the framework is open-sourced at https://github.com/PIN-AI/God-Model.

cs.CR↗
arXiv2024

Efficient Annotator Reliability Assessment and Sample Weighting for Knowledge-Based Misinformation Detection on Social Media

Misinformation spreads rapidly on social media, confusing the truth and targeting potentially vulnerable people. To effectively mitigate the negative impact of misinformation, it must first be accurately detected before applying a mitigation strategy, such as X's community notes, which is currently a manual process. This study takes a knowledge-based approach to misinformation detection, modelling the problem similarly to one of natural language inference. The EffiARA annotation framework is introduced, aiming to utilise inter- and intra-annotator agreement to understand the reliability of each annotator and influence the training of large language models for classification based on annotator reliability. In assessing the EffiARA annotation framework, the Russo-Ukrainian Conflict Knowledge-Based Misinformation Classification Dataset (RUC-MCD) was developed and made publicly available. This study finds that sample weighting using annotator reliability performs the best, utilising both inter- and intra-annotator agreement and soft-label training. The highest classification performance achieved using Llama-3.2-1B was a macro-F1 of 0.757 and 0.740 using TwHIN-BERT-large.

cs.LG↗
arXiv2024

Confidence Regulation Neurons in Language Models

Despite their widespread use, the mechanisms by which large language models (LLMs) represent and regulate uncertainty in next-token predictions remain largely unexplored. This study investigates two critical components believed to influence this uncertainty: the recently discovered entropy neurons and a new set of components that we term token frequency neurons. Entropy neurons are characterized by an unusually high weight norm and influence the final layer normalization (LayerNorm) scale to effectively scale down the logits. Our work shows that entropy neurons operate by writing onto an unembedding null space, allowing them to impact the residual stream norm with minimal direct effect on the logits themselves. We observe the presence of entropy neurons across a range of models, up to 7 billion parameters. On the other hand, token frequency neurons, which we discover and describe here for the first time, boost or suppress each token's logit proportionally to its log frequency, thereby shifting the output distribution towards or away from the unigram distribution. Finally, we present a detailed case study where entropy neurons actively manage confidence in the setting of induction, i.e. detecting and continuing repeated subsequences.

cs.LG↗
arXiv2024

Comparison between parameter-efficient techniques and full fine-tuning: A case study on multilingual news article classification

Adapters and Low-Rank Adaptation (LoRA) are parameter-efficient fine-tuning techniques designed to make the training of language models more efficient. Previous results demonstrated that these methods can even improve performance on some classification tasks. This paper complements the existing research by investigating how these techniques influence the classification performance and computation costs compared to full fine-tuning when applied to multilingual text classification tasks (genre, framing, and persuasion techniques detection; with different input lengths, number of predicted classes and classification difficulty), some of which have limited training data. In addition, we conduct in-depth analyses of their efficacy across different training scenarios (training on the original multilingual data; on the translations into English; and on a subset of English-only data) and different languages. Our findings provide valuable insights into the applicability of the parameter-efficient fine-tuning techniques, particularly to complex multilingual and multilabel classification tasks.

cs.CL↗
arXiv2023

Don't Waste a Single Annotation: Improving Single-Label Classifiers Through Soft Labels

In this paper, we address the limitations of the common data annotation and training methods for objective single-label classification tasks. Typically, when annotating such tasks annotators are only asked to provide a single label for each sample and annotator disagreement is discarded when a final hard label is decided through majority voting. We challenge this traditional approach, acknowledging that determining the appropriate label can be difficult due to the ambiguity and lack of context in the data samples. Rather than discarding the information from such ambiguous annotations, our soft label method makes use of them for training. Our findings indicate that additional annotator information, such as confidence, secondary label and disagreement, can be used to effectively generate soft labels. Training classifiers with these soft labels then leads to improved performance and calibration on the hard label test set.

cs.CL↗
arXiv2023

SheffieldVeraAI at SemEval-2023 Task 3: Mono and multilingual approaches for news genre, topic and persuasion technique classification

This paper describes our approach for SemEval-2023 Task 3: Detecting the category, the framing, and the persuasion techniques in online news in a multi-lingual setup. For Subtask 1 (News Genre), we propose an ensemble of fully trained and adapter mBERT models which was ranked joint-first for German, and had the highest mean rank of multi-language teams. For Subtask 2 (Framing), we achieved first place in 3 languages, and the best average rank across all the languages, by using two separate ensembles: a monolingual RoBERTa-MUPPETLARGE and an ensemble of XLM-RoBERTaLARGE with adapters and task adaptive pretraining. For Subtask 3 (Persuasion Techniques), we train a monolingual RoBERTa-Base model for English and a multilingual mBERT model for the remaining languages, which achieved top 10 for all languages, including 2nd for English. For each subtask, we compared monolingual and multilingual approaches, and considered class imbalance techniques.

cs.CL↗
arXiv2021

Wideband photonic interference cancellation based on free space optical communication

We propose and experimentally demonstrate an interference management system that removes wideband wireless interference by using photonic signal processing and free space optical communication. The receiver separates radio frequency interferences by upconverting the mixed signals to optical frequencies and processing the signals with the photonic circuits. Signals with GHz bandwidth are processed and separated in real-time. The reference signals for interference cancellation are transmitted in a free space optical communication link, which provides large bandwidth for multi-band operation and accelerates the mixed signal separation process by reducing the dimensions of the un-known mixing matrix. Experimental results show that the system achieves 30dB real-time cancellation depth with over 6GHz bandwidth. Multiple radio frequency bands can be processed at the same time with a single system. In addition, multiple radio frequency bands can be processed at the same time with a single system.

eess.SP↗
arXiv2021

Wideband photonic blind source separation with optical pulse sampling

We propose and experimentally demonstrate an optical pulse sampling method for photonic blind source separation. The photonic system processes and separates wideband signals based on the statistical information of the mixed signals and thus the sampling frequency can be orders of magnitude lower than the bandwidth of the signals. The ultra-fast optical pulse functions as a tweezer that collects samples of the signals at very low sampling rates, and each sample is short enough to maintain the statistical properties of the signals. The low sampling frequency reduces the workloads of the analog to digital conversion and digital signal processing systems. In the meantime, the short pulse sampling maintains the accuracy of the sampled signals, so the statistical properties of the undersampling signals are the same as the statistical properties of the original signals. With the optical pulses generated from a mode-locked laser, the optical pulse sampling system is able to process and separate mixed signals with bandwidth over 100GHz and achieves a dynamic range of 30dB.

eess.SP↗
arXiv2021

Hodge Numbers of O'Grady 6 Via Ngô Strings

We give an alternative computation of the Betti and Hodge numbers for manifolds of $OG6$ type using the method of Ngô Strings introduced by de Cataldo, Rapagnetta, and Saccà.

math.AG↗
arXiv2020

A Topological Characterization of the Middle Perversity Intersection Complex for Arbitrary Complex Algebraic Varieties

For an arbitrary complex algebraic variety which is not necessarily pure dimensional, the intersection complex can be defined as the direct sum of the Deligne-Goresky-MacPherson intersection complexes of each irreducible component. We give two axiomatic topological characterizations of the middle perversity direct sum intersection complex, one stratification dependent and the other stratification independent. To accomplish this, we show that this direct sum intersection complex can be constructed using Deligne's construction in the more general context of topologically stratified spaces. A consequence of these characterizations is the invariance of this direct sum intersection complex under homeomorphisms.

math.AG↗
arXiv2015

Dynamical laser spike processing

Novel materials and devices in photonics have the potential to revolutionize optical information processing, beyond conventional binary-logic approaches. Laser systems offer a rich repertoire of useful dynamical behaviors, including the excitable dynamics also found in the time-resolved "spiking" of neurons. Spiking reconciles the expressiveness and efficiency of analog processing with the robustness and scalability of digital processing. We demonstrate that graphene-coupled laser systems offer a unified low-level spike optical processing paradigm that goes well beyond previously studied laser dynamics. We show that this platform can simultaneously exhibit logic-level restoration, cascadability and input-output isolation---fundamental challenges in optical information processing. We also implement low-level spike-processing tasks that are critical for higher level processing: temporal pattern detection and stable recurrent memory. We study these properties in the context of a fiber laser system, but the addition of graphene leads to a number of advantages which stem from its unique properties, including high absorption and fast carrier relaxation. These could lead to significant speed and efficiency improvements in unconventional laser processing devices, and ongoing research on graphene microfabrication promises compatibility with integrated laser platforms.

physics.optics↗
arXiv2014

SIMPEL: Circuit model for photonic spike processing laser neurons

We propose an equivalent circuit model for photonic spike processing laser neurons with an embedded saturable absorber---a simulation model for photonic excitable lasers (SIMPEL). We show that by mapping the laser neuron rate equations into a circuit model, SPICE analysis can be used as an efficient and accurate engine for numerical calculations, capable of generalization to a variety of different laser neuron types found in literature. The development of this model parallels the Hodgkin--Huxley model of neuron biophysics, a circuit framework which brought efficiency, modularity, and generalizability to the study of neural dynamics. We employ the model to study various signal-processing effects such as excitability with excitatory and inhibitory pulses, binary all-or-nothing response, and bistable dynamics.

physics.optics↗