https://hdl.handle.net/20.500.12761/2 2026-06-17T09:51:44Z https://hdl.handle.net/20.500.12761/2041 EgoLife: Towards Egocentric Life Assistant Yang, Jingkang; Liu, Shuai; Guo, Hongming; Dong, Yuhao; Zhang, Xiamengwei; Zhang, Sicheng; Wang, Pengyun; Zhou, Zitang; Xie, Binzhu; Wang, Ziyue; Ouyang, Bei; Lin, Zhengyu; Cominelli, Marco; Cai, Zhongang; Li, Bo; Zhang, Yuanhan; Zhang, Peiyuan; Hong, Fangzhou; Widmer, Joerg; Gringoli, Francesco; Yang, Lei; Liu, Ziwei We introduce EgoLife, a project to develop an egocentric life assistant that accompanies and enhances personal efficiency through AI-powered wearable glasses. To lay the foundation for this assistant, we conducted a comprehensive data collection study where six participants lived together for one week, continuously recording their daily activities - including discussions, shopping, cooking, socializing, and entertainment - using AI glasses for multimodal egocentric video capture, along with synchronized third-person-view video references. This effort resulted in the EgoLife Dataset, a comprehensive 300-hour egocentric, interpersonal, multiview, and multimodal daily life dataset with intensive annotation. Leveraging this dataset, we introduce EgoLifeQA, a suite of long-context, life-oriented question-answering tasks designed to provide meaningful assistance in daily life by addressing practical questions such as recalling past relevant events, monitoring health habits, and offering personalized recommendations. To address the key technical challenges of (1) developing robust visual-audio models for egocentric data, (2) enabling identity recognition, and (3) facilitating long-context question answering over extensive temporal information, we introduce EgoButler, an integrated system comprising EgoGPT and EgoRAG. EgoGPT is an omni-modal model trained on egocentric datasets, achieving state-of-the-art performance on egocentric video understanding. EgoRAG is a retrieval-based component that supports answering ultra-long-context questions. Our experimental studies verify their working mechanisms and reveal critical factors and bottlenecks, guiding future improvements. By releasing our datasets, models, and benchmarks, we aim to stimulate further research in egocentric AI assistants. 2025-06-15T00:00:00Z https://hdl.handle.net/20.500.12761/2040 DFT-Based Spectral Precoding for Scalable Unique-Word OFDM Sharma, Salil; Waqas Haider Shah, Syed; Lacruz, Jesús Omar; Widmer, Joerg Unique-word orthogonal frequency division multiplexing (UW-OFDM) addresses the spectral inefficiency of cyclic prefix (CP) OFDM by embedding the guard interval within the symbol period. However, its practical adoption is hindered by a prohibitive computational bottleneck. Conventional methods for embedding unique word within the DFT window rely on complex matrix inversions at the transmitter (Tx). While alternative zero-padded (ZP) schemes avoid this complexity, they place guard interval outside DFT window, breaking natural circularity and necessitating noise-enhancing Overlap-Add receiver (Rx). This letter proposes a scalable, closed-form isometric Spectral Precoding architecture that resolves this trade-off. By exploiting frequency-domain interpolation, we implicitly construct the required near-zero-energy tail within the DFT window using only deterministic linear transformations. We prove this construction reduces Tx complexity to log-linear order while preserving standard single-tap equalization without Rx noise enhancement. Validated with 5G-NR numerologies, the scheme matches 2026-05-01T00:00:00Z https://hdl.handle.net/20.500.12761/2039 Blind 5G NR LEO Initial Access Receiver Jaminon-De Roeck, Chen; Timothy, Otim; Santaromita, Giuseppe; Giustiniano, Domenico LEO satellite links pose severe challenges for 5G Non-Terrestrial Network (NTN) initial access due to large carrier frequency offsets, rapidly varying Doppler, and low SNR, making fully blind operation essential in cold-start scenarios. Using a real 5G NTN LEO over-the-air testbed emulation, we show that Doppler compensation accuracy directly governs user equipment (UE) connection outcomes, with residual errors below 3 kHz required for reliable synchronization and attachment. We then present a fully blind, feedfor ward receiver that computes bounded reliability scores at each processing stage and uses them for hypothesis gating, burst selection, conservative log-likelihood ratio (LLR)-domain combining, weighting, and final lock de cisions. The design includes a conservative, reliability aware multi-Synchronization Signal Block (SSB) Phys ical Broadcast Channel (PBCH) soft-combining stage that only supplements single-SSB channel quality as sessment. Deterministic results from 3GPP-compliant simulations demonstrate robust blind synchronization under ±100kHz frequency offsets and successful PBCH decoding at low signal-to-noise ratios in urban line-of sight (LOS) and non-line-of-sight (NLOS) conditions. The simulation code used is publicly available. 2026-04-20T00:00:00Z https://hdl.handle.net/20.500.12761/2038 Breaking Sandwich MEV in ERC-20 DeFi via Context-Aware Directional Constraints Arote, Prerna Prabhakar Decentralized exchanges (DEXs) are a core component of decentralized finance (DeFi) but remain vulnerable to Maximal Extractable Value (MEV) attacks, particularly sandwich attacks that exploit transaction ordering across accounts and blocks. Existing defenses often rely on protocol modifications or identity-based assumptions, limiting their practicality and robustness. We present Context-Aware Directional Constraint (CADC), a lightweight application-layer defense implemented as an ERC-20–compatible token. CADC enforces directional consistency within pool-specific execution contexts and maintains a per-direction entry price to prevent both immediate and delayed profit extraction. By constraining the temporal and price conditions required for attack execution, CADC remains effective against multi-address and cross-block strategies. Evaluation on a testnet shows that CADC prevents same-block and delayed sandwich attacks while preserving normal trading behavior, with minimal overhead (2--3k gas per pool interaction, 3--8% of typical swaps). CADC provides a practical and deployable MEV mitigation without requiring changes to the underlying blockchain protocol. 2026-06-01T00:00:00Z