"Fedspeak", the stylized and often nuanced language used by the U.S. Federal Reserve, encodes implicit policy signals and strategic stances. The Federal Open Market Committee strategically employs Fedspeak as a communication tool to shape market expectations and influence both domestic and global economic conditions. As such, automatically parsing and interpreting Fedspeak presents a high-impact challenge, with significant implications for financial forecasting, algorithmic trading, and data-driven policy analysis. Technically, to enrich the semantic and contextual representation of Fedspeak texts, we incorporate domain-specific reasoning grounded in the monetary policy transmission mechanism. We further introduce a dynamic uncertainty decoding module to assess the confidence of model predictions, thereby enhancing both classification accuracy and model reliability. Experimental results demonstrate that our framework achieves state-of-the-art performance on the policy stance analysis task. Moreover, statistical analysis reveals a significant positive correlation between perceptual uncertainty and model error rates, validating the effectiveness of perceptual uncertainty as a diagnostic signal.

Large Language Models (LLMs) may suffer from hallucinations in real-world applications due to the lack of relevant knowledge. In contrast, knowledge graphs encompass extensive, multi-relational structures that store a vast array of symbolic facts. Consequently, integrating LLMs with knowledge graphs has been extensively explored, with Knowledge Graph Question Answering (KGQA) serving as a critical touchstone for the integration. This task requires LLMs to answer natural language questions by retrieving relevant triples from knowledge graphs. However, existing methods face two significant challenges: *excessively long reasoning paths distracting from the answer generation*, and *false-positive relations hindering the path refinement*. In this paper, we propose an iterative interactive KGQA framework that leverages the interactive learning capabilities of LLMs to perform reasoning and Debating over Graphs (DoG). Specifically, DoG employs a subgraph-focusing mechanism, allowing LLMs to perform answer trying after each reasoning step, thereby mitigating the impact of lengthy reasoning paths. On the other hand, DoG utilizes a multi-role debate team to gradually simplify complex questions, reducing the influence of false-positive relations. This debate mechanism ensures the reliability of the reasoning process. Experimental results on five public datasets demonstrate the effectiveness and superiority of our architecture. Notably, DoG outperforms the state-of-the-art method ToG by 23.7% and 9.1% in accuracy on WebQuestions and GrailQA, respectively. Furthermore, the integration experiments with various LLMs on the mentioned datasets highlight the flexibility of DoG.

Large language models (LLMs) have shown significant promise in embodied decisionmaking tasks within virtual open-world environments.Nonetheless, their performance is hindered by the absence of domain-specific knowledge.Methods that finetune on largescale domain-specific data entail prohibitive development costs.This paper introduces Vista-Wise, a cost-effective agent framework that integrates cross-modal domain knowledge and finetunes a dedicated object detection model for visual analysis.It reduces the requirement for domain-specific training data from millions of samples to a few hundred.VistaWise integrates visual information and textual dependencies into a cross-modal knowledge graph (KG), enabling a comprehensive and accurate understanding of multimodal environments.We also equip the agent with a retrieval-based pooling strategy to extract task-related information from the KG, and a desktop-level skill library to support direct operation of the Minecraft desktop client via mouse and keyboard inputs.Experimental results demonstrate that VistaWise achieves state-of-the-art performance across various open-world tasks, highlighting its effectiveness in reducing development costs while enhancing agent performance.

Large Language Model (LLM) agents have demonstrated impressive capabilities in social deduction games (SDGs) like Werewolf, where strategic reasoning and social deception are essential. However, current approaches remain limited to textual information, ignoring crucial multimodal cues such as facial expressions and tone of voice that humans naturally use to communicate. Moreover, existing SDG agents primarily focus on inferring other players' identities without modeling how others perceive themselves or fellow players. To address these limitations, we use One Night Ultimate Werewolf (ONUW) as a testbed and present MultiMind, the first framework integrating multimodal information into SDG agents. MultiMind processes facial expressions and vocal tones alongside verbal content, while employing a Theory of Mind (ToM) model to represent each player's suspicion levels toward others. By combining this ToM model with Monte Carlo Tree Search (MCTS), our agent identifies communication strategies that minimize suspicion directed at itself. Through comprehensive evaluation in both agent-versus-agent simulations and studies with human players, we demonstrate MultiMind's superior performance in gameplay. Our work presents a significant advancement toward LLM agents capable of human-like social reasoning across multimodal domains. Our code is available at https://github.com/CjangCjengh/onuw.

Textbook question answering (TQA) is the task of correctly answering diagram or nondiagram (ND) questions given large multimodal contexts consisting of abundant essays and diagrams. In real-world scenarios, an explainable TQA system plays a key role in deepening humans' understanding of learned knowledge. However, there is no work to investigate how to provide explanations currently. To address this issue, we devise a novel architecture toward span-level eXplanations for TQA (XTQA). In this article, spans are the combinations of sentences within a paragraph. The key idea is to consider the entire textual context of a lesson as candidate evidence and then use our proposed coarse-to-fine grained explanation extracting (EE) algorithm to narrow down the evidence scope and extract the span-level explanations with varying lengths for answering different questions. The EE algorithm can also be integrated into other TQA methods to make them explainable and improve the TQA performance. Experimental results show that XTQA obtains the best overall explanation result [mean intersection over union (mIoU)] of 52.38% on the first 300 questions of CK12-QA test splits, demonstrating the explainability of our method (ND: 150 and diagram: 150). The results also show that XTQA achieves the best TQA performance of 36.46% and 36.95% on the aforementioned splits. We have released our code in https://github.com/dr-majie/opentqa.

In this demo, we illustrate the process of creating a machine economy utilizing W3bstream - an emerging open-source framework designed for connecting smart devices to smart contracts. We will emphasize the flexibility and user-friendliness of W3bstream in empowering decentralized IoT applications.

Textbook Question Answering (TQA) is the task of answering diagram and non-diagram questions given large multi-modal contexts consisting of abundant text and diagrams. Deep text understandings and effective learning of diagram semantics are important for this task due to its specificity. In this paper, we propose a Weakly Supervised learning method for TQA (WSTQ), which regards the incompletely accurate results of essential intermediate procedures for this task as supervision to develop Text Matching (TM) and Relation Detection (RD) tasks and then employs the tasks to motivate itself to learn strong text comprehension and excellent diagram semantics respectively. Specifically, we apply the result of text retrieval to build positive as well as negative text pairs. In order to learn deep text understandings, we first pre-train the text understanding module of WSTQ on TM and then fine-tune it on TQA. We build positive as well as negative relation pairs by checking whether there is any overlap between the items/regions detected from diagrams using object detection. The RD task forces our method to learn the relationships between regions, which are crucial to express the diagram semantics. We train WSTQ on RD and TQA simultaneously, i.e., multitask learning, to obtain effective diagram semantics and then improve the TQA performance. Extensive experiments are carried out on CK12-QA and AI2D to verify the effectiveness of WSTQ. Experimental results show that our method achieves significant accuracy improvements of 5.02% and 4.12% on test splits of the above datasets respectively than the current state-of-the-art baseline. We have released our code on https://github.com/dr-majie/WSTQ.

Delegated Proof-of-Stake (DPoS) is an efficient, decentralized, and flexible consensus framework available in the blockchain industry. However, applying DPoS to the decentralized Internet of Things (IoT) applications is quite challenging due to the nature of IoT systems such as large-scale deployments and huge amount of data. To address the unique challenge for IoT based blockchain applications, we present Roll-DPoS, a randomized delegated proof of stake algorithm. Roll-DPoS inherits all the advantages of the original DPoS consensus framework and further enhances its capability in terms of decentralization as well as extensibility to complex blockchain architectures. A number of modern cryptographic techniques have been utilized to optimize the consensus process with respect to the computational and communication overhead.

Outsourcing data to cloud servers, while increasing service availability and reducing users' burden of managing data, inevitably brings in new concerns such as data privacy, since the server may be honest-but-curious. To mediate the conflicts between data usability and data privacy in such a scenario, research of searchable encryption is of increasing interest. Motivated by the fact that a cloud server, besides its curiosity, may be selfish in order to save its computation and/or download bandwidth, in this paper, we investigate the searchable encryption problem in the presence of a semi-honest-but-curious server, which may execute only a fraction of search operations honestly and return a fraction of search outcome honestly. To fight against this strongest adversary ever, a verifiable SSE (VSSE) scheme is proposed to offer verifiable searchability in additional to the data privacy, both of which are further confirmed by our rigorous security analysis. Besides, we treat the practicality/efficiency as a central requirement of a searchable encryption scheme. To demonstrate the lightweightness of our scheme, we implemented and tested the proposed VSSE on a laptop (serving as the server) and a mobile phone running Android 2.3.4 (serving as the end user). The experimental results optimistically suggest that the proposed scheme satisfies all of our design goals.

The large operation range of passive RFID systems and the ubiquitous deployment of passive tags introduce growing security and privacy threats such as tag skimming/tracking/cloning, in which eavesdropping the communication between the legitimate reader and the victim tag to obtain raw data is a basic tool for the adversary. However, given the fundamentality of eavesdropping, there are limited work investigating its intension/extension for passive RFID systems. In this work, we identify a brand-new attack at physical layer, called Unidirectional Active Eavesdropping, which defeats the customary impression that eavesdropping is a “passive” attack. In this attack, the adversary transmits an un-modulated carrier at a certain frequency, while a valid reader and a tag interacts at another frequency. When a passive tag modulates the amplitude of reader's signal, it causes fluctuations on the blank carrier as well. By carefully examining the amplitude of the backscattered version of both blank carrier and reader's carrier, the eavesdropper is able to recognize tag's responses more confidently. Besides the formalization and the theoretic analysis, we set out to fill the literature's gap by demonstrating this new attack towards a popular family of passive RFID systems, namely EPCglobal UHF Class-1 Gen-2, using software-defined radio devices and a programmable passive tag. Our empirically results further confirm that the active eavesdropping achieves a significant improvement in the reliability of the intercepted communication.

Security concerns of RFID systems engaged a lot of researchers to design and to cryptanalyze RFID mutual authentication protocols. A suitable mutual authentication protocol for an RFID system should provide mutual authentication along with user privacy. In addition, such protocol must be resistant to active and passive attacks, e.g. man-in-the-middle attack, reply attack, reader-/tag-impersonation, denial of service and traceability attack. Among them, tag-impersonation refers to a process that the adversary's tag fools the legitimate reader to authenticate it as a valid tag. In this paper we exam the security of two RFID mutual authentication protocols, i.e., [6] and [17], under tag impersonation attack. We found that these two protocols share a same vulnerability in each session, the tag and the reader generates a random value respectively and they use the exclusive or (XOR) of those random values in the authentication process. We exploit this vulnerability to present two effective and efficient tag impersonation attacks against these protocols, e.g., the success probabilities of our attacks are "1" and the complexity is at most two runs of each protocol. At last, we exhibit the improved version of these protocols, which are immune from tag impersonation attacks.

In GLOBECOM'10, Adamo et. al. proposed an interesting encryption scheme, called Error Correction-Based Cipher (ECBC), working at the physical layer. This scheme, together with its ancestor, Secret Error Correcting Code (SECC), belongs to the family of Joint Encryption and Error Correction (JEEC), which combines error correction and data encryption as one process to enable efficient implementations. In this paper, we provide rigorous investigation on the security of ECBC and SECC to unveil their cryptographic strengths under chosen-plaintext attacks. For ECBC, we found a 3-stage differential-style attack, which breaks the scheme with O(k × 2deg(f)+ 2k) effort, where deg(f) is the degree of the core cryptographic function f. For SECC, we found a similar attack of complexity O(k × 2k+1). Both of the attacks are significantly improved from exhaustive search, e.g., O(22k+kn+n × 2k) for ECBC and O(2kn+ (k+n) × 2k) for SECC. In addition, we exhibit that f used in ECBC's implementation is particularly vulnerable to our attack, which allows the attacker to recover the secret generator matrix in O(1). To mitigate this vulnerability, we propose a secure yet lightweight construction of f achieving the maximum degree. Finally, the core part of our attack against ECBC has been implemented utilizing GPU acceleration and demonstrated on a cluster GPU instance provided by Amazon EC2. Experimental results confirm that the original implementation of ECBC scheme can be broken in (almost) constant time (<;0.4 second) regardless of k, whereas the ECBC scheme enhanced by our proposed f can withstand this attack to the maximum extent.

The family of WG stream ciphers has good randomness properties. In this paper, we parameterize WG-7 stream cipher for RFID tags, where the modest computation/storage capabilities and the necessity to keep their prices low present a challenging problem that goes beyond the well-studied cryptography. The rigorous security analysis of WG-7 indicates that it is secure against time/memory/data trade off attack, differential attack, algebraic attack, correlation attack and Discrete Fourier Transform (DFT) attack. Furthermore, we offer efficient implementation of WG-7 on the 4-bit microcontroller ATAM893-D and the 8-bit microcontroller ATmega8 from ATmel. The experimental results show that WG-7 outperforms most of previous proposals in terms of throughput and implementation complexity. Moreover, we propose a mutual authentication protocol based on WG-7, which provides the untraceability, resistance of tag impersonation and reader impersonation. With its verified cryptographic properties, low implementation complexity and ideal throughput, WG-7 is a promising candidate for RFID applications.

The paper introduces a novel message authentication framework over broadcast channels, where a symmetric cryptography-based physical layer assisted message authentication (PLAA) scheme is introduced in wireless networks. The proposed framework integrate the conventional message authentication schemes and the physical layer authentication mechanisms by taking advantage of temporal and spatial uniqueness in physical layer channel responses, aiming to achieving fast authentication while minimising the packet transmission overhead. Our claims through extensive analysis and simulation will be verified via comparing with public key infrastructure-based PLAA scheme and traditional upper layer authentication schemes.