Draiven
AI Insights
- Decentralized coordination raises challenges in maintaining coherent semantic adjudication across heterogeneous agents, especially when contextual information is unevenly distributed. (ML: 0.96)
- Semantic Auditability: The ability to preserve the semantic context and rationale underlying coordination decisions, enabling post-hoc reconstruction of why a particular coordination outcome was reached. (ML: 0.94)
- Intention-Based Access Control (IBAC): An authorization model that evaluates governance decisions against the expressed intent, its contextual constraints, and organizational policies. (ML: 0.92)
- The protocol may conflict with regulatory or organizational environments that mandate immutable contracts or pre-certified execution paths. (ML: 0.91)
- Liquid Interfaces prioritize semantic flexibility, adaptive negotiation, and contextual interpretation over deterministic execution guarantees. (ML: 0.88)
- Liquid Interfaces position governance in distributed systems not as an external control layer, but as an emergent consequence of semantically mediated coordination. (ML: 0.87)
- Liquid Interfaces rely on agents declaring capabilities and constraints as part of the negotiation process, introducing a tension between openness and trust. (ML: 0.85)
- Liquid Interfaces are designed for environments characterized by semantic uncertainty, evolving capabilities, and open-ended interaction. (ML: 0.83)
- Liquid Interface: A protocol that enables semantically mediated coordination among autonomous agents, prioritizing flexibility, adaptability, and contextual interpretation over deterministic execution guarantees. (ML: 0.78)
- The paradigm is not suitable for domains that require strict real-time constraints, bounded worst-case latency, or hard safety guarantees enforced through static verification. (ML: 0.74)
Abstract
Contemporary software architectures struggle to support autonomous agents whose reasoning is adaptive, probabilistic, and context-dependent, while system integration remains dominated by static interfaces and deterministic contracts. This paper introduces Liquid Interfaces, a coordination paradigm in which interfaces are not persistent technical artifacts, but ephemeral relational events that emerge through intention articulation and semantic negotiation at runtime.We formalize this model and present the Liquid Interface Protocol (LIP),which governs intention-driven interaction, negotiated execution, and enforce ephemerality under semantic uncertainty. We further discuss the governance implications of this approach and describe a reference architecture that demonstrates practical feasibility. Liquid Interfaces provide a principled foundation for adaptive coordination in agent-based systems
Contemporary software architectures struggle to support autonomous agents whose reasoning is adaptive, probabilistic, and context-dependent, while system integration remains dominated by static interfaces and deterministic contracts. This paper introduces Liquid Interfaces, a coordination paradigm in which interfaces are not persistent technical artifacts, but ephemeral relational events that emerge through intention articulation and semantic negotiation at runtime.We formalize this model and present the Liquid Interface Protocol (LIP),which governs intention-driven interaction, negotiated execution, and enforce ephemerality under semantic uncertainty. We further discuss the governance implications of this approach and describe a reference architecture that demonstrates practical feasibility. Liquid Interfaces provide a principled foundation for adaptive coordination in agent-based systems
Why we are recommending this paper?
Due to your Interest in Ontology for Products
This paper introduces Liquid Interfaces, a coordination paradigm directly relevant to building robust product ontologies and knowledge graphs. Given your interest in product categorization and knowledge management, this work offers a valuable framework for structuring complex systems.
Due to your Interest in Ontology for Products
This paper introduces Liquid Interfaces, a coordination paradigm directly relevant to building robust product ontologies and knowledge graphs. Given your interest in product categorization and knowledge management, this work offers a valuable framework for structuring complex systems.
Harbin Institute of Technology
AI Insights
- The adaptive sampling strategy may lead to biased models if not properly calibrated. (ML: 0.97)
- Adaptive Sampling: A strategy that selects relevant clients and samples for each FL round, based on the model's performance and data distribution. (ML: 0.96)
- The proposed approach may not be suitable for large-scale datasets due to its computational complexity. (ML: 0.94)
- Previous works have addressed FCL challenges using various methods, including model pruning and knowledge distillation. (ML: 0.94)
- FCL-AS uses an adaptive sampling strategy to select relevant clients and samples for each FL round, based on the model's performance and data distribution. (ML: 0.92)
- FCL-AS is more efficient than traditional federated learning approaches due to its adaptive sampling strategy. (ML: 0.89)
- The paper discusses the challenges of federated continual learning (FCL), which involves updating models on decentralized data in a sequential manner. (ML: 0.88)
- Federated Continual Learning (FCL): A learning paradigm where models are updated sequentially on decentralized data. (ML: 0.87)
- The authors propose a novel approach called Federated Continual Learning with Adaptive Sampling (FCL-AS) to address these challenges. (ML: 0.87)
- The proposed FCL-AS approach outperforms existing methods in terms of accuracy and convergence speed. (ML: 0.74)
Abstract
Federated Continual Learning (FCL) leverages inter-client collaboration to balance new knowledge acquisition and prior knowledge retention in non-stationary data. However, existing batch-based FCL methods lack adaptability to streaming scenarios featuring category overlap between old and new data and absent task identifiers, leading to indistinguishability of old and new knowledge, uncertain task assignments for samples, and knowledge confusion.To address this, we propose streaming federated continual learning setting: per federated learning (FL) round, clients process streaming data with disjoint samples and potentially overlapping categories without task identifiers, necessitating sustained inference capability for all prior categories after each FL round.Next, we introduce FedKACE: 1) an adaptive inference model switching mechanism that enables unidirectional switching from local model to global model to achieve a trade-off between personalization and generalization; 2) a adaptive gradient-balanced replay scheme that reconciles new knowledge learning and old knowledge retention under overlapping-class scenarios; 3) a kernel spectral boundary buffer maintenance that preserves high-information and high-boundary-influence samples to optimize cross-round knowledge retention. Experiments across multiple scenarios and regret analysis demonstrate the effectiveness of FedKACE.
Federated Continual Learning (FCL) leverages inter-client collaboration to balance new knowledge acquisition and prior knowledge retention in non-stationary data. However, existing batch-based FCL methods lack adaptability to streaming scenarios featuring category overlap between old and new data and absent task identifiers, leading to indistinguishability of old and new knowledge, uncertain task assignments for samples, and knowledge confusion.To address this, we propose streaming federated continual learning setting: per federated learning (FL) round, clients process streaming data with disjoint samples and potentially overlapping categories without task identifiers, necessitating sustained inference capability for all prior categories after each FL round.Next, we introduce FedKACE: 1) an adaptive inference model switching mechanism that enables unidirectional switching from local model to global model to achieve a trade-off between personalization and generalization; 2) a adaptive gradient-balanced replay scheme that reconciles new knowledge learning and old knowledge retention under overlapping-class scenarios; 3) a kernel spectral boundary buffer maintenance that preserves high-information and high-boundary-influence samples to optimize cross-round knowledge retention. Experiments across multiple scenarios and regret analysis demonstrate the effectiveness of FedKACE.
Why we are recommending this paper?
Due to your Interest in Continual Generalized Category Discovery
Focusing on continual learning and knowledge graphs, this paper addresses the challenge of adapting to evolving product categories β a core interest for you. The streaming federated approach aligns with your interest in dynamic knowledge discovery.
Due to your Interest in Continual Generalized Category Discovery
Focusing on continual learning and knowledge graphs, this paper addresses the challenge of adapting to evolving product categories β a core interest for you. The streaming federated approach aligns with your interest in dynamic knowledge discovery.
The University of Hong Kong
AI Insights
- product polarization: a situation in which some firms produce high-quality products and others produce low-quality products. (ML: 0.97)
- It highlights inefficiencies specific to oligopoly markets, such as the potential for firms with low standalone values to dominate common-characteristics provision. (ML: 0.94)
- Their results highlight inefficiencies specific to oligopoly markets, such as the potential for firms with low standalone values to dominate common-characteristics provision. (ML: 0.93)
- product concentration: a situation in which all firms produce similar or identical products. (ML: 0.93)
- The authors derive conditions under which each type of equilibrium exists and compare their welfare implications. (ML: 0.92)
- The paper explores the relationship between product differentiation, concentration, and polarization in oligopoly markets. (ML: 0.92)
- The paper provides new insights into the relationship between product differentiation, concentration, and polarization in oligopoly markets. (ML: 0.90)
- product differentiation: a situation in which firms produce differentiated products that are close substitutes for one another. (ML: 0.90)
- standalone value: the value of a firm's output if it were to produce alone without any competition from other firms. (ML: 0.89)
- It develops a framework for analyzing equilibria with different levels of product differentiation, including product concentration and polarization. (ML: 0.73)
Abstract
Building on the generalized hedonic-linear model of Pellegrino (2025), this paper studies optimal product differentiation when a representative consumer has preferences over product characteristics. Under multiproduct monopoly, the monopolist's choice of product characteristics is always aligned with the social planner's optimum, despite underproduction. By contrast, under oligopoly, multiple equilibria can arise that differ qualitatively in their patterns of characteristics design. We show that, while oligopoly equilibria exhibiting product differentiation yield higher welfare than those with product concentration, the degree of product differentiation under oligopoly remains below the socially optimal level. As a result, social welfare under oligopoly is typically lower than under monopoly, highlighting a key advantage of coordination in characteristics design. We extend the analysis to settings with overlapping ownership structures and show that common ownership can improve welfare by inducing firms to soften competition through increased product differentiation rather than output reduction.
Building on the generalized hedonic-linear model of Pellegrino (2025), this paper studies optimal product differentiation when a representative consumer has preferences over product characteristics. Under multiproduct monopoly, the monopolist's choice of product characteristics is always aligned with the social planner's optimum, despite underproduction. By contrast, under oligopoly, multiple equilibria can arise that differ qualitatively in their patterns of characteristics design. We show that, while oligopoly equilibria exhibiting product differentiation yield higher welfare than those with product concentration, the degree of product differentiation under oligopoly remains below the socially optimal level. As a result, social welfare under oligopoly is typically lower than under monopoly, highlighting a key advantage of coordination in characteristics design. We extend the analysis to settings with overlapping ownership structures and show that common ownership can improve welfare by inducing firms to soften competition through increased product differentiation rather than output reduction.
Why we are recommending this paper?
Due to your Interest in Product Categorization
This paper directly tackles product differentiation through a hedonic lens, aligning with your interest in taxonomy design and product categorization. The focus on consumer preferences and product characteristics is highly relevant.
Due to your Interest in Product Categorization
This paper directly tackles product differentiation through a hedonic lens, aligning with your interest in taxonomy design and product categorization. The focus on consumer preferences and product characteristics is highly relevant.
Universit e Paris Saclay
AI Insights
- Contextual objectivity is based on the idea that physical systems are described by modalities (possible states) rather than definite properties. (ML: 0.96)
- The paper concludes by highlighting the potential applications of contextual objectivity in fields such as quantum computing and quantum information processing. (ML: 0.90)
- Born rule: A fundamental principle in quantum mechanics that describes how probabilities are assigned to measurement outcomes. (ML: 0.89)
- Modality: A possible state of a physical system, which is used to describe its properties and behavior. (ML: 0.86)
- Contextual objectivity: A new ontology for understanding quantum phenomena, based on the idea that physical systems are described by modalities (possible states) rather than definite properties. (ML: 0.84)
- The paper discusses the concept of contextual objectivity in quantum mechanics, which is a new ontology for understanding quantum phenomena. (ML: 0.82)
- Contextual objectivity offers a new and coherent way to understand quantum phenomena, one that complementsβand in some respects surpassesβexisting ontological interpretations. (ML: 0.78)
- They also discuss the use of infinite-dimensional Hilbert spaces to describe quantum systems, which is a key aspect of their approach. (ML: 0.77)
- The authors argue that this approach can resolve several long-standing problems in quantum mechanics, including the measurement problem and the Born rule. (ML: 0.70)
- Infinite-dimensional Hilbert space: A mathematical framework used to describe quantum systems, where the number of dimensions is infinite. (ML: 0.65)
Abstract
This note presents a concise and non-polemical comparison of several major interpretations of quantum mechanics, with a particular emphasis on the distinction between FAPP-solutions ("For All Practical Purposes'') versus ontological solutions to the measurement problem. Building on this distinction, we argue that the Contexts-Systems-Modalities (CSM) framework, supplemented by the operator-algebraic description of macroscopic contexts, provides a conceptually complete, non-FAPP ontology that naturally incorporates irreversibility and the physical structure of measurement devices. This approach differs significantly from other ontological interpretations such as Bohmian mechanics, spontaneous collapse, or many-worlds, and highlights the major role of contextual quantization in shaping quantum theory.
This note presents a concise and non-polemical comparison of several major interpretations of quantum mechanics, with a particular emphasis on the distinction between FAPP-solutions ("For All Practical Purposes'') versus ontological solutions to the measurement problem. Building on this distinction, we argue that the Contexts-Systems-Modalities (CSM) framework, supplemented by the operator-algebraic description of macroscopic contexts, provides a conceptually complete, non-FAPP ontology that naturally incorporates irreversibility and the physical structure of measurement devices. This approach differs significantly from other ontological interpretations such as Bohmian mechanics, spontaneous collapse, or many-worlds, and highlights the major role of contextual quantization in shaping quantum theory.
Why we are recommending this paper?
Due to your Interest in Ontology for Products
Exploring ontological solutions to the measurement problem, this paperβs focus on contextual understanding resonates with your interest in knowledge management and ontology development. The comparison of different interpretations offers a valuable perspective.
Due to your Interest in Ontology for Products
Exploring ontological solutions to the measurement problem, this paperβs focus on contextual understanding resonates with your interest in knowledge management and ontology development. The comparison of different interpretations offers a valuable perspective.
The Hong Kong Polytechnic University
AI Insights
- The proposed LLM-empowered editing approach enhances logical reasoning, leading to improved accuracy and interpretability. (ML: 0.97)
- Graph Neural Networks (GNNs): A type of neural network designed for processing graph-structured data, which can learn complex relationships between nodes in a graph. (ML: 0.94)
- Temporal Knowledge Graph Reasoning: The process of reasoning over knowledge graphs to answer temporal queries, such as 'What events occurred during the presidency of Donald Trump?' Hits@1, Hits@3, Hits@10: Evaluation metrics used to measure the performance of a model in answering temporal queries. (ML: 0.94)
- Large Language Models (LLMs): Pre-trained models that have been trained on vast amounts of text data and can generate human-like responses to input prompts. (ML: 0.94)
- The integration of GNN structural knowledge and LLM semantic context enables IGETR to capture more nuanced and contextually relevant relations. (ML: 0.94)
- IGETR's integration of GNNs and LLMs enables it to outperform baseline approaches across multiple evaluation metrics. (ML: 0.93)
- Ablation studies confirm the effectiveness of the LLM-based path editing module and demonstrate that performance improvements originate from genuine reasoning enhancements rather than merely retrieving internal historical data. (ML: 0.93)
- IGETR, a novel framework that combines graph neural networks (GNNs) with large language models (LLMs), demonstrates superior performance across multiple metrics compared to baseline approaches. (ML: 0.92)
- Hits@k represents the proportion of correct answers among the top k retrieved results. (ML: 0.91)
- The proposed framework effectively leverages the strengths of both GNNs and LLMs, leading to improved accuracy and interpretability in temporal knowledge graph reasoning. (ML: 0.88)
Abstract
Temporal knowledge graph reasoning (TKGR) aims to predict future events by inferring missing entities with dynamic knowledge structures. Existing LLM-based reasoning methods prioritize contextual over structural relations, struggling to extract relevant subgraphs from dynamic graphs. This limits structural information understanding, leading to unstructured, hallucination-prone inferences especially with temporal inconsistencies. To address this problem, we propose IGETR (Integration of Graph and Editing-enhanced Temporal Reasoning), a hybrid reasoning framework that combines the structured temporal modeling capabilities of Graph Neural Networks (GNNs) with the contextual understanding of LLMs. IGETR operates through a three-stage pipeline. The first stage aims to ground the reasoning process in the actual data by identifying structurally and temporally coherent candidate paths through a temporal GNN, ensuring that inference starts from reliable graph-based evidence. The second stage introduces LLM-guided path editing to address logical and semantic inconsistencies, leveraging external knowledge to refine and enhance the initial paths. The final stage focuses on integrating the refined reasoning paths to produce predictions that are both accurate and interpretable. Experiments on standard TKG benchmarks show that IGETR achieves state-of-the-art performance, outperforming strong baselines with relative improvements of up to 5.6% on Hits@1 and 8.1% on Hits@3 on the challenging ICEWS datasets. Additionally, we execute ablation studies and additional analyses confirm the effectiveness of each component.
Temporal knowledge graph reasoning (TKGR) aims to predict future events by inferring missing entities with dynamic knowledge structures. Existing LLM-based reasoning methods prioritize contextual over structural relations, struggling to extract relevant subgraphs from dynamic graphs. This limits structural information understanding, leading to unstructured, hallucination-prone inferences especially with temporal inconsistencies. To address this problem, we propose IGETR (Integration of Graph and Editing-enhanced Temporal Reasoning), a hybrid reasoning framework that combines the structured temporal modeling capabilities of Graph Neural Networks (GNNs) with the contextual understanding of LLMs. IGETR operates through a three-stage pipeline. The first stage aims to ground the reasoning process in the actual data by identifying structurally and temporally coherent candidate paths through a temporal GNN, ensuring that inference starts from reliable graph-based evidence. The second stage introduces LLM-guided path editing to address logical and semantic inconsistencies, leveraging external knowledge to refine and enhance the initial paths. The final stage focuses on integrating the refined reasoning paths to produce predictions that are both accurate and interpretable. Experiments on standard TKG benchmarks show that IGETR achieves state-of-the-art performance, outperforming strong baselines with relative improvements of up to 5.6% on Hits@1 and 8.1% on Hits@3 on the challenging ICEWS datasets. Additionally, we execute ablation studies and additional analyses confirm the effectiveness of each component.
Why we are recommending this paper?
Due to your Interest in Knowledge Graphs
This paper addresses the need for reasoning with temporal knowledge graphs, a critical component of your interest in knowledge graphs and TKGR. The focus on extracting relevant subgraphs aligns with your interest in structured knowledge representation.
Due to your Interest in Knowledge Graphs
This paper addresses the need for reasoning with temporal knowledge graphs, a critical component of your interest in knowledge graphs and TKGR. The focus on extracting relevant subgraphs aligns with your interest in structured knowledge representation.
Indian Institute of Technology Guwahati
AI Insights
- Regular language: A language that can be recognized by a deterministic finite automaton (DFA). (ML: 0.93)
- The paper does not provide a complete characterization of the language of all 1-11-representations of a given graph G. (ML: 0.92)
- The study of repetition patterns in 1-11-representations of graphs has led to new insights into the structure and properties of these representations. (ML: 0.92)
- The language of all permutational 1-11-representations of a given graph G is regular. (ML: 0.91)
- The language of all 1-11-representations of a given graph G is regular. (ML: 0.90)
- The regularity of the language of all 1-11-representations of a given graph G is a key result that highlights the connection between graph theory and formal language theory. (ML: 0.90)
- The paper assumes that the input alphabet V is finite, which may not be the case for all graphs. (ML: 0.90)
- The study of 1-11-representations of graphs has been ongoing since the work of Thue in 1906. (ML: 0.85)
- Deterministic finite automaton (DFA): A 5-tuple A = (Q, V, Ξ΄, q0, F), where Q is a finite set of states, V is a finite input alphabet, Ξ΄: Q Γ V β Q is the transition function, q0 β Q is the initial state, and F β Q is the set of accepting states. (ML: 0.84)
- Cube-free representations can always be obtained in the permutation framework, while square-free representations may be unavoidable. (ML: 0.80)
Abstract
A 1-11-representation of a graph $G(V,E)$ is a word over the alphabet $V$ such that two distinct vertices $x$ and $y$ are adjacent if and only if the restricted word $w{x,y}$ (obtained from $w$ by deleting all letters except $x$ and $y$) contains at most one occurrence of $xx$ or $yy$. Although every graph admits a 1-11-representation, the repetition patterns that may or must appear in such representations have not been fully studied. In this paper, we study cube-free and square-free 1-11-representations of graphs. We first show that cubes cannot always be avoided in 1-11-representations of minimum length by providing a graph for which every minimum-length 1-11-representation necessarily contains a cube. We then focus on permutational 1-11-representations, where the representing word is a concatenation of permutations of the vertex set. In this setting, we prove that any cube appearing in a permutational 1-11-representation can be removed without changing the represented graph. As a consequence, every permutational 1-11-representation attaining the permutational 1-11-representation number is cube-free. We further show that this behaviour does not extend to squares by providing a graph for which every permutational 1-11-representation with the minimum number of permutations necessarily contains a square. Finally, we prove that the language of all 1-11-representations of a given graph is regular. Moreover, we show that the language of all permutational 1-11-representations of a graph is also regular.
A 1-11-representation of a graph $G(V,E)$ is a word over the alphabet $V$ such that two distinct vertices $x$ and $y$ are adjacent if and only if the restricted word $w{x,y}$ (obtained from $w$ by deleting all letters except $x$ and $y$) contains at most one occurrence of $xx$ or $yy$. Although every graph admits a 1-11-representation, the repetition patterns that may or must appear in such representations have not been fully studied. In this paper, we study cube-free and square-free 1-11-representations of graphs. We first show that cubes cannot always be avoided in 1-11-representations of minimum length by providing a graph for which every minimum-length 1-11-representation necessarily contains a cube. We then focus on permutational 1-11-representations, where the representing word is a concatenation of permutations of the vertex set. In this setting, we prove that any cube appearing in a permutational 1-11-representation can be removed without changing the represented graph. As a consequence, every permutational 1-11-representation attaining the permutational 1-11-representation number is cube-free. We further show that this behaviour does not extend to squares by providing a graph for which every permutational 1-11-representation with the minimum number of permutations necessarily contains a square. Finally, we prove that the language of all 1-11-representations of a given graph is regular. Moreover, we show that the language of all permutational 1-11-representations of a graph is also regular.
Why we are recommending this paper?
Due to your Interest in Graphs for Products
Due to your Interest in Graphs for Products
University of
AI Insights
- The problem statement is a mathematical proof of the cohomology and PoincarΓ© series of graph products of Koszul algebras. (ML: 0.78)
- The cohomology and PoincarΓ© series of graph products of Koszul algebras are computed using the Normal Form Theorem and other techniques. (ML: 0.77)
- The cohomology of ΞpGq is computed as HβpΞGqΒ»AβpΞGqβE{I!pΞGq, where I!pΞGq is the two-sided ideal generated by the family S!pΞGq. (ML: 0.72)
- Koszul algebra graph product cohomology PoincarΓ© series The solution involves a detailed mathematical proof that uses various results from algebraic topology and combinatorics. (ML: 0.71)
- The solution involves using the Normal Form Theorem from Green [11, Theorem3.9] to show that the complex pP nFpbFpΞEpGq, d nqnPNis acyclic. (ML: 0.63)
- The gocha and PoincarΓ© series of ΞG are computed using the cohomology of ΞpGq. (ML: 0.61)
Abstract
Let $p$ be a prime. We resolve a question posed by MinΓ‘Δ-Rogelstad-TΓ’n. We relate the Zassenhaus and the lower central series of pro-$p$ groups under a torsion-freeness condition. We also study graph products of (pro-$p$) groups under natural assumptions. In particular, we compute their graded Lie algebras associated with the previous filtrations, as well as their cohomology over $\mathbb{F}_p$. Our approach relies on various filtrations of amalgamated products, as studied in Leoni's PhD thesis. Explicit examples are provided using the Koszul property. As a concrete application, we compute the cohomology over $\mathbb{F}_p$ and the graded Lie algebras associated with the filtrations of graph products of fundamental groups of surfaces. These groups furnish new examples satisfying the torsion-freeness condition, which arises in the question of MinΓ‘Δ-Rogelstad-TΓ’n.
Let $p$ be a prime. We resolve a question posed by MinΓ‘Δ-Rogelstad-TΓ’n. We relate the Zassenhaus and the lower central series of pro-$p$ groups under a torsion-freeness condition. We also study graph products of (pro-$p$) groups under natural assumptions. In particular, we compute their graded Lie algebras associated with the previous filtrations, as well as their cohomology over $\mathbb{F}_p$. Our approach relies on various filtrations of amalgamated products, as studied in Leoni's PhD thesis. Explicit examples are provided using the Koszul property. As a concrete application, we compute the cohomology over $\mathbb{F}_p$ and the graded Lie algebras associated with the filtrations of graph products of fundamental groups of surfaces. These groups furnish new examples satisfying the torsion-freeness condition, which arises in the question of MinΓ‘Δ-Rogelstad-TΓ’n.
Why we are recommending this paper?
Due to your Interest in Graphs for Products
Due to your Interest in Graphs for Products
Southeast University
AI Insights
- Previous work has shown that LLMs can suffer from world knowledge forgetting when adapting to new tasks. (ML: 0.97)
- Some methods have been proposed to alleviate this problem, but they often require significant modifications to the original model or additional training data. (ML: 0.96)
- The performance of the model may degrade over time if it is not properly fine-tuned. (ML: 0.95)
- The proposed method may not be suitable for all types of tasks or domains. (ML: 0.95)
- LORAMOE is designed to adapt to new tasks without forgetting previously learned knowledge, and can be used for various applications such as question answering, text classification, and language translation. (ML: 0.92)
- The paper discusses the challenges of training large language models (LLMs) on a single task, and proposes a method called LORAMOE that alleviates world knowledge forgetting in LLMs via Moe-style plugin. (ML: 0.91)
- LORAMOE's ability to adapt to new tasks without forgetting previously learned knowledge makes it a promising approach for continual learning. (ML: 0.91)
- LLMs: Large Language Models Moe-style plugin: A type of model that uses a mixture-of-experts architecture to adapt to new tasks The proposed method, LORAMOE, is effective in alleviating world knowledge forgetting in LLMs and can be used for various applications. (ML: 0.90)
Abstract
Continual learning for pre-trained vision-language models requires balancing three competing objectives: retaining pre-trained knowledge, preserving knowledge from a sequence of learned tasks, and maintaining the plasticity to acquire new knowledge. This paper presents a simple but effective approach called KeepLoRA to effectively balance these objectives. We first analyze the knowledge retention mechanism within the model parameter space and find that general knowledge is mainly encoded in the principal subspace, while task-specific knowledge is encoded in the residual subspace. Motivated by this finding, KeepLoRA learns new tasks by restricting LoRA parameter updates in the residual subspace to prevent interfering with previously learned capabilities. Specifically, we infuse knowledge for a new task by projecting its gradient onto a subspace orthogonal to both the principal subspace of pre-trained model and the dominant directions of previous task features. Our theoretical and empirical analyses confirm that KeepLoRA balances the three objectives and achieves state-of-the-art performance. The implementation code is available at https://github.com/MaolinLuo/KeepLoRA.
Continual learning for pre-trained vision-language models requires balancing three competing objectives: retaining pre-trained knowledge, preserving knowledge from a sequence of learned tasks, and maintaining the plasticity to acquire new knowledge. This paper presents a simple but effective approach called KeepLoRA to effectively balance these objectives. We first analyze the knowledge retention mechanism within the model parameter space and find that general knowledge is mainly encoded in the principal subspace, while task-specific knowledge is encoded in the residual subspace. Motivated by this finding, KeepLoRA learns new tasks by restricting LoRA parameter updates in the residual subspace to prevent interfering with previously learned capabilities. Specifically, we infuse knowledge for a new task by projecting its gradient onto a subspace orthogonal to both the principal subspace of pre-trained model and the dominant directions of previous task features. Our theoretical and empirical analyses confirm that KeepLoRA balances the three objectives and achieves state-of-the-art performance. The implementation code is available at https://github.com/MaolinLuo/KeepLoRA.
Why we are recommending this paper?
Due to your Interest in Continual Generalized Category Discovery
Due to your Interest in Continual Generalized Category Discovery
Amazon
AI Insights
- Evidence-based fact verification: A method of verifying facts by analyzing evidence from multiple sources. (ML: 0.98)
- Fact-checking is the process of verifying the accuracy of information. (ML: 0.97)
- The paper cites various studies on fact-checking and explainable AI, including research on contrastive explanations, counterfactual explanations, and evidence-based fact verification. (ML: 0.97)
- Contrastive explanation: An explanation that highlights the differences between a model's predictions and actual outcomes. (ML: 0.96)
- Explainable AI refers to methods that make it easier for humans to understand how an AI system arrived at a particular decision or prediction. (ML: 0.96)
- The paper discusses various methods for fact-checking and explainable AI, including contrastive explanations, counterfactual explanations, and evidence-based fact verification. (ML: 0.96)
- The paper discusses various methods for fact-checking and explainable AI, including contrastive explanations, counterfactual explanations, and evidence-based fact verification. (ML: 0.96)
- The paper discusses various methods for fact-checking and explainable AI, including contrastive explanations, counterfactual explanations, and evidence-based fact verification. (ML: 0.96)
- The paper concludes that explainable AI is essential for building trust in AI systems, and that various methods can be used to improve the transparency and accountability of AI decision-making. (ML: 0.95)
- Counterfactual explanation: An explanation that describes what would have happened if certain conditions had been different. (ML: 0.95)
- The paper does not provide a comprehensive overview of all existing methods for fact-checking and explainable AI. (ML: 0.94)
- Some of the methods discussed in the paper may be too complex or require significant computational resources to implement. (ML: 0.90)
Abstract
Claim verification is a core component of automated fact-checking systems, aimed at determining the truthfulness of a statement by assessing it against reliable evidence sources such as documents or knowledge bases. This work presents KG-CRAFT, a method that improves automatic claim verification by leveraging large language models (LLMs) augmented with contrastive questions grounded in a knowledge graph. KG-CRAFT first constructs a knowledge graph from claims and associated reports, then formulates contextually relevant contrastive questions based on the knowledge graph structure. These questions guide the distillation of evidence-based reports, which are synthesised into a concise summary that is used for veracity assessment by LLMs. Extensive evaluations on two real-world datasets (LIAR-RAW and RAWFC) demonstrate that our method achieves a new state-of-the-art in predictive performance. Comprehensive analyses validate in detail the effectiveness of our knowledge graph-based contrastive reasoning approach in improving LLMs' fact-checking capabilities.
Claim verification is a core component of automated fact-checking systems, aimed at determining the truthfulness of a statement by assessing it against reliable evidence sources such as documents or knowledge bases. This work presents KG-CRAFT, a method that improves automatic claim verification by leveraging large language models (LLMs) augmented with contrastive questions grounded in a knowledge graph. KG-CRAFT first constructs a knowledge graph from claims and associated reports, then formulates contextually relevant contrastive questions based on the knowledge graph structure. These questions guide the distillation of evidence-based reports, which are synthesised into a concise summary that is used for veracity assessment by LLMs. Extensive evaluations on two real-world datasets (LIAR-RAW and RAWFC) demonstrate that our method achieves a new state-of-the-art in predictive performance. Comprehensive analyses validate in detail the effectiveness of our knowledge graph-based contrastive reasoning approach in improving LLMs' fact-checking capabilities.
Why we are recommending this paper?
Due to your Interest in Knowledge Graphs
Due to your Interest in Knowledge Graphs
Karlsruhe Institute of Technology KIT
AI Insights
- Large language models may contain software architectural knowledge, but their accuracy and completeness are uncertain. (ML: 0.96)
- Automated knowledge management in the software life cycle is a growing area of research. (ML: 0.96)
- The extraction and analysis of software architectural knowledge from large language models can be used to improve software development processes. (ML: 0.94)
- A taxonomy for design decisions in software architecture documentation is necessary for ensuring that software systems meet their requirements. (ML: 0.92)
- Machine learning-based approaches can be used to improve automated traceability maintenance and consistency checking. (ML: 0.92)
- Automated consistency checking between different views of a software system can help ensure that the system meets its requirements. (ML: 0.91)
- The use of ontologies and knowledge graphs can facilitate the representation and analysis of software architectural knowledge. (ML: 0.90)
- The use of large language models for software architectural knowledge extraction and analysis is an emerging trend. (ML: 0.89)
- Consistency checking between software architecture and informal documentation is crucial for ensuring that software systems meet their requirements. (ML: 0.85)
- Traceability link recovery and maintenance are essential for understanding the relationships between different components of a software system. (ML: 0.84)
Abstract
Software architecture is inherently knowledge-centric. The architectural knowledge is distributed across heterogeneous software artifacts such as requirements documents, design diagrams, code, and documentation, making it difficult for developers to access and utilize this knowledge effectively. Moreover, as systems evolve, inconsistencies frequently emerge between these artifacts, leading to architectural erosion and impeding maintenance activities. We envision an automated pipeline that systematically extracts architectural knowledge from diverse artifacts, links them, identifies and resolves inconsistencies, and consolidates this knowledge into a structured knowledge base. This knowledge base enables critical activities such as architecture conformance checking and change impact analysis, while supporting natural language question-answering to improve access to architectural knowledge. To realize this vision, we plan to develop specialized extractors for different artifact types, design a unified knowledge representation schema, implement consistency checking mechanisms, and integrate retrieval-augmented generation techniques for conversational knowledge access.
Software architecture is inherently knowledge-centric. The architectural knowledge is distributed across heterogeneous software artifacts such as requirements documents, design diagrams, code, and documentation, making it difficult for developers to access and utilize this knowledge effectively. Moreover, as systems evolve, inconsistencies frequently emerge between these artifacts, leading to architectural erosion and impeding maintenance activities. We envision an automated pipeline that systematically extracts architectural knowledge from diverse artifacts, links them, identifies and resolves inconsistencies, and consolidates this knowledge into a structured knowledge base. This knowledge base enables critical activities such as architecture conformance checking and change impact analysis, while supporting natural language question-answering to improve access to architectural knowledge. To realize this vision, we plan to develop specialized extractors for different artifact types, design a unified knowledge representation schema, implement consistency checking mechanisms, and integrate retrieval-augmented generation techniques for conversational knowledge access.
Why we are recommending this paper?
Due to your Interest in Knowledge Management
Due to your Interest in Knowledge Management
Universidad Politcnica de Madrid UPM
AI Insights
- Act-r: A cognitive architecture for modeling cognition. (ML: 0.96)
- Knowledge Graph: A graph-based data structure that represents knowledge in a structured and machine-readable format. (ML: 0.95)
- Lida: A systems-level architecture for cognition, emotion, and learning. (ML: 0.93)
- The proposed methodology for implementing knowledge graphs in ROS 2 systems offers a robust framework for enhancing knowledge management and decision-making in autonomous missions. (ML: 0.92)
- Soar: An architecture for general intelligence. (ML: 0.92)
- The structured representation of tasks and the tailored data extraction facilitated precise control and real-time decision-making, underscoring the methodologyβs potential to improve the operational capabilities of autonomous systems. (ML: 0.87)
- The proposed methodology has the potential to improve the operational capabilities of autonomous systems by enhancing knowledge management and decision-making. (ML: 0.84)
- The practical application of this methodology demonstrates its effectiveness in real-world scenarios, highlighting its versatility in handling complex missions within the Aerostack2 framework. (ML: 0.67)
- The practical application of this methodology is demonstrated through a mission to locate a person using drones, implemented within the Aerostack2 framework. (ML: 0.59)
- Aerostack2: A software framework for developing multi-robot aerial systems. (ML: 0.54)
Abstract
This paper presents a comprehensive methodology for implementing knowledge graphs in ROS 2 systems, aiming to enhance the efficiency and intelligence of autonomous robotic missions. The methodology encompasses several key steps: defining initial and target conditions, structuring tasks and subtasks, planning their sequence, representing task-related data in a knowledge graph, and designing the mission using a high-level language. Each step builds on the previous one to ensure a cohesive process from initial setup to final execution. A practical implementation within the Aerostack2 framework is demonstrated through a simulated search and rescue mission in a Gazebo environment, where drones autonomously locate a target. This implementation highlights the effectiveness of the methodology in improving decision-making and mission performance by leveraging knowledge graphs.
This paper presents a comprehensive methodology for implementing knowledge graphs in ROS 2 systems, aiming to enhance the efficiency and intelligence of autonomous robotic missions. The methodology encompasses several key steps: defining initial and target conditions, structuring tasks and subtasks, planning their sequence, representing task-related data in a knowledge graph, and designing the mission using a high-level language. Each step builds on the previous one to ensure a cohesive process from initial setup to final execution. A practical implementation within the Aerostack2 framework is demonstrated through a simulated search and rescue mission in a Gazebo environment, where drones autonomously locate a target. This implementation highlights the effectiveness of the methodology in improving decision-making and mission performance by leveraging knowledge graphs.
Why we are recommending this paper?
Due to your Interest in Knowledge Management
Due to your Interest in Knowledge Management
Interests not found
We did not find any papers that match the below interests.
Try other terms also consider if the content exists in arxiv.org.
- MECE Mutually Exclusive, Collectively Exhaustive.
- Taxonomy of Products
You can edit or add more interests any time.
π¬ Help Shape Our Pricing
We're exploring pricing options to make this project sustainable. Take 3 minutes to share what you'd be willing to pay (if anything). Your input guides our future investment.
Help us improve your experience!
This project is on its early stages your feedback can be pivotal on the future of the project.
Let us know what you think about this week's papers and suggestions!
Give Feedback