In reinforcement learning, especially in sparse-reward domains, many environment steps are required to observe reward information. In order to increase the frequency of such observations, "potential-based reward shaping" (PBRS) has been proposed as a method of providing a more dense reward signal while leaving the optimal policy invariant. However, the required potential function must be carefully designed with task-dependent knowledge to not deter training performance. In this work, we propose a bootstrapped method of reward shaping, termed BS-RS, in which the agent's current estimate of the state-value function acts as the potential function for PBRS. We provide convergence proofs for the tabular setting, give insights into training dynamics for deep RL, and show that the proposed method improves training speed in the Atari suite.
论文检索
输入标题、作者或关键词，从 11,272 篇学术成果中精准定位
Little Is Enough: Boosting Privacy by Sharing Only Hard Labels in Federated Semi-Supervised Learning
PDF ↗In many critical applications, sensitive data is inherently distributed and cannot be centralized due to privacy concerns. A wide range of federated learning approaches have been proposed to train models locally at each client without sharing their sensitive data, typically by exchanging model parameters, or probabilistic predictions (soft labels) on a public dataset or a combination of both. However, these methods still disclose private information and restrict local models to those that can be trained using gradient-based methods. We propose a federated co-training (FEDCT) approach that improves privacy by sharing only definitive (hard) labels on a public unlabeled dataset. Clients use a consensus of these shared labels as pseudo-labels for local training. This federated co-training approach empirically enhances privacy without compromising model quality. In addition, it allows the use of local models that are not suitable for parameter aggregation in traditional federated learning, such as gradient-boosted decision trees, rule ensembles, and random forests. Furthermore, we observe that FEDCT performs effectively in federated fine-tuning of large language models, where its pseudo-labeling mechanism is particularly beneficial. Empirical evaluations and theoretical analyses suggest its applicability across a range of federated learning scenarios.
Exponential-family harmoniums (EFHs) generalize the restricted Boltzmann machine beyond Bernoulli random variables to other exponential families. Here we show how to extend the EFH beyond standard exponential families (Poisson, Gaussian, etc.), by allowing the sufficient statistics for the hidden units to be arbitrary functions of the observed data, parameterized by deep neural networks. This rules out the standard sampling scheme, block Gibbs sampling, so we replace it with a form of Langevin dynamics within Gibbs, inspired by a recent method for training Gaussian restricted Boltzmann machines (GRBMs). With Gibbs-Langevin, the GRBM can successfully model small datasets like MNIST and CelebA-32, but struggles with CIFAR-10, and cannot scale to larger images because it lacks convolutions. In contrast, our neural-network EFHs (NN-EFHs) generate high-quality samples from CIFAR-10 and scale well to CelebA-HQ. On these datasets, the NN-EFH achieves FID scores that are 25--50% lower than a standard energy-based model with a similar neural-network architecture and the same number of parameters; and competitive with noise-conditional score networks, which utilize more complex neural networks (U-nets) and require considerably more sampling steps.
Cross-Domain Few-Shot Learning (CD-FSL) aims to transfer knowledge from seen source domains to unseen target domains, which is crucial for evaluating the generalization and robustness of models. Recent studies focus on utilizing visual styles to bridge the domain gap between different domains. However, the serious dilemma of gradient instability and local optimization problem occurs in those style-based CD-FSL methods. This paper addresses these issues and proposes a novel crop-global style perturbation method, called Self-Versatility Adversarial Style Perturbation (SVasP), which enhances the gradient stability and escapes from poor sharp minima jointly. Specifically, SVasP simulates more diverse potential target domain adversarial styles via diversifying input patterns and aggregating localized crop style gradients, to serve as global style perturbation stabilizers within one image, a concept we refer to as self-versatility. Then a novel objective function is proposed to maximize visual discrepancy while maintaining semantic consistency between global, crop, and adversarial features. Having the stabilized global style perturbation in the training phase, one can obtain a flattened minima in the loss landscape, boosting the transferability of the model to the target domains. Extensive experiments on multiple benchmark datasets demonstrate that our method significantly outperforms existing state-of-the-art methods.
Physical commonsense is an essential aspect of human cognition, involving an intuitive understanding of the physical properties and interactions of everyday objects and materials. Though physical commonsense reasoning should inherently be a multisensory task, integrating both video and audio signals, existing physical audiovisual commonsense reasoning (PACR) models predominantly rely on visual information. This reliance leads to spurious correlations and undermines the models’ reasoning and generalization abilities. To counteract this, we introduce a model-agnostic Counterfactual Physical Audiovisual Commonsense Reasoning (CF-PACR) framework aimed at mitigating visual bias-induced spurious effects. Specifically, we construct a traditional PACR model using both audio and visual information as the factual reasoning model. Subsequently, in the counterfactual reasoning model, we isolate visual information to estimate direct effects. Finally, we subtract the direct effects from the total effects across modalities to derive indirect effects, thereby mitigating visual biases. Extensive experiments validate the effectiveness and generalizability of CF-PACR in alleviating the spurious correlations between visual modality and model predictions.
There has been a longstanding dispute over which formalism is the best for representing knowledge in AI. The well-known “declarative vs. procedural controversy” is concerned with the choice of utilizing declarations or procedures as the primary mode of knowledge representation. The ongoing debate between symbolic AI and connectionist AI also revolves around the question of whether knowledge should be represented implicitly (e.g., as parametric knowledge in deep learning and large language models) or explicitly (e.g., as logical theories in traditional knowledge representation and reasoning). To address these issues, we propose a general framework to capture various knowledge representation formalisms in which we are interested. Within the framework, we find a family of universal knowledge representation formalisms, and prove that all universal formalisms are recursively isomorphic. Moreover, we show that all pairwise intertranslatable formalisms that admit the padding property are also recursively isomorphic. These imply that, up to an offline compilation, all universal (or natural and equally expressive) representation formalisms are in fact the same, which thus provides a partial answer to the aforementioned dispute.
Knowledge-based questions are typically employed to evaluate LLM's knowledge boundaries; meanwhile, numerous studies focus on question generation as a means to enhance the capabilities of both models and individuals. However, there is a lack of in-depth exploration about what constitutes a good question from the perspective of knowledge cognition. This paper proposes aligning the complete knowledge underlying questions with educational criteria effectively employed in physics courses, thereby developing novel knowledge-intensive metrics of question quality. To this end, we propose Meta-Fact Checking (MFC), which transforms questions into knowledge graph (KG) triples utilizing LLMs through few-shot prompting, thereby quantifying question quality based on the patterns observed within these triples. MFC introduces a novel interaction mechanism for KGs that communicates meta-facts, illustrating the types of knowledge that KGs can offer to the LLM for reasoning questions, rather than relying solely on the original triples. This strategy ensures that MFC remains unaffected by unexplored triples that LLM has not yet encountered within KGs compared to the retrieve-while-reasoning routine. Experiments across multiple datasets and LLMs demonstrate that MFC significantly improves the accuracy and efficiency of both question answering and assessing. This research marks a pioneering effort to automate the evaluation of question quality based on cognitive capabilities.
Text-based knowledge graph completion methods take advantage of pre-trained language models (PLM) to enhance intrinsic semantic connections of raw triplets with detailed text descriptions. Typical methods in this branch map an input query (textual descriptions associated with an entity and a relation) and its candidate entities into feature vectors, respectively, and then maximize the probability of valid triples. These methods are gaining promising performance and increasing attention for the rapid development of large language models. According to the property of the language models, the more related and specific context information the input query provides, the more discriminative the resultant embedding will be. In this paper, through observation and validation, we find a neglected fact that the relation-aware neighbors of the head entities in queries could act as effective contexts for more precise link prediction. Driven by this finding, we propose a relation-aware anchor enhanced knowledge graph completion method (RAA-KGC). Specifically, in our method, to provide a reference of what might the target entity be like, we first generate anchor entities within the relation-aware neighborhood of the head entity. Then, by pulling the query embedding towards the neighborhoods of the anchors, it is tuned to be more discriminative for target entity matching. The results of our extensive experiments not only validate the efficacy of RAA-KGC but also reveal that by integrating our relation-aware anchor enhancement strategy, the performance of current leading methods can be notably enhanced without substantial modifications.
SPAC: Sparse Partitioning and Adaptive Core Tensor Pruning Model for Knowledge Graph Completion
PDF ↗Tensor decomposition (TD) models are promising solutions for knowledge graph completion due to their simple structures but powerful representation capacities. The TD models typically adopt Tucker decomposition with a structured core tensor. Some models with a sparse core tensor, such as DistMult and ComplEx, are too simple and thus limit the interaction between embedding components, while other models with a dense core tensor are too complex and may lead to significant overfitting. To address these issues, we propose a new TD model called SPAC (Sparse Partitioning and Adaptive Core tensor pruning) model for knowledge graph completion. Specifically, SPAC captures coarse and fine-grained semantic information using a hybrid core tensor, where auxiliary cores are used to model sparse interactions and main cores for dense interactions. Moreover, SPAC introduces a gating mechanism to control the output of intermediate variables, enhancing the interaction between different partition groups. Furthermore, SPAC employs an adaptive pruning approach to dynamically adjust the shape of the core tensor. Due to the elaborate model design, the proposed TD model enhances expressive capacity and reduces the number of parameters in the core tensor. Experiments are conducted on datasets FB15k-237, WN18RR, and YAGO3-10. The results demonstrate that SPAC outperforms state-of-the-art tensor decomposition models, including MEIM and Tucker models. A series of ablation studies show that the gating mechanism and adaptive pruning strategy in SPAC are crucial for the performance improvement.
Querying temporal data has recently gained traction in several artificial intelligence applications. As operational domains of intelligent agents are constantly being expanded, there is a strong need for representing domain knowledge. This comes in the form of ontologies, which are predominantly expressed in description logics and enrich time-stamped data to temporal knowledge bases. For modeling highly complex system environments, expressive description logics are often the formalism of choice. Querying such temporal knowledge bases is a challenging task, but recently a first practical solution has been put forward. We propose a novel approach to the query answering problem based on two well-known rewriting rules from temporal logic. After a careful theoretical analysis of our algorithm, we show in a practical evaluation on several benchmarks that it outperforms state of the art, sometimes by orders of magnitude. Based on our findings, we also propose a fragment of temporal conjunctive queries which guides users towards well-performing queries.
Prompt optimization automatically refines prompting expressions, unlocking the full potential of LLMs in downstream tasks. However, current prompt optimization methods are costly to train and lack sufficient interpretability. This paper proposes enhancing LLMs' reasoning performance by eliciting their causal inference ability from prompting instructions to correct answers. Specifically, we introduce the Self-Causal Instruction Enhancement (SCIE) method, which enables LLMs to generate high-quality, low-quantity observational data, then estimates the causal effect based on these data, and ultimately generates instructions with the optimized causal effect. In SCIE, the instructions are treated as the treatment, and textual features are used to process natural language, establishing causal relationships through treatments between instructions and downstream tasks. Additionally, we propose applying Object-Relational (OR) principles, where the uncovered causal relationships are treated as the inheritable class across task objects, ensuring low-cost reusability. Extensive experiments demonstrate that our method effectively generates instructions that enhance reasoning performance with reduced training cost of prompts, leveraging interpretable textual features to provide actionable insights.
DatalogMTL is a powerful rule-based language for temporal reasoning. Due to its high expressive power and flexible modeling capabilities, it is suitable for a wide range of applications, including tasks from industrial and financial sectors. However, due its high computational complexity, practical reasoning in DatalogMTL is highly challenging. To address this difficulty, we introduce a new reasoning method for DatalogMTL which exploits the magic sets technique—a rewriting approach developed for (non-temporal) Datalog to simulate top-down evaluation with bottom-up reasoning. We have implemented this approach and evaluated it on publicly available benchmarks, showing that the proposed approach significantly and consistently outperformed state-of-the-art reasoning techniques.
Traffic Scenario Logic: A Spatial-Temporal Logic for Modeling and Reasoning of Urban Traffic Scenarios
PDF ↗Formal representations of traffic scenarios can be used to generate test cases for the safety verification of autonomous driving. However, most existing methods are limited to highway or highly simplified intersection scenarios due to the intricacy and diversity of traffic scenarios. In response, we propose Traffic Scenario Logic (TSL), which is a spatial-temporal logic designed for modeling and reasoning of urban pedestrian-free traffic scenarios. TSL provides a formal representation of the urban road network that can be derived from OpenDRIVE, i.e., the de facto industry standard of high-definition maps for autonomous driving, enabling the representation of a broad range of traffic scenarios without discretization approximations. We implemented the reasoning of TSL using Telingo, i.e., a solver for temporal programs based on Answer Set Programming, and tested it on different urban road layouts. Demonstrations show the effectiveness of TSL in test scenario generation and its potential value in areas like decision-making and control verification of autonomous driving. The code for TSL reasoning has been open-sourced.
SR-FoT: A Syllogistic-Reasoning Framework of Thought for Large Language Models Tackling Knowledge-based Reasoning Tasks
PDF ↗Deductive reasoning is a crucial logical capability that assists us in solving complex problems based on existing knowledge. Although augmented by Chain-of-Thought prompts, Large Language Models (LLMs) might not follow the correct reasoning paths. Enhancing the deductive reasoning abilities of LLMs, and leveraging their extensive built-in knowledge for various reasoning tasks, remains an open question. Attempting to mimic the human deductive reasoning paradigm, we propose a multi-stage Syllogistic-Reasoning Framework of Thought (SR-FoT) that enables LLMs to perform syllogistic deductive reasoning to handle complex knowledge-based reasoning tasks. Our SR-FoT begins by interpreting the question and then uses the interpretation and the original question to propose a suitable major premise. It proceeds by generating and answering minor premise questions in two stages to match the minor premises. Finally, it guides LLMs to use the previously generated major and minor premises to perform syllogistic deductive reasoning to derive the answer to the original question. Extensive and thorough experiments on knowledge-based reasoning tasks have demonstrated the effectiveness and advantages of our SR-FoT.
Datalog is a logic programming language widely used in knowledge representation and reasoning (KRR), program analysis, and social media mining due to its expressiveness and high performance. Traditionally, Datalog engines use either row-oriented or column-oriented storage. Engines like VLog and Nemo favor column-oriented storage for efficiency on limited-resource machines, while row-oriented engines like Soufflé use advanced datastructures with locking to perform better on multi-core CPUs. The advent of modern datacenter GPUs, such as the NVIDIA H100 with its ability to run over 16k threads simultaneously and high memory bandwidth, has reopened the debate on which storage layout is more effective. This paper presents the first column-oriented Datalog engines tailored to the strengths of modern GPUs. We present VFLog, a CUDA-based Datalog runtime library with a column-oriented GPU datastructure that supports all necessary relational algebra operations. Our results demonstrate over 200x performance gains over SOTA CPU-based column-oriented Datalog engines and a 2.5x speedup over GPU Datalog engines in various workloads, including KRR.
MTVHunter: Smart Contracts Vulnerability Detection Based on Multi-Teacher Knowledge Translation
PDF ↗Smart contracts, closely intertwined with cryptocurrency transactions, have sparked widespread concerns about considerable financial losses of security issues. To counteract this, a variety of tools have been developed to identify vulnerability in smart contract. However, they fail to overcome two challenges at the same time when faced with smart contract bytecode: (i) strong interference caused by enormous non-relevant instructions; (ii) missing semantics of bytecode due to incomplete data and control flow dependencies. In this paper, we propose a multi-teacher based bytecode vulnerability detection method, namely Multi-Teacher Vulnerability Hunter (MTVHunter), which delivers effective denoising and missing semantic to bytecode under multi-teacher guidance. Specifically, we first propose an instruction denoising teacher to eliminate noise interference by abstract vulnerability pattern and further reflect in contract embeddings. Secondly, we design a novel semantic complementary teacher with neuron distillation, which effectively extracts necessary semantic from source code to replenish the bytecode. Particularly, the proposed neuron distillation accelerate this semantic filling by turning the knowledge transition into a regression task. We conduct experiments on 229,178 real-world smart contracts that concerns four types of common vulnerabilities. Extensive experiments show MTVHunter achieves significantly performance gains over state-of-the-art approaches.
Knowledge stored in large language models requires timely updates to reflect the dynamic nature of real-world information. To update the knowledge, most knowledge editing methods focus on the low layers, since recent probes into the knowledge recall process reveal that the answer information is enriched in low layers. However, these probes only and could only reveal critical recall stages for the original answers, while the goal of editing is to rectify model's prediction for the target answers. This inconsistency indicates that both the probe approaches and the associated editing methods are deficient. To mitigate the inconsistency and identify critical editing regions, we propose a contrast-based probe approach, and locate two crucial stages where the model behavior diverges between the original and target answers: Information Enrichment in low layers and Probability Promotion in high layers. Building upon the insights, we develop the Joint knowledge Editing for information Enrichment and probability Promotion (JEEP) method, which jointly edits both the low and high layers to modify the two critical recall stages. Considering the mutual interference and growing forgetting due to dual modifications, JEEP is designed to ensure that updates to distinct regions share the same objectives and are complementary. We rigorously evaluate JEEP by editing up to thousands of facts on various models, i.e., GPT-J (6B) and LLaMA (7B), and addressing diverse editing objectives, i.e., adding factual and counterfactual knowledge. In all tested scenarios, JEEP achieves best performances, validating the effectiveness of the revealings of our probe approach and the designs of our editing method.
Learning Theorem Rationale for Improving the Mathematical Reasoning Capability of Large Language Models
PDF ↗Large language models (LLMs) have achieved significant progress in mathematical reasoning, especially in elementary math. However, they remain indisposed on tackling complex questions at high-school or college levels, which put forward a more advanced requirement of mastering relevant mathematical theorems. For we humans, whether selecting the appropriate theorems according to the provided question is a crucial factor affecting the quality of the ultimate solutions, yet which has been neglected by previous research in the field of LLM reasoning. In this paper, we propose a novel approach to enhance the LLM's capability of utilizing the mathematical theorems to specific problems, which we refer to as Theorem Rationale (TR). To this end, a new dataset encompassing problem-theorem-solution triples is deliberately established for transferring principles of TR. Furthermore, we develop an evolving strategy to boost hierarchical instructions oriented on the theorems to alleviate difficulty in acquiring the curated data and facilitate the digestion of theorem application from various perspectives. Evaluations on a wide range of public datasets exhibit that the model fine-tuned with our dataset achieves consistent improvements at varying mathematical levels compared to the backbone. And further ablation studies illustrate the effectiveness of our proposed evolutionary strategies on enhancing the model's capability of math problem-solving. Overall, extensive experiments reveal the potential of our proposed method which highlights the significance of aligning the problems with the concrete theorems for LLMs to alleviate hallucination and improve the models' mathematical reasoning capabilities.
In this paper, we introduce a syntactic framework for analyzing and handling inconsistencies in propositional bases. Our approach focuses on examining the relationships between variable occurrences within conflicts. We propose two dual concepts: Minimal Inconsistency Relation (MIR) and Maximal Consistency Relation (MCR). Each MIR is a minimal equivalence relation on variable occurrences that results in inconsistency, while each MCR is a maximal equivalence relation designed to prevent inconsistency. Notably, MIRs capture conflicts overlooked by minimal inconsistent subsets. Using MCRs, we develop a series of non-explosive inference relations. The main strategy involves restoring consistency by modifying the propositional base according to each MCR, followed by employing the classical inference relation to derive conclusions. Additionally, we propose an unusual semantics that assigns truth values to variable occurrences instead of the variables themselves. The associated inference relations are established through Boolean interpretations compatible with the occurrence-based models.
In recent years, Machine Learning (ML) models have achieved remarkable success in various domains. However, these models also tend to demonstrate unsafe behaviors, precluding their deployment in safety-critical systems. To cope with this issue, ample research focuses on developing methods that guarantee the safe behaviour of a given ML model. A prominent example is shielding which incorporates an ex- ternal component (a “shield”) that blocks unwanted behavior. Despite significant progress, shielding suffers from a main setback: it is currently geared towards properties encoded solely in propositional logics (e.g., LTL) and is unsuitable for richer logics. This, in turn, limits the widespread applicability of shielding in many real-world systems. In this work, we address this gap, and extend shielding to LTL modulo theories, by building upon recent advances in reactive synthesis modulo theories. This allowed us to develop a novel approach for generating shields conforming to complex safety specifications in these more expressive, logics. We evaluated our shields and demonstrate their ability to handle rich data with temporal dynamics. To the best of our knowledge, this is the first approach for synthesizing shields for such expressivity.