While auxiliary information has become a key to enhancing Large Language Models (LLMs), relatively little is known about how LLMs merge these contexts, specifically contexts generated by LLMs and those retrieved from external sources. To investigate this, we formulate a systematic framework to identify whether LLMs' responses, derived from the integration of generated and retrieved contexts, are attributed to either generated or retrieved contexts. To easily trace the origin of the response, we construct datasets with conflicting contexts, i.e., each question is paired with both generated and retrieved contexts, yet only one of them contains the correct answer. Our experiments reveal a significant bias in several LLMs (GPT-4/3.5 and Llama2) to favor generated contexts, even when they provide incorrect information. We further identify two key factors contributing to this bias: i) contexts generated by LLMs typically show greater similarity to the questions, increasing their likelihood of being selected; ii) the segmentation process used in retrieved contexts disrupts their completeness, thereby hindering their full utilization in LLMs. Our analysis enhances the understanding of how LLMs merge diverse contexts, offering valuable insights for advancing current augmentation methods for LLMs.
相關內容
When large language models are aligned via supervised fine-tuning, they may encounter new factual information that was not acquired through pre-training. It is often conjectured that this can teach the model the behavior of hallucinating factually incorrect responses, as the model is trained to generate facts that are not grounded in its pre-existing knowledge. In this work, we study the impact of such exposure to new knowledge on the capability of the fine-tuned model to utilize its pre-existing knowledge. To this end, we design a controlled setup, focused on closed-book QA, where we vary the proportion of the fine-tuning examples that introduce new knowledge. We demonstrate that large language models struggle to acquire new factual knowledge through fine-tuning, as fine-tuning examples that introduce new knowledge are learned significantly slower than those consistent with the model's knowledge. However, we also find that as the examples with new knowledge are eventually learned, they linearly increase the model's tendency to hallucinate. Taken together, our results highlight the risk in introducing new factual knowledge through fine-tuning, and support the view that large language models mostly acquire factual knowledge through pre-training, whereas fine-tuning teaches them to use it more efficiently.
Gaussian Splatting has garnered widespread attention due to its exceptional performance. Consequently, SLAM systems based on Gaussian Splatting have emerged, leveraging its capabilities for rapid real-time rendering and high-fidelity mapping. However, current Gaussian Splatting SLAM systems usually struggle with large scene representation and lack effective loop closure adjustments and scene generalization capabilities. To address these issues, we introduce NGM-SLAM, the first GS-SLAM system that utilizes neural radiance field submaps for progressive scene expression, effectively integrating the strengths of neural radiance fields and 3D Gaussian Splatting. We have developed neural implicit submaps as supervision and achieve high-quality scene expression and online loop closure adjustments through Gaussian rendering of fused submaps. Our results on multiple real-world scenes and large-scale scene datasets demonstrate that our method can achieve accurate gap filling and high-quality scene expression, supporting both monocular, stereo, and RGB-D inputs, and achieving state-of-the-art scene reconstruction and tracking performance.
In recent years, the rise of large language models (LLMs) has made it possible to directly achieve named entity recognition (NER) without any demonstration samples or only using a few samples through in-context learning (ICL). However, standard ICL only helps LLMs understand task instructions, format and input-label mapping, but neglects the particularity of the NER task itself. In this paper, we propose a new prompting framework P-ICL to better achieve NER with LLMs, in which some point entities are leveraged as the auxiliary information to recognize each entity type. With such significant information, the LLM can achieve entity classification more precisely. To obtain optimal point entities for prompting LLMs, we also proposed a point entity selection method based on K-Means clustering. Our extensive experiments on some representative NER benchmarks verify the effectiveness of our proposed strategies in P-ICL and point entity selection.
Large Multimodal Models (LMMs) have achieved impressive success in visual understanding and reasoning, remarkably improving the performance of mathematical reasoning in a visual context. Yet, a challenging type of visual math lies in the multimodal graph theory problem, which demands that LMMs understand the graphical structures accurately and perform multi-step reasoning on the visual graph. Additionally, exploring multimodal graph theory problems will lead to more effective strategies in fields like biology, transportation, and robotics planning. To step forward in this direction, we are the first to design a benchmark named VisionGraph, used to explore the capabilities of advanced LMMs in solving multimodal graph theory problems. It encompasses eight complex graph problem tasks, from connectivity to shortest path problems. Subsequently, we present a Description-Program-Reasoning (DPR) chain to enhance the logical accuracy of reasoning processes through graphical structure description generation and algorithm-aware multi-step reasoning. Our extensive study shows that 1) GPT-4V outperforms Gemini Pro in multi-step graph reasoning; 2) All LMMs exhibit inferior perception accuracy for graphical structures, whether in zero/few-shot settings or with supervised fine-tuning (SFT), which further affects problem-solving performance; 3) DPR significantly improves the multi-step graph reasoning capabilities of LMMs and the GPT-4V (DPR) agent achieves SOTA performance.
Optimizing Information Freshness in IoT Systems with Update Rate Constraints: A Token-Based Approach
In Internet of Things (IoT) status update systems, where information is sampled and subsequently transmitted from a source to a destination node, the imperative necessity lies in maintaining the timeliness of information and updating the system with optimal frequency. Optimizing information freshness in resource-limited status update systems often involves Constrained Markov Decision Process (CMDP) problems with update rate constraints. Solving CMDP problems, especially with multiple constraints, is a challenging task. To address this, we present a token-based approach that transforms CMDP into an unconstrained MDP, simplifying the solution process. We apply this approach to systems with one and two update rate constraints for optimizing Age of Incorrect Information (AoII) and Age of Information (AoI) metrics, respectively, and explore the analytical and numerical aspects. Additionally, we introduce an iterative triangle bisection method for solving the CMDP problems with two constraints, comparing its results with the token-based MDP approach. Our findings show that the token-based approach yields superior performance over baseline policies, converging to the optimal policy as the maximum number of tokens increases.
Movable antenna (MA) technology, which can reconfigure wireless channels by flexibly moving antenna positions in a specified region, has great potential for improving communication performance. In this paper, we consider a new setup of MAs-enabled multicasting, where we adopt a simple setting in which a linear MA array-enabled source (${\rm{S}}$) transmits a common message to two single-antenna users ${\rm{U}}_1$ and ${\rm{U}}_2$. We aim to maximize the minimum rate among these two users, by jointly optimizing the transmit beamforming and antenna positions at ${\rm{S}}$. Instead of utilizing the widely-used alternating optimization (AO) approach, we reveal, with rigorous proof, that the above two variables can be optimized separately: i) the optimal antenna positions can be firstly determined via the successive convex approximation technique, based on the rule of maximizing the correlation between ${\rm{S}}$-${\rm{U}}_1$ and ${\rm{S}}$-${\rm{U}}_2$ channels; ii) afterwards, the optimal closed-form transmit beamforming can be derived via simple arguments. Compared to AO, this new approach yields the same performance but reduces the computational complexities significantly. Moreover, it can provide insightful conclusions which are not possible with AO.
Autonomous Vehicles (AVs) are prone to revolutionise the transportation industry. However, they must be thoroughly tested to avoid safety violations. Simulation testing plays a crucial role in finding safety violations of Automated Driving Systems (ADSs). This paper proposes PAFOT, a position-based approach testing framework, which generates adversarial driving scenarios to expose safety violations of ADSs. We introduce a 9-position grid which is virtually drawn around the Ego Vehicle (EV) and modify the driving behaviours of Non-Playable Characters (NPCs) to move within this grid. PAFOT utilises a single-objective genetic algorithm to search for adversarial test scenarios. We demonstrate PAFOT on a well-known high-fidelity simulator, CARLA. The experimental results show that PAFOT can effectively generate safety-critical scenarios to crash ADSs and is able to find collisions in a short simulation time. Furthermore, it outperforms other search-based testing techniques by finding more safety-critical scenarios under the same driving conditions within less effective simulation time.
Large Language Models (LLMs) have made significant strides in information acquisition. However, their overreliance on potentially flawed parametric knowledge leads to hallucinations and inaccuracies, particularly when handling long-tail, domain-specific queries. Retrieval Augmented Generation (RAG) addresses this limitation by incorporating external, non-parametric knowledge. Nevertheless, the retrieved long-context documents often contain noisy, irrelevant information alongside vital knowledge, negatively diluting LLMs' attention. Inspired by the supportive role of essential concepts in individuals' reading comprehension, we propose a novel concept-based RAG framework with the Abstract Meaning Representation (AMR)-based concept distillation algorithm. The proposed algorithm compresses the cluttered raw retrieved documents into a compact set of crucial concepts distilled from the informative nodes of AMR by referring to reliable linguistic features. The concepts explicitly constrain LLMs to focus solely on vital information in the inference process. We conduct extensive experiments on open-domain question-answering datasets to empirically evaluate the proposed method's effectiveness. The results indicate that the concept-based RAG framework outperforms other baseline methods, particularly as the number of supporting documents increases, while also exhibiting robustness across various backbone LLMs. This emphasizes the distilled concepts are informative for augmenting the RAG process by filtering out interference information. To the best of our knowledge, this is the first work introducing AMR to enhance the RAG, presenting a potential solution to augment inference performance with semantic-based context compression.
As the complexity and destructiveness of Advanced Persistent Threat (APT) increase, there is a growing tendency to identify a series of actions undertaken to achieve the attacker's target, called attack investigation. Currently, analysts construct the provenance graph to perform causality analysis on Point-Of-Interest (POI) event for capturing critical events (related to the attack). However, due to the vast size of the provenance graph and the rarity of critical events, existing attack investigation methods suffer from problems of high false positives, high overhead, and high latency. To this end, we propose SPARSE, an efficient and real-time system for constructing critical component graphs (i.e., consisting of critical events) from streaming logs. Our key observation is 1) Critical events exist in a suspicious semantic graph (SSG) composed of interaction flows between suspicious entities, and 2) Information flows that accomplish attacker's goal exist in the form of paths. Therefore, SPARSE uses a two-stage framework to implement attack investigation (i.e., constructing the SSG and performing path-level contextual analysis). First, SPARSE operates in a state-based mode where events are consumed as streams, allowing easy access to the SSG related to the POI event through semantic transfer rule and storage strategy. Then, SPARSE identifies all suspicious flow paths (SFPs) related to the POI event from the SSG, quantifies the influence of each path to filter irrelevant events. Our evaluation on a real large-scale attack dataset shows that SPARSE can generate a critical component graph (~ 113 edges) in 1.6 seconds, which is 2014 X smaller than the backtracking graph (~ 227,589 edges). SPARSE is 25 X more effective than other state-of-the-art techniques in filtering irrelevant edges.
Deep Learning has revolutionized the fields of computer vision, natural language understanding, speech recognition, information retrieval and more. However, with the progressive improvements in deep learning models, their number of parameters, latency, resources required to train, etc. have all have increased significantly. Consequently, it has become important to pay attention to these footprint metrics of a model as well, not just its quality. We present and motivate the problem of efficiency in deep learning, followed by a thorough survey of the five core areas of model efficiency (spanning modeling techniques, infrastructure, and hardware) and the seminal work there. We also present an experiment-based guide along with code, for practitioners to optimize their model training and deployment. We believe this is the first comprehensive survey in the efficient deep learning space that covers the landscape of model efficiency from modeling techniques to hardware support. Our hope is that this survey would provide the reader with the mental model and the necessary understanding of the field to apply generic efficiency techniques to immediately get significant improvements, and also equip them with ideas for further research and experimentation to achieve additional gains.
小貼士
登錄享
注冊地址： 北京市海淀區羊坊店路18號2幢3層301-191