We consider the problem of sampling from a discrete and structured distribution as a sequential decision problem, where the objective is to find a stochastic policy such that objects are sampled at the end of this sequential process proportionally to some predefined reward. While we could use maximum entropy Reinforcement Learning (MaxEnt RL) to solve this problem for some distributions, it has been shown that in general, the distribution over states induced by the optimal policy may be biased in cases where there are multiple ways to generate the same object. To address this issue, Generative Flow Networks (GFlowNets) learn a stochastic policy that samples objects proportionally to their reward by approximately enforcing a conservation of flows across the whole Markov Decision Process (MDP). In this paper, we extend recent methods correcting the reward in order to guarantee that the marginal distribution induced by the optimal MaxEnt RL policy is proportional to the original reward, regardless of the structure of the underlying MDP. We also prove that some flow-matching objectives found in the GFlowNet literature are in fact equivalent to well-established MaxEnt RL algorithms with a corrected reward. Finally, we study empirically the performance of multiple MaxEnt RL and GFlowNet algorithms on multiple problems involving sampling from discrete distributions.

Federated Learning is a machine learning approach that enables the training of a deep learning model among several participants with sensitive data that wish to share their own knowledge without compromising the privacy of their data. In this research, the authors employ a secured Federated Learning method with an additional layer of privacy and proposes a method for addressing the non-IID challenge. Moreover, differential privacy is compared with chaotic-based encryption as layer of privacy. The experimental approach assesses the performance of the federated deep learning model with differential privacy using both IID and non-IID data. In each experiment, the Federated Learning process improves the average performance metrics of the deep neural network, even in the case of non-IID data.

Artificial intelligence applications in healthcare are increasing every day. These applications have the ability to advance the healthcare industry by, for instance, supporting clinical decision making, risk prediction, developing early warning systems for patients, increasing the accuracy and timeliness of diagnosis, improving patient-physician interaction, and optimizing operations and resource allocation [21]. Federated learning is a new approach for distributed artificial intelligence that aims to have several agents train a deep learning model in a collaborative and secure way, without sharing any private data. This training is done the following way: a central server defines a deep learning model and sends it to the agents, who train the model in their private data. Then, they send the parameters of the model (weights or gradients) back to the server, who aggregates these data in order to find a global federated model, which in turn is delivered back to the agents to be retrained in their data. This process is iterated until convergence. In the initial definition of the federated learning approach, the aggregation step is done by averaging the model parameters. Nevertheless, other aggregation methods may be of more interest since they can improve the performance of the model by giving more weight to different agents depending on their size or the performance of the local models in their data.

Improving the controllability, portability, and inference speed of diffusion language models (DLMs) is a key challenge in natural language generation. While recent research has shown significant success in complex text generation with language models, the memory and computational power are still very demanding and fall short of expectations, which naturally results in low portability and instability for the models. To mitigate these issues, numerous well-established methods were proposed for neural network quantization. To further enhance their portability of independent deployment as well as improve their stability evaluated by language perplexity, we propose a novel approach called the Quantized Embedding Controllable Diffusion Language Model (QE-CDLM). QE-CDLM builds upon the recent successful controllable DLMs by remodeling the task-specific embedding space via quantization. This leads to a gradient-based controller for the generation tasks, and more stable intermediate latent variables are obtained, which naturally brings in an accelerated convergence as well as better controllability. Additionally, the adaption fine-tuning method is employed to reduce tunable weights. Experimental results on five challenging fine-grained control tasks demonstrate that QE-CDLM compares favorably to existing methods in terms of quality and feasibility, achieving better perplexity and lightweight fine-tuning.

Multi-span answer extraction, also known as the task of multi-span question answering (MSQA), is critical for real-world applications, as it requires extracting multiple pieces of information from a text to answer complex questions. Despite the active studies and rapid progress in English MSQA research, there is a notable lack of publicly available MSQA benchmark in Chinese. Previous efforts for constructing MSQA datasets predominantly emphasized entity-centric contextualization, resulting in a bias towards collecting factoid questions and potentially overlooking questions requiring more detailed descriptive responses. To overcome these limitations, we present CLEAN, a comprehensive Chinese multi-span question answering dataset that involves a wide range of open-domain subjects with a substantial number of instances requiring descriptive answers. Additionally, we provide established models from relevant literature as baselines for CLEAN. Experimental results and analysis show the characteristics and challenge of the newly proposed CLEAN dataset for the community. Our dataset, CLEAN, will be publicly released at zhiyiluo.site/misc/clean_v1.0_ sample.json.

Mitigating the hallucinations of Large Language Models (LLMs) and enhancing them is a crucial task. Although some existing methods employ model self-enhancement techniques, they fall short of effectively addressing unknown factual hallucinations. Using Knowledge Graph (KG) enhancement approaches fails to address the generalization across different KG sources and the enhancement of open-ended answer questions simultaneously. To tackle these limitations, there is a framework that combines Pseudo-Graph Generation and Atomic Knowledge Verification proposed. The enhancement of LLM using KG in an open-ended question-answering setting is implemented by leveraging the Pseudo-Graph Generation. Atomic Knowledge Verification utilizes atomic-level knowledge querying and verification to achieve generalizability under different KG sources. Compared to the baseline, this approach yields a minimum improvement of 11.5 in the ROUGE-L score for open-ended questions. For precise questions, we observe a minimum accuracy improvement of 7.5. Moreover, there is also demonstration that this framework exhibits generalizability across different KG sources. In summary, our results pave the way for enhancing LLMs by incorporating Pseudo- and Multisource-KGs, particularly in the context of open-ended questions.

All text-based language problems can be reduced to either generation or embedding. Current models only perform well at one or the other. We introduce generative representational instruction tuning (GRIT) whereby a large language model is trained to handle both generative and embedding tasks by distinguishing between them through instructions. Compared to other open models, our resulting GritLM 7B sets a new state of the art on the Massive Text Embedding Benchmark (MTEB) and outperforms all models up to its size on a range of generative tasks. By scaling up further, GritLM 8x7B outperforms all open generative language models that we tried while still being among the best embedding models. Notably, we find that GRIT matches training on only generative or embedding data, thus we can unify both at no performance loss. Among other benefits, the unification via GRIT speeds up Retrieval-Augmented Generation (RAG) by > 60% for long documents, by no longer requiring separate retrieval and generation models. Models, code, etc. are freely available at https://github.com/ContextualAI/gritlm.

Ensuring artificial intelligence behaves in such a way that is aligned with human values is commonly referred to as the alignment challenge. Prior work has shown that rational agents, behaving in such a way that maximizes a utility function, will inevitably behave in such a way that is not aligned with human values, especially as their level of intelligence goes up. Prior work has also shown that there is no "one true utility function"; solutions must include a more holistic approach to alignment. This paper describes oblivious agents: agents that are architected in such a way that their effective utility function is an aggregation of a known and hidden sub-functions. The hidden component, to be maximized, is internally implemented as a black box, preventing the agent from examining it. The known component, to be minimized, is knowledge of the hidden sub-function. Architectural constraints further influence how agent actions can evolve its internal environment model. We show that an oblivious agent, behaving rationally, constructs an internal approximation of designers' intentions (i.e., infers alignment), and, as a consequence of its architecture and effective utility function, behaves in such a way that maximizes alignment; i.e., maximizing the approximated intention function. We show that, paradoxically, it does this for whatever utility function is used as the hidden component and, in contrast with extant techniques, chances of alignment actually improve as agent intelligence grows.

Large Language Models (LLMs) can make up answers that are not real, and this is known as hallucination. This research aims to see if, how, and to what extent LLMs are aware of hallucination. More specifically, we check whether and how an LLM reacts differently in its hidden states when it answers a question right versus when it hallucinates. To do this, we introduce an experimental framework which allows examining LLM's hidden states in different hallucination situations. Building upon this framework, we conduct a series of experiments with language models in the LLaMA family (Touvron et al., 2023). Our empirical findings suggest that LLMs react differently when processing a genuine response versus a fabricated one. We then apply various model interpretation techniques to help understand and explain the findings better. Moreover, informed by the empirical observations, we show great potential of using the guidance derived from LLM's hidden representation space to mitigate hallucination. We believe this work provides insights into how LLMs produce hallucinated answers and how to make them occur less often.

This paper presents a study of the effectiveness of Neural Network (NN) techniques for deconvolution inverse problems relevant for applications in Quantum Field Theory, but also in more general contexts. We consider NN's asymptotic limits, corresponding to Gaussian Processes (GPs), where non-linearities in the parameters of the NN can be neglected. Using these resulting GPs, we address the deconvolution inverse problem in the case of a quantum harmonic oscillator simulated through Monte Carlo techniques on a lattice. In this simple toy model, the results of the inversion can be compared with the known analytical solution. Our findings indicate that solving the inverse problem with a NN yields less performing results than those obtained using the GPs derived from NN's asymptotic limits. Furthermore, we observe the trained NN's accuracy approaching that of GPs with increasing layer width. Notably, one of these GPs defies interpretation as a probabilistic model, offering a novel perspective compared to established methods in the literature. Our results suggest the need for detailed studies of the training dynamics in more realistic set-ups.