Despite their impressive capacities, Large language models (LLMs) often struggle with the hallucination issue of generating inaccurate or fabricated content even when they possess correct knowledge. In this paper, we extend the exploration of the correlation between hidden-state prediction changes and output factuality into a deeper, token-wise level. Based on the insights , we propose cross-layer Entropy eNhanced Decoding (END), a decoding method that mitigates hallucinations without requiring extra training. END leverages inner probability changes across layers to individually quantify the factual knowledge required for each candidate token, and adjusts the final predicting distribution to prioritize tokens with higher factuality. Experiments on both hallucination and QA benchmarks demonstrate that END significantly enhances the truthfulness and informativeness of generated content while maintaining robust QA accuracy. Moreover, our work provides a deeper perspective on understanding the correlations between inherent knowledge and output factuality.

Neural representations induced by naturalistic stimuli offer insights into how humans respond to peripheral stimuli in daily life. The key to understanding the general neural mechanisms underlying naturalistic stimuli processing involves aligning neural activities across individuals and extracting inter-subject shared neural representations. Targeting the Electroencephalogram (EEG) technique, known for its rich spatial and temporal information, this study presents a general framework for Contrastive Learning of Shared SpatioTemporal EEG Representations across individuals (CL-SSTER). Harnessing the representational capabilities of contrastive learning, CL-SSTER utilizes a neural network to maximize the similarity of EEG representations across individuals for identical stimuli, contrasting with those for varied stimuli. The network employed spatial and temporal convolutions to simultaneously learn the spatial and temporal patterns inherent in EEG. The versatility of CL-SSTER was demonstrated on three EEG datasets, including a synthetic dataset, a speech audio EEG dataset, and an emotional video EEG dataset. CL-SSTER attained the highest inter-subject correlation (ISC) values compared to the state-of-the-art ISC methods. The latent representations generated by CL-SSTER exhibited reliable spatiotemporal EEG patterns, which can be explained by specific aspects of the stimuli. CL-SSTER serves as an interpretable and scalable foundational framework for the identification of inter-subject shared neural representations in the realm of naturalistic neuroscience.

Long-context processing is a critical ability that constrains the applicability of large language models. Although there exist various methods devoted to enhancing the long-context processing ability of large language models (LLMs), they are developed in an isolated manner and lack systematic analysis and integration of their strengths, hindering further developments. In this paper, we introduce UniMem, a unified framework that reformulates existing long-context methods from the view of memory augmentation of LLMs. UniMem is characterized by four key dimensions: Memory Management, Memory Writing, Memory Reading, and Memory Injection, providing a systematic theory for understanding various long-context methods. We reformulate 16 existing methods based on UniMem and analyze four representative methods: Transformer-XL, Memorizing Transformer, RMT, and Longformer into equivalent UniMem forms to reveal their design principles and strengths. Based on these analyses, we propose UniMix, an innovative approach that integrates the strengths of these algorithms. Experimental results show that UniMix achieves superior performance in handling long contexts with significantly lower perplexity than baselines.

Can replay, as a widely observed neural activity pattern in brain regions, particularly in the hippocampus and neocortex, emerge in an artificial agent? If yes, does it contribute to the tasks? In this work, without heavy dependence on complex assumptions, we discover naturally emergent replay under task-optimized paradigm using a recurrent neural network-based reinforcement learning model, which mimics the hippocampus and prefrontal cortex, as well as their intercommunication and the sensory cortex input. The emergent replay in the hippocampus, which results from the episodic memory and cognitive map as well as environment observations, well resembles animal experimental data and serves as an effective indicator of high task performance. The model also successfully reproduces local and nonlocal replay, which matches the human experimental data. Our work provides a new avenue for understanding the mechanisms behind replay.

Nowadays, open-source large language models like LLaMA have emerged. Recent developments have incorporated supervised fine-tuning (SFT) and reinforcement learning fine-tuning (RLFT) to align these models with human goals. However, SFT methods treat all training data with mixed quality equally, while RLFT methods require high-quality pairwise or ranking-based preference data. In this study, we present a novel framework, named OpenChat, to advance open-source language models with mixed-quality data. Specifically, we consider the general SFT training data, consisting of a small amount of expert data mixed with a large proportion of sub-optimal data, without any preference labels. We propose the C(onditioned)-RLFT, which regards different data sources as coarse-grained reward labels and learns a class-conditioned policy to leverage complementary data quality information. Interestingly, the optimal policy in C-RLFT can be easily solved through single-stage, RL-free supervised learning, which is lightweight and avoids costly human preference labeling. Through extensive experiments on three standard benchmarks, our openchat-13b fine-tuned with C-RLFT achieves the highest average performance among all 13b open-source language models. Moreover, we use AGIEval to validate the model generalization performance, in which only openchat-13b surpasses the base model. Finally, we conduct a series of analyses to shed light on the effectiveness and robustness of OpenChat. Our code, data, and models are publicly available at https://github.com/imoneoi/openchat.

Optimization of discrete structures aims at generating a new structure with the better property given an existing one, which is a fundamental problem in machine learning. Different from the continuous optimization, the realistic applications of discrete optimization (e.g., text generation) are very challenging due to the complex and long-range constraints, including both syntax and semantics, in discrete structures. In this work, we present SAGS, a novel Simulated Annealing framework for Graph and Sequence optimization. The key idea is to integrate powerful neural networks into metaheuristics (e.g., simulated annealing, SA) to restrict the search space in discrete optimization. We start by defining a sophisticated objective function, involving the property of interest and pre-defined constraints (e.g., grammar validity). SAGS searches from the discrete space towards this objective by performing a sequence of local edits, where deep generative neural networks propose the editing content and thus can control the quality of editing. We evaluate SAGS on paraphrase generation and molecule generation for sequence optimization and graph optimization, respectively. Extensive results show that our approach achieves state-of-the-art performance compared with existing paraphrase generation methods in terms of both automatic and human evaluations. Further, SAGS also significantly outperforms all the previous methods in molecule generation.

Two main routes of learning methods exist at present including neuroscience-inspired methods and machine learning methods. Both have own advantages, but neither currently can solve all learning problems well. Integrating them into one network may provide better learning abilities for general tasks. On the other hand, spiking neural network embodies "computation" in spatiotemporal domain with unique features of rich coding scheme and threshold switching, which is very suitable for low power and high parallel neuromorphic computing. Here, we report a spike-based general learning model that integrates two learning routes by introducing a brain-inspired meta-local module and a two-phase parametric modelling. The hybrid model can meta-learn general local plasticity, and receive top-down supervision information for multi-scale learning. We demonstrate that this hybrid model facilitates learning of many general tasks, including fault-tolerance learning, few-shot learning and multiple-task learning. Furthermore, the implementation of the hybrid model on the Tianjic neuromorphic platform proves that it can fully utilize the advantages of neuromorphic hardware architecture and promote energy-efficient on-chip applications.

In early years, text classification is typically accomplished by feature-based machine learning models; recently, deep neural networks, as a powerful learning machine, make it possible to work with raw input as the text stands. However, exiting end-to-end neural networks lack explicit interpretation of the prediction. In this paper, we propose a novel framework, JUMPER, inspired by the cognitive process of text reading, that models text classification as a sequential decision process. Basically, JUMPER is a neural system that scans a piece of text sequentially and makes classification decisions at the time it wishes. Both the classification result and when to make the classification are part of the decision process, which is controlled by a policy network and trained with reinforcement learning. Experimental results show that a properly trained JUMPER has the following properties: (1) It can make decisions whenever the evidence is enough, therefore reducing total text reading by 30-40% and often finding the key rationale of prediction. (2) It achieves classification accuracy better than or comparable to state-of-the-art models in several benchmark and industrial datasets.

Attentional Neural Network is a new framework that integrates top-down cognitive bias and bottom-up feature extraction in one coherent architecture. The top-down influence is especially effective when dealing with high noise or difficult segmentation problems. Our system is modular and extensible. It is also easy to train and cheap to run, and yet can accommodate complex behaviors. We obtain classification accuracy better than or competitive with state of art results on the MNIST variation dataset, and successfully disentangle overlaid digits with high success rates. We view such a general purpose framework as an essential foundation for a larger system emulating the cognitive abilities of the whole brain.

Recent experimental data indicate that the strengthening or weakening of synaptic connections between neurons depends on the relative timing of pre-and postsynaptic action potentials. A Hebbian synaptic modification rule based on these data leads to a stable state in which the excitatory and inhibitory inputs to a neuron are balanced, producing an irregular pattern of firing. It has been proposed that neurons in vivo operate in such a mode.