Parameter-efficient fine-tuning (PEFT) methods optimize large language models (LLMs) by modifying or introducing a small number of parameters to enhance alignment with downstream tasks. However, they can result in catastrophic forgetting, where LLMs prioritize new knowledge at the expense of comprehensive world knowledge. A promising approach to mitigate this issue is to recall prior memories based on the original knowledge. To this end, we propose a model-agnostic PEFT framework, IMSM, which Interweaves Memories of a Siamese Large Language Model. Specifically, our siamese LLM is equipped with an existing PEFT method. Given an incoming query, it generates two distinct memories based on the pre-trained and fine-tuned parameters. IMSM then incorporates an interweaving mechanism that regulates the contributions of both original and enhanced memories when generating the next token. This framework is theoretically applicable to all open-source LLMs and existing PEFT methods. We conduct extensive experiments across various benchmark datasets, evaluating the performance of popular open-source LLMs using the proposed IMSM, in comparison to both classical and leading PEFT methods. Our findings indicate that IMSM maintains comparable time and space efficiency to backbone PEFT methods while significantly improving performance and effectively mitigating catastrophic forgetting.

The scientific impact of academic papers is influenced by intricate factors such as dynamic popularity and inherent contribution. Existing models typically rely on static graphs for citation count estimation, failing to differentiate among its sources. In contrast, we propose distinguishing effects derived from various factors and predicting citation increments as estimated potential impacts within the dynamic context. In this research, we introduce a novel model, DPPDCC, which Disentangles the Potential impacts of Papers into Diffusion, Conformity, and Contribution values. It encodes temporal and structural features within dynamic heterogeneous graphs derived from the citation networks and applies various auxiliary tasks for disentanglement. By emphasizing comparative and co-cited/citing information and aggregating snapshots evolutionarily, DPPDCC captures knowledge flow within the citation network. Afterwards, popularity is outlined by contrasting augmented graphs to extract the essence of citation diffusion and predicting citation accumulation bins for quantitative conformity modeling. Orthogonal constraints ensure distinct modeling of each perspective, preserving the contribution value. To gauge generalization across publication times and replicate the realistic dynamic context, we partition data based on specific time points and retain all samples without strict filtering. Extensive experiments on three datasets validate DPPDCC's superiority over baselines for papers published previously, freshly, and immediately, with further analyses confirming its robustness. Our codes and supplementary materials can be found at https://github.com/ECNU-Text-Computing/DPPDCC.

The potential impact of an academic paper is determined by various factors, including its popularity and contribution. Existing models usually estimate original citation counts based on static graphs and fail to differentiate values from nuanced perspectives. In this study, we propose a novel graph neural network to Disentangle the Potential impacts of Papers into Diffusion, Conformity, and Contribution values (called DPPDCC). Given a target paper, DPPDCC encodes temporal and structural features within the constructed dynamic heterogeneous graph. Particularly, to capture the knowledge flow, we emphasize the importance of comparative and co-cited/citing information between papers and aggregate snapshots evolutionarily. To unravel popularity, we contrast augmented graphs to extract the essence of diffusion and predict the accumulated citation binning to model conformity. We further apply orthogonal constraints to encourage distinct modeling of each perspective and preserve the inherent value of contribution. To evaluate models' generalization for papers published at various times, we reformulate the problem by partitioning data based on specific time points to mirror real-world conditions. Extensive experimental results on three datasets demonstrate that DPPDCC significantly outperforms baselines for previously, freshly, and immediately published papers. Further analyses confirm its robust capabilities. We will make our datasets and codes publicly available.

The potential impact of a paper is often quantified by how many citations it will receive. However, most commonly used models may underestimate the influence of newly published papers over time, and fail to encapsulate this dynamics of citation network into the graph. In this study, we construct hierarchical and heterogeneous graphs for target papers with an annual perspective. The constructed graphs can record the annual dynamics of target papers' scientific context information. Then, a novel graph neural network, Hierarchical and Heterogeneous Contrastive Graph Learning Model (H2CGL), is proposed to incorporate heterogeneity and dynamics of the citation network. H2CGL separately aggregates the heterogeneous information for each year and prioritizes the highly-cited papers and relationships among references, citations, and the target paper. It then employs a weighted GIN to capture dynamics between heterogeneous subgraphs over years. Moreover, it leverages contrastive learning to make the graph representations more sensitive to potential citations. In particular, co-cited or co-citing papers of the target paper with large citation gaps are taken as hard negative samples, while randomly dropping low-cited papers could generate positive samples. Extensive experimental results on two scholarly datasets demonstrate that the proposed H2 CGL significantly outperforms a series of baseline approaches for both previously and freshly published papers. Additional analyses highlight the significance of the proposed modules. Our codes and settings have been released on Github. (https://github.com/ECNU-Text-Computing/H2CGL)

Automatically assessing academic papers has enormous potential to reduce peer-review burden and individual bias. Existing studies strive for building sophisticated deep neural networks to identify academic value based on comprehensive data, e.g., academic graphs and full papers. However, these data are not always easy to access. And the content of the paper rather than other features outside the paper should matter in a fair assessment. Furthermore, while BERT models can maintain general semantics by pre-training on large-scale corpora, they tend to be over-smoothing due to stacked self-attention layers among unfiltered input tokens. Therefore, it is nontrivial to figure out distinguishable value of an academic paper from its limited content. In this study, we propose a novel deep neural network, namely Dual-view Graph Convolutions Enhanced BERT (DGC-BERT), for academic paper acceptance estimation. We combine the title and abstract of the paper as input. Then, a pre-trained BERT model is employed to extract the paper's general representations. Apart from hidden representations of the final layer, we highlight the first and last few layers as lexical and semantic views. In particular, we re-examine the dual-view filtered self-attention matrices via constructing two graphs, respectively. After that, two multi-hop Graph Convolutional Networks (GCNs) are separately employed to capture pivotal and distant dependencies between the tokens. Moreover, the dual-view representations are facilitated by each other with biaffine attention modules. And a re-weighting gate is proposed to further streamline the dual-view representations with the help of the original BERT representation. Finally, whether the submitted paper could be acceptable is predicted based on the original language model features cooperated with the dual-view dependencies. Extensive data analyses and the full paper based MHCNN studies provide insights into the task and structural functions. Comparison experiments on two benchmark datasets demonstrate that the proposed DGC-BERT significantly outperforms alternative approaches, especially the state-of-the-art models like MHCNN and BERT variants. Additional analyses reveal significance and explainability of the proposed modules in the DGC-BERT. Our codes and settings have been released on Github (https://github.com/ECNU-Text-Computing/DGC-BERT).