Min Zhang

Hong Chen, Chengtao Lv, Liang Ding et al.

Large language models (LLMs) have significantly advanced the field of natural language processing, while the expensive memory and computation consumption impede their practical deployment. Quantization emerges as one of the most effective methods for improving the computational efficiency of LLMs. However, existing ultra-low-bit quantization always causes severe accuracy drops. In this paper, we empirically relieve the micro and macro characteristics of ultra-low bit quantization and present a novel Dual-Binarization method for LLMs, namely DB-LLM. For the micro-level, we take both the accuracy advantage of 2-bit-width and the efficiency advantage of binarization into account, introducing Flexible Dual Binarization (FDB). By splitting 2-bit quantized weights into two independent sets of binaries, FDB ensures the accuracy of representations and introduces flexibility, utilizing the efficient bitwise operations of binarization while retaining the inherent high sparsity of ultra-low bit quantization. For the macro-level, we find the distortion that exists in the prediction of LLM after quantization, which is specified as the deviations related to the ambiguity of samples. We propose the Deviation-Aware Distillation (DAD) method, enabling the model to focus differently on various samples. Comprehensive experiments show that our DB-LLM not only significantly surpasses the current State-of-The-Art (SoTA) in ultra-low bit quantization (eg, perplexity decreased from 9.64 to 7.23), but also achieves an additional 20\% reduction in computational consumption compared to the SOTA method under the same bit-width. Our code will be released soon.

Junchao Wu, Runzhe Zhan, Derek F. Wong et al.

The efficacy of detectors for texts generated by large language models (LLMs) substantially depends on the availability of large-scale training data. However, white-box zero-shot detectors, which require no such data, are limited by the accessibility of the source model of the LLM-generated text. In this paper, we propose a simple yet effective black-box zero-shot detection approach based on the observation that, from the perspective of LLMs, human-written texts typically contain more grammatical errors than LLM-generated texts. This approach involves calculating the Grammar Error Correction Score (GECScore) for the given text to differentiate between human-written and LLM-generated text. Experimental results show that our method outperforms current state-of-the-art (SOTA) zero-shot and supervised methods, achieving an average AUROC of 98.62% across XSum and Writing Prompts dataset. Additionally, our approach demonstrates strong reliability in the wild, exhibiting robust generalization and resistance to paraphrasing attacks. Data and code are available at: https://github.com/NLP2CT/GECScore.

Yan Gao, Yazheng Yang, Zhibin Lan et al.

Code-switching (CS) speech translation (ST) refers to translating speech that alternates between two or more languages into a target language text, which poses significant challenges due to the complexity of semantic modeling and the scarcity of CS data. Previous studies tend to rely on the model itself to implicitly learn semantic modeling during training, and resort to inefficient and costly manual annotations for these two challenges. To mitigate these limitations, we propose enhancing Large Language Models (LLMs) with a Mixture of Experts (MoE) speech projector, where each expert specializes in the semantic subspace of a specific language, enabling fine-grained modeling of speech features. Additionally, we introduce a multi-stage training paradigm that utilizes readily available monolingual automatic speech recognition (ASR) and monolingual ST data, facilitating speech-text alignment and improving translation capabilities. During training, we leverage a combination of language-specific loss and intra-group load balancing loss to guide the MoE speech projector in efficiently allocating tokens to the appropriate experts, across expert groups and within each group, respectively. To bridge the data gap across different training stages and improve adaptation to the CS scenario, we further employ a transition loss, enabling smooth transitions of data between stages, to effectively address the scarcity of high-quality CS speech translation data. Extensive experiments on widely used datasets demonstrate the effectiveness and generality of our approach.

Xiaobo Liang, Haoke Zhang, Helan hu et al.

The rapid advancement of large language models has given rise to a plethora of applications across a myriad of real-world tasks, mainly centered on aligning with human intent. However, the complexities inherent in human intent necessitate a dependence on labor-intensive and time-consuming human evaluation. To alleviate this constraint, we delve into the paradigm of employing open-source large language models as evaluators, aligning with the prevailing trend of utilizing GPT-4. Particularly, we present a step-by-step evaluation framework: \textbf{Fennec}, capable of \textbf{F}ine-grained \textbf{E}valuatio\textbf{N} and correctio\textbf{N} \textbf{E}xtended through bran\textbf{C}hing and bridging. Specifically, the branching operation dissects the evaluation task into various dimensions and granularities, thereby alleviating the challenges associated with evaluation. Concurrently, the bridging operation amalgamates diverse training datasets, augmenting the variety of evaluation tasks. In experimental trials, our 7B model consistently outperforms open-source larger-scale evaluation models across various widely adopted benchmarks in terms of both \textit{Agreement} and \textit{Consistency}, closely approaching the capabilities of GPT-4. We employ the fine-grained correction capabilities induced by the evaluation model to refine multiple model responses, and the results show that the refinement elevates the quality of responses, leading to an improvement of 1-2 points on the MT-Bench. Our code is available at Github\footnote{\url{https://github.com/dropreg/Fennec}}.

Zhuo Li, Guodong Du, Weiyang Guo et al.

Aligning large language models (LLMs) to simultaneously satisfy multiple objectives remains a significant challenge, especially given the diverse and often conflicting nature of human preferences. Existing alignment methods struggle to balance trade-offs effectively, often requiring costly retraining or yielding suboptimal results across the Pareto frontier of preferences. In this paper, we introduce \textit{HoE}(Hierarchical Mixture-of-Experts), a \textit{lightweight}, \textit{parameter-efficient}, and \textit{plug-and-play} approach that eliminates the need for model training, while enabling LLMs to adapt across the entire Pareto frontier and accommodate diverse user preferences. In particular, \textit{HoE} consists of three hierarchical components: LoRA Experts, Router Experts and Preference Routing, reaching optimal Pareto frontiers and achieving a trade-off between parameter size, training cost, and performance. We evaluate \textit{HoE} across various tasks on 14 objectives and 200 different preferences among 6 benchmarks, demonstrating superior performance over 15 recent baselines. Code is available in the supplementary materials.

Shou'ang Wei, Xinyun Wang, Shuzhen Bi et al.

The emergence of Large Language Models (LLMs) presents transformative opportunities for education, generating numerous novel application scenarios. However, significant challenges remain: evaluation metrics vary substantially across different educational scenarios, while many emerging scenarios lack appropriate assessment metrics. Current benchmarks predominantly measure general intelligence rather than pedagogical capabilities. To address this gap, we introduce ELMES, an open-source automated evaluation framework specifically designed for assessing LLMs in educational settings. ELMES features a modular architecture that enables researchers to create dynamic, multi-agent dialogues through simple configuration files, facilitating flexible scenario design without requiring extensive programming expertise. The framework incorporates a hybrid evaluation engine that objectively quantifies traditionally subjective pedagogical metrics using an LLM-as-a-Judge methodology. We conduct systematic benchmarking of state-of-the-art LLMs across four critical educational scenarios: Knowledge Point Explanation, Guided Problem-Solving Teaching, Interdisciplinary Lesson Plan Generation, and Contextualized Question Generation, employing fine-grained metrics developed in collaboration with education specialists. Our results demonstrate distinct capability distributions among models, revealing context-specific strengths and limitations. ELMES provides educators and researchers with an accessible evaluation framework that significantly reduces adaptation barriers for diverse educational applications while advancing the practical implementation of LLMs in pedagogy. The framework is publicly available at \emph{https://github.com/sii-research/elmes.git}.

Yifan Lu, Yigeng Zhou, Jing Li et al.

Multi-hop question answering (MHQA) poses a significant challenge for large language models (LLMs) due to the extensive knowledge demands involved. Knowledge editing, which aims to precisely modify the LLMs to incorporate specific knowledge without negatively impacting other unrelated knowledge, offers a potential solution for addressing MHQA challenges with LLMs. However, current solutions struggle to effectively resolve issues of knowledge conflicts. Most parameter-preserving editing methods are hindered by inaccurate retrieval and overlook secondary editing issues, which can introduce noise into the reasoning process of LLMs. In this paper, we introduce KEDKG, a novel knowledge editing method that leverages a dynamic knowledge graph for MHQA, designed to ensure the reliability of answers. KEDKG involves two primary steps: dynamic knowledge graph construction and knowledge graph augmented generation. Initially, KEDKG autonomously constructs a dynamic knowledge graph to store revised information while resolving potential knowledge conflicts. Subsequently, it employs a fine-grained retrieval strategy coupled with an entity and relation detector to enhance the accuracy of graph retrieval for LLM generation. Experimental results on benchmarks show that KEDKG surpasses previous state-of-the-art models, delivering more accurate and reliable answers in environments with dynamic information.

Dongfang Li, Zetian Sun, Xinshuo Hu et al.

Large Language Models (LLMs) need to adapt to the continuous changes in data, tasks, and user preferences. Due to their massive size and the high costs associated with training, LLMs are not suitable for frequent retraining. However, updates are necessary to keep them in sync with rapidly evolving human knowledge. To address these challenges, this paper proposes the Compression Memory Training (CMT) method, an efficient and effective online adaptation framework for LLMs that features robust knowledge retention capabilities. Inspired by human memory mechanisms, CMT compresses and extracts information from new documents to be stored in a memory bank. When answering to queries related to these new documents, the model aggregates these document memories from the memory bank to better answer user questions. The parameters of the LLM itself do not change during training and inference, reducing the risk of catastrophic forgetting. To enhance the encoding, retrieval, and aggregation of memory, we further propose three new general and flexible techniques, including memory-aware objective, self-matching and top-aggregation. Extensive experiments conducted on three continual learning datasets (i.e., StreamingQA, SQuAD and ArchivalQA) demonstrate that the proposed method improves model adaptability and robustness across multiple base LLMs (e.g., +4.07 EM & +4.19 F1 in StreamingQA with Llama-2-7b).

Xinping Zhao, Jindi Yu, Zhenyu Liu et al.

As we all know, hallucinations prevail in Large Language Models (LLMs), where the generated content is coherent but factually incorrect, which inflicts a heavy blow on the widespread application of LLMs. Previous studies have shown that LLMs could confidently state non-existent facts rather than answering ``I don't know''. Therefore, it is necessary to resort to external knowledge to detect and correct the hallucinated content. Since manual detection and correction of factual errors is labor-intensive, developing an automatic end-to-end hallucination-checking approach is indeed a needful thing. To this end, we present Medico, a Multi-source evidence fusion enhanced hallucination detection and correction framework. It fuses diverse evidence from multiple sources, detects whether the generated content contains factual errors, provides the rationale behind the judgment, and iteratively revises the hallucinated content. Experimental results on evidence retrieval (0.964 HR@5, 0.908 MRR@5), hallucination detection (0.927-0.951 F1), and hallucination correction (0.973-0.979 approval rate) manifest the great potential of Medico. A video demo of Medico can be found at https://youtu.be/RtsO6CSesBI.

Zecheng Pan, Zhikang Chen, Ding Li et al.

Merging models fine-tuned for different tasks into a single unified model has become an increasingly important direction for building versatile, efficient multi-task systems. Existing approaches predominantly rely on parameter interpolation in weight space, which we show introduces significant distribution shift in the feature space and undermines task-specific knowledge. In this paper, we propose OTMF (Optimal Transport-based Masked Fusion), a novel model merging framework rooted in optimal transport theory to address the distribution shift that arises from naive parameter interpolation. Instead of directly aggregating features or weights, OTMF aligns the semantic geometry of task-specific models by discovering common masks applied to task vectors through optimal transport plans. These masks selectively extract transferable and task-agnostic components while preserving the unique structural identities of each task. To ensure scalability in real-world settings, OTMF further supports a continual fusion paradigm that incrementally integrates each new task vector without revisiting previous ones, maintaining a bounded memory footprint and enabling efficient fusion across a growing number of tasks. We conduct comprehensive experiments on multiple vision and language benchmarks, and results show that OTMF achieves state-of-the-art performance in terms of both accuracy and efficiency. These findings highlight the practical and theoretical value of our approach to model merging.

Yachao Li, Junhui Li, Jing Jiang et al.

Existing large language models (LLMs) for machine translation are typically fine-tuned on sentence-level translation instructions and achieve satisfactory performance at the sentence level. However, when applied to document-level translation, these models face a significant challenge, particularly when dealing with documents containing over 512 tokens. This challenge arises from the issue of sentence-level coverage, where subsequent sentences in the document remain untranslated. As a result, the document-level translation capability of LLMs fine-tuned on sentence-level translation instructions is significantly limited. We conjecture that the primary cause of LLMs' weak document-level translation performance is the absence of document-to-document mapping ability. To address the issue, we propose an approach that combines sentence-level and document-level translation instructions of varying lengths to fine-tune LLMs. Our proposed translation mixed-instructions enable LLMs (Llama-2~7B and 13B) to maintain consistent translation performance from the sentence level to documents containing as many as 2048 tokens. Extensive experimental results show that the proposed approach significantly enhances the document-level translation capabilities of LLMs on 10 language pairs, effectively mitigating the sentence-level coverage issue in document-level translation. Experimentation on discourse phenomena has demonstrated that our document-level translation approach significantly improves translation quality, both in terms of BLEU score and discourse coherence.

Spyros Angelopoulos, Shahin Kamali, Kimia Shadkami

Bin packing is a classic optimization problem with a wide range of applications, from load balancing to supply chain management. In this work, we study the online variant of the problem, in which a sequence of items of various sizes must be placed into a minimum number of bins of uniform capacity. The online algorithm is enhanced with a (potentially erroneous) prediction concerning the frequency of item sizes in the sequence. We design and analyze online algorithms with efficient tradeoffs between the consistency (i.e., the competitive ratio assuming no prediction error) and the robustness (i.e., the competitive ratio under adversarial error), and whose performance degrades near-optimally as a function of the prediction error. This is the first theoretical and experimental study of online bin packing under competitive analysis, in the realistic setting of learnable predictions. Previous work addressed only extreme cases with respect to the prediction error, and relied on overly powerful and error-free oracles.

Hongqiu Wu, Hai Zhao, Min Zhang

Code summarization (CS) is becoming a promising area in recent language understanding, which aims to generate sensible human language automatically for programming language in the format of source code, serving in the most convenience of programmer developing. It is well known that programming languages are highly structured. Thus previous works attempt to apply structure-based traversal (SBT) or non-sequential models like Tree-LSTM and graph neural network (GNN) to learn structural program semantics. However, it is surprising that incorporating SBT into advanced encoder like Transformer instead of LSTM has been shown no performance gain, which lets GNN become the only rest means modeling such necessary structural clue in source code. To release such inconvenience, we propose structure-induced Transformer, which encodes sequential code inputs with multi-view structural clues in terms of a newly-proposed structure-induced self-attention mechanism. Extensive experiments show that our proposed structure-induced Transformer helps achieve new state-of-the-art results on benchmarks.

Houquan Zhou, Yu Zhang, Zhenghua Li et al.

In the pre deep learning era, part-of-speech tags have been considered as indispensable ingredients for feature engineering in dependency parsing. But quite a few works focus on joint tagging and parsing models to avoid error propagation. In contrast, recent studies suggest that POS tagging becomes much less important or even useless for neural parsing, especially when using character-based word representations. Yet there are not enough investigations focusing on this issue, both empirically and linguistically. To answer this, we design and compare three typical multi-task learning framework, i.e., Share-Loose, Share-Tight, and Stack, for joint tagging and parsing based on the state-of-the-art biaffine parser. Considering that it is much cheaper to annotate POS tags than parse trees, we also investigate the utilization of large-scale heterogeneous POS tag data. We conduct experiments on both English and Chinese datasets, and the results clearly show that POS tagging (both homogeneous and heterogeneous) can still significantly improve parsing performance when using the Stack joint framework. We conduct detailed analysis and gain more insights from the linguistic aspect.

Baijun Ji, Zhirui Zhang, Xiangyu Duan et al.

Transfer learning between different language pairs has shown its effectiveness for Neural Machine Translation (NMT) in low-resource scenario. However, existing transfer methods involving a common target language are far from success in the extreme scenario of zero-shot translation, due to the language space mismatch problem between transferor (the parent model) and transferee (the child model) on the source side. To address this challenge, we propose an effective transfer learning approach based on cross-lingual pre-training. Our key idea is to make all source languages share the same feature space and thus enable a smooth transition for zero-shot translation. To this end, we introduce one monolingual pre-training method and two bilingual pre-training methods to obtain a universal encoder for different languages. Once the universal encoder is constructed, the parent model built on such encoder is trained with large-scale annotated data and then directly applied in zero-shot translation scenario. Experiments on two public datasets show that our approach significantly outperforms strong pivot-based baseline and various multilingual NMT approaches.

Wei Jiang, Zhenghua Li, Yu Zhang et al.

This paper describes a simple UCCA semantic graph parsing approach. The key idea is to convert a UCCA semantic graph into a constituent tree, in which extra labels are deliberately designed to mark remote edges and discontinuous nodes for future recovery. In this way, we can make use of existing syntactic parsing techniques. Based on the data statistics, we recover discontinuous nodes directly according to the output labels of the constituent parser and use a biaffine classification model to recover the more complex remote edges. The classification model and the constituent parser are simultaneously trained under the multi-task learning framework. We use the multilingual BERT as extra features in the open tracks. Our system ranks the first place in the six English/German closed/open tracks among seven participating systems. For the seventh cross-lingual track, where there is little training data for French, we propose a language embedding approach to utilize English and German training data, and our result ranks the second place.