Conghui Zhu

Zhen Li, Bing Xu, Conghui Zhu et al.

Compared with unimodal data, multimodal data can provide more features to help the model analyze the sentiment of data. Previous research works rarely consider token-level feature fusion, and few works explore learning the common features related to sentiment in multimodal data to help the model fuse multimodal features. In this paper, we propose a Contrastive Learning and Multi-Layer Fusion (CLMLF) method for multimodal sentiment detection. Specifically, we first encode text and image to obtain hidden representations, and then use a multi-layer fusion module to align and fuse the token-level features of text and image. In addition to the sentiment analysis task, we also designed two contrastive learning tasks, label based contrastive learning and data based contrastive learning tasks, which will help the model learn common features related to sentiment in multimodal data. Extensive experiments conducted on three publicly available multimodal datasets demonstrate the effectiveness of our approach for multimodal sentiment detection compared with existing methods. The codes are available for use at https://github.com/Link-Li/CLMLF

Hongli Zhou, Hui Huang, Yunfei Long et al.

Recently, there has been a trend of evaluating the Large Language Model (LLM) quality in the flavor of LLM-as-a-Judge, namely leveraging another LLM to evaluate the current output quality. However, existing judges are proven to be biased, namely they would favor answers which present better superficial quality (such as verbosity, fluency) while ignoring the instruction following ability. In this work, we propose systematic research about the bias of LLM-as-a-Judge. Specifically, for closed-source judge models, we apply calibration to mitigate the significance of superficial quality, both on probability level and prompt level. For open-source judge models, we propose to mitigate the bias by contrastive training, with curated negative samples that deviate from instruction but present better superficial quality. We apply our methods on the bias evaluation benchmark, and experiment results show our methods mitigate the bias by a large margin while maintaining a satisfactory evaluation accuracy.

Hui Huang, Jiaheng Liu, Yancheng He et al.

Complex instruction-following with elaborate constraints is imperative for Large Language Models (LLMs). While existing methods have constructed data for complex instruction alignment, they all rely on a more advanced model, especially GPT-4, limiting their application. In this paper, we propose a Multi-granularity Self-Contrastive Training (MuSC) framework, to improve the complex instruction alignment without relying on a stronger model. Our method is conducted on both coarse and fine granularity. On coarse-granularity, we construct constraint-aware preference data based on instruction decomposition and recombination. On fine-granularity, we perform token-aware preference optimization with dynamic token-level supervision. Our method is evaluated on open-sourced models, and experiment results show our method achieves significant improvement on both complex and general instruction-following benchmarks, surpassing previous self-alignment methods.

Xiangyu Lu, Wang Xu, Haoyu Wang et al.

Real-time speech conversation is essential for natural and efficient human-machine interactions, requiring duplex and streaming capabilities. Traditional Transformer-based conversational chatbots operate in a turn-based manner and exhibit quadratic computational complexity that grows as the input size increases. In this paper, we propose DuplexMamba, a Mamba-based end-to-end multimodal duplex model for speech-to-text conversation. DuplexMamba enables simultaneous input processing and output generation, dynamically adjusting to support real-time streaming. Specifically, we develop a Mamba-based speech encoder and adapt it with a Mamba-based language model. Furthermore, we introduce a novel duplex decoding strategy that enables DuplexMamba to process input and generate output simultaneously. Experimental results demonstrate that DuplexMamba successfully implements duplex and streaming capabilities while achieving performance comparable to several recently developed Transformer-based models in automatic speech recognition (ASR) tasks and voice assistant benchmark evaluations. Our code and model are released.

Yudi Zhang, Weilin Zhao, Xu Han et al. · tsinghua

Speculative decoding and quantization effectively accelerate memory-bound inference of large language models. Speculative decoding mitigates the memory bandwidth bottleneck by verifying multiple tokens within a single forward pass, which increases computational effort. Quantization achieves this optimization by compressing weights and activations into lower bit-widths and also reduces computations via low-bit matrix multiplications. To further leverage their strengths, we investigate the integration of these two techniques. Surprisingly, experiments applying the advanced speculative decoding method EAGLE-2 to various quantized models reveal that the memory benefits from 4-bit weight quantization are diminished by the computational load from speculative decoding. Specifically, verifying a tree-style draft incurs significantly more time overhead than a single-token forward pass on 4-bit weight quantized models. This finding led to our new speculative decoding design: a hierarchical framework that employs a small model as an intermediate stage to turn tree-style drafts into sequence drafts, leveraging the memory access benefits of the target quantized model. Experimental results show that our hierarchical approach achieves a 2.78$\times$ speedup across various tasks for the 4-bit weight Llama-3-70B model on an A100 GPU, outperforming EAGLE-2 by 1.31$\times$. Code available at https://github.com/AI9Stars/SpecMQuant.

Yibin Huang, Wang Xu, Wanyue Zhang et al.

Spatial intelligence is a critical frontier for Multimodal Large Language Models (MLLMs), empowering them to comprehend the physical world. Drawing inspiration from human perception mechanisms, existing studies attempt to construct a coherent spatial understanding via grid-based cognitive maps from multi-frame visual inputs. However, current grid-based map methods rely on discretized raster representations, which limit the model's ability in fine-grained spatial reasoning. To overcome this limitation, we propose Video2Layout, a framework for reconstructing metric-grounded spatial layouts from video. The framework employs continuous object boundary coordinates to quantify inter-object physical distances and object size. This empowers the model with quantitative spatial computation capabilities, effectively alleviating the inherent ambiguity when describing spatial relationships in natural language. Specifically, our method comprises two core stages. First, in supervised fine-tuning stage, we construct a high-quality dataset from the AI2THOR simulator, which enables the model to learn the mapping from visual inputs to precise boundary coordinates. Subsequently, a reinforcement fine-tuning stage further enhances the model's real-world generalization capabilities. To systematically evaluate the correlation between cognitive map accuracy and image quantity, as well as how the quantity of image inputs affects spatial reasoning accuracy, we introduce QVS-Bench, a diagnostic benchmark designed to analyze the relevant mechanisms. Evaluated on QVS-Bench and mainstream spatial reasoning benchmarks, our model, V2LO-7B achieves an average improvement of 4.92% over the model trained on grid maps, validating the superiority of our method. Our code is available at https://github.com/ybrrraway/Video2Layout.

Hongyun Zhou, Xiangyu Lu, Wang Xu et al.

Low-Rank Adaptation (LoRA) is currently the most commonly used Parameter-efficient fine-tuning (PEFT) method, it introduces auxiliary parameters for each layer to fine-tune the pre-trained model under limited computing resources. However, it still faces resource consumption challenges during training when scaling up to larger models. Most previous studies have tackled this issue by using pruning techniques, which involve removing LoRA parameters deemed unimportant. Nonetheless, these efforts only analyze LoRA parameter features to evaluate their importance, such as parameter count, size, and gradient. In fact, the output of LoRA (product of LoRA parameter and hidden state), directly impacts the final results. Preliminary experiments indicate that a fraction of LoRA elements possesses significantly high output values, substantially influencing the layer output. Motivated by the observation, we propose LoRA-drop. Concretely, LoRA-drop evaluates the importance of LoRA based on the LoRA output. Then we retain LoRA for important layers and the other layers share the same LoRA. We conduct abundant experiments with models of different scales on NLU and NLG tasks. Results demonstrate that LoRA-drop can achieve performance comparable to full fine-tuning and LoRA, while retaining 50\% of the LoRA parameters on average.

Hongli Zhou, Hui Huang, Ziqing Zhao et al.

The evaluation of large language models (LLMs) via benchmarks is widespread, yet inconsistencies between different leaderboards and poor separability among top models raise concerns about their ability to accurately reflect authentic model capabilities. This paper provides a critical analysis of benchmark effectiveness, examining mainstream prominent LLM benchmarks using results from diverse models. We first propose Pseudo-Siamese Network for Item Response Theory (PSN-IRT), an enhanced Item Response Theory framework that incorporates a rich set of item parameters within an IRT-grounded architecture. PSN-IRT can be utilized for accurate and reliable estimations of item characteristics and model abilities. Based on PSN-IRT, we conduct extensive analysis on 11 LLM benchmarks comprising 41,871 items, revealing significant and varied shortcomings in their measurement quality. Furthermore, we demonstrate that leveraging PSN-IRT is able to construct smaller benchmarks while maintaining stronger alignment with human preference.

Hui Huang, Yancheng He, Wei Liu et al.

The widespread adoption of reinforcement learning-based alignment highlights the growing importance of reward models. Various benchmarks have been built to evaluate reward models in various domains and scenarios. However, a significant gap remains in assessing reward models for long-form generation, despite its critical role in real-world applications. To bridge this, we introduce Long-form RewardBench, the first reward modeling testbed specifically designed for long-form generation. Our benchmark encompasses five key subtasks: QA, RAG, Chat, Writing, and Reasoning. We collected instruction and preference data through a meticulously designed multi-stage data collection process, and conducted extensive experiments on 20+ mainstream reward models, including both classifiers and generative models. Our findings reveal that current models still lack long-form reward modeling capabilities. Furthermore, we designed a novel Long-form Needle-in-a-Haystack Test, which revealed a correlation between reward modeling performance and the error's position within a response, as well as the overall response length, with distinct characteristics observed between classification and generative models. Finally, we demonstrate that classifiers exhibit better generalizability compared to generative models trained on the same data. As the first benchmark for long-form reward modeling, this work aims to offer a robust platform for visualizing progress in this crucial area.

Hongli Zhou, Hui Huang, Rui Zhang et al.

Large language model (LLM)-based judges are widely adopted for automated evaluation and reward modeling, yet their judgments are often affected by judgment biases. Accurately evaluating these biases is essential for ensuring the reliability of LLM-based judges. However, existing studies typically investigate limited biases under a single judge formulation, either generative or discriminative, lacking a comprehensive evaluation. To bridge this gap, we propose JudgeBiasBench, a benchmark for systematically quantifying biases in LLM-based judges. JudgeBiasBench defines a taxonomy of judgment biases across 4 dimensions, and constructs bias-augmented evaluation instances through a controlled bias injection pipeline, covering 12 representative bias types. We conduct extensive experiments across both generative and discriminative judges, revealing that current judges exhibit significant and diverse bias patterns that often compromise the reliability of automated evaluation. To mitigate judgment bias, we propose bias-aware training that explicitly incorporates bias-related attributes into the training process, encouraging judges to disentangle task-relevant quality from bias-correlated cues. By adopting reinforcement learning for generative judges and contrastive learning for discriminative judges, our methods effectively reduce judgment biases while largely preserving general evaluation capability.

Wenxin Zhu, Andong Chen, Yuchen Song et al.

With the remarkable success of Multimodal Large Language Models (MLLMs) in perception tasks, enhancing their complex reasoning capabilities has emerged as a critical research focus. Existing models still suffer from challenges such as opaque reasoning paths and insufficient generalization ability. Chain-of-Thought (CoT) reasoning, which has demonstrated significant efficacy in language models by enhancing reasoning transparency and output interpretability, holds promise for improving model reasoning capabilities when extended to the multimodal domain. This paper provides a systematic review centered on "Multimodal Chain-of-Thought" (MCoT). First, it analyzes the background and theoretical motivations for its inception from the perspectives of technical evolution and task demands. Then, it introduces mainstream MCoT methods from three aspects: CoT paradigms, the post-training stage, and the inference stage, while also analyzing their underlying mechanisms. Furthermore, the paper summarizes existing evaluation benchmarks and metrics, and discusses the application scenarios of MCoT. Finally, it analyzes the challenges currently facing MCoT and provides an outlook on its future research directions.

Jitao Xu, Hongyun Zhou, Lei Shen et al.

Large Language Models (LLMs) can benefit from useful experiences to improve their performance on specific tasks. However, finding helpful experiences for different LLMs is not obvious, since it is unclear what experiences suit specific LLMs. Previous studies intended to automatically find useful experiences using LLMs, while it is difficult to ensure the effectiveness of the obtained experience. In this paper, we propose Stochastic Experience Optimization (SEO), an iterative approach that finds optimized model-specific experience without modifying model parameters through experience update in natural language. In SEO, we propose a stochastic validation method to ensure the update direction of experience, avoiding unavailing updates. Experimental results on three tasks for three LLMs demonstrate that experiences optimized by SEO can achieve consistently improved performance. Further analysis indicates that SEO-optimized experience can generalize to out-of-distribution data, boosting the performance of LLMs on similar tasks.

Changzhen Ji, Yating Zhang, Xiaozhong Liu et al.

Conversation generation as a challenging task in Natural Language Generation (NLG) has been increasingly attracting attention over the last years. A number of recent works adopted sequence-to-sequence structures along with external knowledge, which successfully enhanced the quality of generated conversations. Nevertheless, few works utilized the knowledge extracted from similar conversations for utterance generation. Taking conversations in customer service and court debate domains as examples, it is evident that essential entities/phrases, as well as their associated logic and inter-relationships can be extracted and borrowed from similar conversation instances. Such information could provide useful signals for improving conversation generation. In this paper, we propose a novel reading and memory framework called Deep Reading Memory Network (DRMN) which is capable of remembering useful information of similar conversations for improving utterance generation. We apply our model to two large-scale conversation datasets of justice and e-commerce fields. Experiments prove that the proposed model outperforms the state-of-the-art approaches.

Changzhen Ji, Xin Zhou, Conghui Zhu et al.

The multi-role judicial debate composed of the plaintiff, defendant, and judge is an important part of the judicial trial. Different from other types of dialogue, questions are raised by the judge, The plaintiff, plaintiff's agent defendant, and defendant's agent would be to debating so that the trial can proceed in an orderly manner. Question generation is an important task in Natural Language Generation. In the judicial trial, it can help the judge raise efficient questions so that the judge has a clearer understanding of the case. In this work, we propose an innovative end-to-end question generation model-Trial Brain Model (TBM) to build a Trial Brain, it can generate the questions the judge wants to ask through the historical dialogue between the plaintiff and the defendant. Unlike prior efforts in natural language generation, our model can learn the judge's questioning intention through predefined knowledge. We do experiments on real-world datasets, the experimental results show that our model can provide a more accurate question in the multi-role court debate scene.

Changzhen Ji, Xin Zhou, Yating Zhang et al.

In the past few years, audiences from different fields witness the achievements of sequence-to-sequence models (e.g., LSTM+attention, Pointer Generator Networks, and Transformer) to enhance dialogue content generation. While content fluency and accuracy often serve as the major indicators for model training, dialogue logics, carrying critical information for some particular domains, are often ignored. Take customer service and court debate dialogue as examples, compatible logics can be observed across different dialogue instances, and this information can provide vital evidence for utterance generation. In this paper, we propose a novel network architecture - Cross Copy Networks(CCN) to explore the current dialog context and similar dialogue instances' logical structure simultaneously. Experiments with two tasks, court debate and customer service content generation, proved that the proposed algorithm is superior to existing state-of-art content generation models.

Guanhua Zhang, Bing Bai, Jian Liang et al.

Recent studies show that crowd-sourced Natural Language Inference (NLI) datasets may suffer from significant biases like annotation artifacts. Models utilizing these superficial clues gain mirage advantages on the in-domain testing set, which makes the evaluation results over-estimated. The lack of trustworthy evaluation settings and benchmarks stalls the progress of NLI research. In this paper, we propose to assess a model's trustworthy generalization performance with cross-datasets evaluation. We present a new unified cross-datasets benchmark with 14 NLI datasets, and re-evaluate 9 widely-used neural network-based NLI models as well as 5 recently proposed debiasing methods for annotation artifacts. Our proposed evaluation scheme and experimental baselines could provide a basis to inspire future reliable NLI research.

Guanhua Zhang, Bing Bai, Junqi Zhang et al.

With the recent proliferation of the use of text classifications, researchers have found that there are certain unintended biases in text classification datasets. For example, texts containing some demographic identity-terms (e.g., "gay", "black") are more likely to be abusive in existing abusive language detection datasets. As a result, models trained with these datasets may consider sentences like "She makes me happy to be gay" as abusive simply because of the word "gay." In this paper, we formalize the unintended biases in text classification datasets as a kind of selection bias from the non-discrimination distribution to the discrimination distribution. Based on this formalization, we further propose a model-agnostic debiasing training framework by recovering the non-discrimination distribution using instance weighting, which does not require any extra resources or annotations apart from a pre-defined set of demographic identity-terms. Experiments demonstrate that our method can effectively alleviate the impacts of the unintended biases without significantly hurting models' generalization ability.

Conghui Zhu, Guanlin Li, Lemao Liu et al.

Despite the great success of NMT, there still remains a severe challenge: it is hard to interpret the internal dynamics during its training process. In this paper we propose to understand learning dynamics of NMT by using a recent proposed technique named Loss Change Allocation (LCA)~\citep{lan-2019-loss-change-allocation}. As LCA requires calculating the gradient on an entire dataset for each update, we instead present an approximate to put it into practice in NMT scenario. %motivated by the lesson from sgd. Our simulated experiment shows that such approximate calculation is efficient and is empirically proved to deliver consistent results to the brute-force implementation. In particular, extensive experiments on two standard translation benchmark datasets reveal some valuable findings.

Guanlin Li, Lemao Liu, Conghui Zhu et al.

Generalization to unseen instances is our eternal pursuit for all data-driven models. However, for realistic task like machine translation, the traditional approach measuring generalization in an average sense provides poor understanding for the fine-grained generalization ability. As a remedy, this paper attempts to identify and understand generalization barrier words within an unseen input sentence that \textit{cause} the degradation of fine-grained generalization. We propose a principled definition of generalization barrier words and a modified version which is tractable in computation. Based on the modified one, we propose three simple methods for barrier detection by the search-aware risk estimation through counterfactual generation. We then conduct extensive analyses on those detected generalization barrier words on both Zh$\Leftrightarrow$En NIST benchmarks from various perspectives. Potential usage of the detected barrier words is also discussed.

Chaoqun Duan, Kehai Chen, Rui Wang et al.

Existing neural machine translation (NMT) systems utilize sequence-to-sequence neural networks to generate target translation word by word, and then make the generated word at each time-step and the counterpart in the references as consistent as possible. However, the trained translation model tends to focus on ensuring the accuracy of the generated target word at the current time-step and does not consider its future cost which means the expected cost of generating the subsequent target translation (i.e., the next target word). To respond to this issue, we propose a simple and effective method to model the future cost of each target word for NMT systems. In detail, a time-dependent future cost is estimated based on the current generated target word and its contextual information to boost the training of the NMT model. Furthermore, the learned future context representation at the current time-step is used to help the generation of the next target word in the decoding. Experimental results on three widely-used translation datasets, including the WMT14 German-to-English, WMT14 English-to-French, and WMT17 Chinese-to-English, show that the proposed approach achieves significant improvements over strong Transformer-based NMT baseline.

Chaoqun Duan, Lei Cui, Shuming Ma et al.

Automatic live commenting aims to provide real-time comments on videos for viewers. It encourages users engagement on online video sites, and is also a good benchmark for video-to-text generation. Recent work on this task adopts encoder-decoder models to generate comments. However, these methods do not model the interaction between videos and comments explicitly, so they tend to generate popular comments that are often irrelevant to the videos. In this work, we aim to improve the relevance between live comments and videos by modeling the cross-modal interactions among different modalities. To this end, we propose a multimodal matching transformer to capture the relationships among comments, vision, and audio. The proposed model is based on the transformer framework and can iteratively learn the attention-aware representations for each modality. We evaluate the model on a publicly available live commenting dataset. Experiments show that the multimodal matching transformer model outperforms the state-of-the-art methods.

Guanhua Zhang, Bing Bai, Junqi Zhang et al.

Natural language inference (NLI) aims at predicting the relationship between a given pair of premise and hypothesis. However, several works have found that there widely exists a bias pattern called annotation artifacts in NLI datasets, making it possible to identify the label only by looking at the hypothesis. This irregularity makes the evaluation results over-estimated and affects models' generalization ability. In this paper, we consider a more trust-worthy setting, i.e., cross-dataset evaluation. We explore the impacts of annotation artifacts in cross-dataset testing. Furthermore, we propose a training framework to mitigate the impacts of the bias pattern. Experimental results demonstrate that our methods can alleviate the negative effect of the artifacts and improve the generalization ability of models.

Guanhua Zhang, Bing Bai, Jian Liang et al.

Natural Language Sentence Matching (NLSM) has gained substantial attention from both academics and the industry, and rich public datasets contribute a lot to this process. However, biased datasets can also hurt the generalization performance of trained models and give untrustworthy evaluation results. For many NLSM datasets, the providers select some pairs of sentences into the datasets, and this sampling procedure can easily bring unintended pattern, i.e., selection bias. One example is the QuoraQP dataset, where some content-independent naive features are unreasonably predictive. Such features are the reflection of the selection bias and termed as the leakage features. In this paper, we investigate the problem of selection bias on six NLSM datasets and find that four out of them are significantly biased. We further propose a training and evaluation framework to alleviate the bias. Experimental results on QuoraQP suggest that the proposed framework can improve the generalization ability of trained models, and give more trustworthy evaluation results for real-world adoptions.

Yiming Cui, Conghui Zhu, Xiaoning Zhu et al.

Pivot language is employed as a way to solve the data sparseness problem in machine translation, especially when the data for a particular language pair does not exist. The combination of source-to-pivot and pivot-to-target translation models can induce a new translation model through the pivot language. However, the errors in two models may compound as noise, and still, the combined model may suffer from a serious phrase sparsity problem. In this paper, we directly employ the word lexical model in IBM models as an additional resource to augment pivot phrase table. In addition, we also propose a phrase table pruning method which takes into account both of the source and target phrasal coverage. Experimental result shows that our pruning method significantly outperforms the conventional one, which only considers source side phrasal coverage. Furthermore, by including the entries in the lexicon model, the phrase coverage increased, and we achieved improved results in Chinese-to-Japanese translation using English as pivot language.