Authors: Sikun Guo, Amir Hassan Shariatmadari, Guangzhi Xiong, Albert Huang, Eric Xie, Stefan Bekiranov, Aidong Zhang
Abstract: Large Language Models (LLMs) have transformed how people interact with artificial intelligence (AI) systems, achieving state-of-the-art results in various tasks, including scientific discovery and hypothesis generation. However, the lack of a comprehensive and systematic evaluation framework for generating research ideas using LLMs poses a significant obstacle to understanding and assessing their generative capabilities in scientific discovery. To address this gap, we propose IdeaBench, a benchmark system that includes a comprehensive dataset and an evaluation framework for standardizing the assessment of research idea generation using LLMs. Our dataset comprises titles and abstracts from a diverse range of influential papers, along with their referenced works. To emulate the human process of generating research ideas, we profile LLMs as domain-specific researchers and ground them in the same context considered by human researchers. This maximizes the utilization of the LLMs' parametric knowledge to dynamically generate new research ideas. We also introduce an evaluation framework for assessing the quality of generated research ideas. Our evaluation framework is a two-stage process: first, using GPT-4o to rank ideas based on user-specified quality indicators such as novelty and feasibility, enabling scalable personalization; and second, calculating relative ranking based "Insight Score" to quantify the chosen quality indicator. The proposed benchmark system will be a valuable asset for the community to measure and compare different LLMs, ultimately advancing the automation of the scientific discovery process.
Authors: Lin Wang, Xiaocui Yang, Shi Feng, Daling Wang, Yifei Zhang
Abstract: Multimodal conversation, a crucial form of human communication, carries rich emotional content, making the exploration of the causes of emotions within it a research endeavor of significant importance. However, existing research on the causes of emotions typically uses clause selection methods to locate the reason utterance, without providing a detailed explanation of the emotional causes. In this paper, we propose a new task, \textbf{M}ultimodal \textbf{C}onversation \textbf{E}motion \textbf{C}ause \textbf{E}xplanation (MCECE), aiming to generate a detailed explanation of the emotional cause to the target utterance within a multimodal conversation scenario. Building upon the MELD dataset, we develop a new dataset (ECEM) that integrates video clips with detailed explanations of character emotions, facilitating an in-depth examination of the causal factors behind emotional expressions in multimodal conversations.A novel approach, FAME-Net, is further proposed, that harnesses the power of Large Language Models (LLMs) to analyze visual data and accurately interpret the emotions conveyed through facial expressions in videos. By exploiting the contagion effect of facial emotions, FAME-Net effectively captures the emotional causes of individuals engaged in conversations. Our experimental results on the newly constructed dataset show that FAME-Net significantly outperforms several excellent large language model baselines. Code and dataset are available at \url{https://github.com/3222345200/ECEMdataset.git}
Authors: Geoff Keeling, Winnie Street, Martyna Stachaczyk, Daria Zakharova, Iulia M. Comsa, Anastasiya Sakovych, Isabella Logothesis, Zejia Zhang, Blaise Ag\"uera y Arcas, Jonathan Birch
Abstract: Pleasure and pain play an important role in human decision making by providing a common currency for resolving motivational conflicts. While Large Language Models (LLMs) can generate detailed descriptions of pleasure and pain experiences, it is an open question whether LLMs can recreate the motivational force of pleasure and pain in choice scenarios - a question which may bear on debates about LLM sentience, understood as the capacity for valenced experiential states. We probed this question using a simple game in which the stated goal is to maximise points, but where either the points-maximising option is said to incur a pain penalty or a non-points-maximising option is said to incur a pleasure reward, providing incentives to deviate from points-maximising behaviour. Varying the intensity of the pain penalties and pleasure rewards, we found that Claude 3.5 Sonnet, Command R+, GPT-4o, and GPT-4o mini each demonstrated at least one trade-off in which the majority of responses switched from points-maximisation to pain-minimisation or pleasure-maximisation after a critical threshold of stipulated pain or pleasure intensity is reached. LLaMa 3.1-405b demonstrated some graded sensitivity to stipulated pleasure rewards and pain penalties. Gemini 1.5 Pro and PaLM 2 prioritised pain-avoidance over points-maximisation regardless of intensity, while tending to prioritise points over pleasure regardless of intensity. We discuss the implications of these findings for debates about the possibility of LLM sentience.
Authors: Jianyi Zhang, Da-Cheng Juan, Cyrus Rashtchian, Chun-Sung Ferng, Heinrich Jiang, Yiran Chen
Abstract: Large language models (LLMs) have demonstrated remarkable capabilities, but their outputs can sometimes be unreliable or factually incorrect. To address this, we introduce Self Logits Evolution Decoding (SLED), a novel decoding framework that enhances the truthfulness of LLMs without relying on external knowledge bases or requiring further fine-tuning. From an optimization perspective, our SLED framework leverages the latent knowledge embedded within the LLM by contrasting the output logits from the final layer with those from early layers. It then utilizes an approximate gradient approach to enable latent knowledge to guide the self-refinement of outputs, thereby effectively improving factual accuracy. Extensive experiments have been conducted on established benchmarks across a diverse range of model families (LLaMA 2, LLaMA 3, Gemma) and scales (from 2B to 70B), including more advanced architectural configurations such as the mixture of experts (MoE). Our evaluation spans a wide variety of tasks, including multi-choice, open-generation, and adaptations to chain-of-thought reasoning tasks. The results demonstrate that SLED consistently improves factual accuracy by up to 20\% compared to existing decoding methods while maintaining natural language fluency and negligible latency overhead. Furthermore, it can be flexibly combined with other decoding methods to further enhance their performance.
Authors: Xinyi Leng, Jason Liang, Jack Mauro, Xu Wang, Andrea L. Bertozzi, James Chapman, Junyuan Lin, Bohan Chen, Chenchen Ye, Temple Daniel, P. Jeffrey Brantingham
Abstract: Narrative data spans all disciplines and provides a coherent model of the world to the reader or viewer. Recent advancement in machine learning and Large Language Models (LLMs) have enable great strides in analyzing natural language. However, Large language models (LLMs) still struggle with complex narrative arcs as well as narratives containing conflicting information. Recent work indicates LLMs augmented with external knowledge bases can improve the accuracy and interpretability of the resulting models. In this work, we analyze the effectiveness of applying knowledge graphs (KGs) in understanding true-crime podcast data from both classical Natural Language Processing (NLP) and LLM approaches. We directly compare KG-augmented LLMs (KGLLMs) with classical methods for KG construction, topic modeling, and sentiment analysis. Additionally, the KGLLM allows us to query the knowledge base in natural language and test its ability to factually answer questions. We examine the robustness of the model to adversarial prompting in order to test the model's ability to deal with conflicting information. Finally, we apply classical methods to understand more subtle aspects of the text such as the use of hearsay and sentiment in narrative construction and propose future directions. Our results indicate that KGLLMs outperform LLMs on a variety of metrics, are more robust to adversarial prompts, and are more capable of summarizing the text into topics.
Authors: Yuxiang Guo, Lu Yin, Bo Jiang, Jiaqi Zhang
Abstract: Aligning large language models (LLMs) with human intent is critical for enhancing their performance across a variety of tasks. Standard alignment techniques, such as Direct Preference Optimization (DPO), often rely on the binary Bradley-Terry (BT) model, which can struggle to capture the complexities of human preferences -- particularly in the presence of noisy or inconsistent labels and frequent ties. To address these limitations, we introduce the Tie-rank Oriented Bradley-Terry model (TOBT), an extension of the BT model that explicitly incorporates ties, enabling more nuanced preference representation. Building on this, we propose Tie-rank Oriented Direct Preference Optimization (TODO), a novel alignment algorithm that leverages TOBT's ternary ranking system to improve preference alignment. In evaluations on Mistral-7B and Llama 3-8B models, TODO consistently outperforms DPO in modeling preferences across both in-distribution and out-of-distribution datasets. Additional assessments using MT Bench and benchmarks such as Piqa, ARC-c, and MMLU further demonstrate TODO's superior alignment performance. Notably, TODO also shows strong results in binary preference alignment, highlighting its versatility and potential for broader integration into LLM alignment. The implementation details can be found in https://github.com/XXares/TODO.
Authors: Aliyah R. Hsu, James Zhu, Zhichao Wang, Bin Bi, Shubham Mehrotra, Shiva K. Pentyala, Katherine Tan, Xiang-Bo Mao, Roshanak Omrani, Sougata Chaudhuri, Regunathan Radhakrishnan, Sitaram Asur, Claire Na Cheng, Bin Yu
Abstract: LLMs have demonstrated impressive proficiency in generating coherent and high-quality text, making them valuable across a range of text-generation tasks. However, rigorous evaluation of this generated content is crucial, as ensuring its quality remains a significant challenge due to persistent issues such as factual inaccuracies and hallucinations. This paper introduces two fine-tuned general-purpose LLM autoevaluators, REC-12B and REC-70B, specifically designed to evaluate generated text across several dimensions: faithfulness, instruction following, coherence, and completeness. These models not only provide ratings for these metrics but also offer detailed explanations and verifiable citations, thereby enhancing trust in the content. Moreover, the models support various citation modes, accommodating different requirements for latency and granularity. Extensive evaluations on diverse benchmarks demonstrate that our general-purpose LLM auto-evaluator, REC-70B, outperforms state-of-the-art LLMs, excelling in content evaluation by delivering better quality explanations and citations with minimal bias. It achieves Rank \#1 as a generative model on the RewardBench leaderboard\footnote{\url{https://huggingface.co/spaces/allenai/reward-bench}} under the model name \texttt{TextEval-Llama3.1-70B}. Our REC dataset and models are released at \url{https://github.com/adelaidehsu/REC}.
URLs: https://huggingface.co/spaces/allenai/reward-bench, https://github.com/adelaidehsu/REC
Authors: Rohan Sanghera, Arun James Thirunavukarasu, Marc El Khoury, Jessica O'Logbon, Yuqing Chen, Archie Watt, Mustafa Mahmood, Hamid Butt, George Nishimura, Andrew Soltan
Abstract: Large language models (LLMs) excel in tasks requiring processing and interpretation of input text. Abstract screening is a labour-intensive component of systematic review involving repetitive application of inclusion and exclusion criteria on a large volume of studies identified by a literature search. Here, LLMs (GPT-3.5 Turbo, GPT-4 Turbo, GPT-4o, Llama 3 70B, Gemini 1.5 Pro, and Claude Sonnet 3.5) were trialled on systematic reviews in a full issue of the Cochrane Library to evaluate their accuracy in zero-shot binary classification for abstract screening. Trials over a subset of 800 records identified optimal prompting strategies and demonstrated superior performance of LLMs to human researchers in terms of sensitivity (LLMmax = 1.000, humanmax = 0.775), precision (LLMmax = 0.927, humanmax = 0.911), and balanced accuracy (LLMmax = 0.904, humanmax = 0.865). The best performing LLM-prompt combinations were trialled across every replicated search result (n = 119,691), and exhibited consistent sensitivity (range 0.756-1.000) but diminished precision (range 0.004-0.096). 66 LLM-human and LLM-LLM ensembles exhibited perfect sensitivity with a maximal precision of 0.458, with less observed performance drop in larger trials. Significant variation in performance was observed between reviews, highlighting the importance of domain-specific validation before deployment. LLMs may reduce the human labour cost of systematic review with maintained or improved accuracy and sensitivity. Systematic review is the foundation of evidence-based medicine, and LLMs can contribute to increasing the efficiency and quality of this mode of research.
Authors: Yukun Li, Sijia Wang, Lifu Huang, Li-Ping Liu
Abstract: One important approach to improving the reliability of large language models (LLMs) is to provide accurate confidence estimations regarding the correctness of their answers. However, developing a well-calibrated confidence estimation model is challenging, as mistakes made by LLMs can be difficult to detect. We propose a novel method combining the LLM's self-consistency with labeled data and training an auxiliary model to estimate the correctness of its responses to questions. This auxiliary model predicts the correctness of responses based solely on their consistent information. To set up the learning problem, we use a weighted graph to represent the consistency among the LLM's multiple responses to a question. Correctness labels are assigned to these responses based on their similarity to the correct answer. We then train a graph neural network to estimate the probability of correct responses. Experiments demonstrate that the proposed approach substantially outperforms several of the most recent methods in confidence calibration across multiple widely adopted benchmark datasets. Furthermore, the proposed approach significantly improves the generalization capability of confidence calibration on out-of-domain (OOD) data.
Authors: Bo Yuan, Jiazi Hu
Abstract: Course evaluation is a critical component in higher education pedagogy. It not only serves to identify limitations in existing course designs and provide a basis for curricular innovation, but also to offer quantitative insights for university administrative decision-making. Traditional evaluation methods, primarily comprising student surveys, instructor self-assessments, and expert reviews, often encounter challenges, including inherent subjectivity, feedback delays, inefficiencies, and limitations in addressing innovative teaching approaches. Recent advancements in large language models (LLMs) within artificial intelligence (AI) present promising new avenues for enhancing course evaluation processes. This study explores the application of LLMs in automated course evaluation from multiple perspectives and conducts rigorous experiments across 100 courses at a major university in China. The findings indicate that: (1) LLMs can be an effective tool for course evaluation; (2) their effectiveness is contingent upon appropriate fine-tuning and prompt engineering; and (3) LLM-generated evaluation results demonstrate a notable level of rationality and interpretability.
Authors: Siyuan Chen, Qingyi Si, Chenxu Yang, Yunzhi Liang, Zheng Lin, Huan Liu, Weiping Wang
Abstract: The advent of large language models (LLMs) has significantly propelled the advancement of Role-Playing Agents (RPAs). However, current Role-Playing Agents predominantly focus on mimicking a character's fundamental attributes while neglecting the replication of linguistic style, and they are incapable of effectively replicating characters when performing tasks beyond multi-turn dialogues, which results in generated responses that lack authenticity. The reason current RPAs lack this capability is due to the nature of existing character datasets, which lack collections of character quotations and are limited to multi-turn dialogue tasks, constraining the RPA's performance across other task domains and failing to mimic a character's linguistic style. To address this gap, we developed a multi-task role-playing dataset named MRstyle, which encompasses a substantial number of real individuals along with their quotations and covers seven different tasks. On this basis, we develop StyleRPA, a Multi-Task Role-Playing Agent (MRPA) that significantly outperforms recent open-source LLMs and RPAs baselines on 7 tasks including Dialogue, Dictionary, Composition, Story Generation, Product Description, Music Commentary, and Open Question Answering. The code and data will be released.
Authors: Haneul Yoo, Cheonbok Park, Sangdoo Yun, Alice Oh, Hwaran Lee
Abstract: Large language models (LLMs) now exhibit near human-level performance in various tasks, but their performance drops drastically after a handful of high-resource languages due to the imbalance in pre-training data. Inspired by the human process of second language acquisition, particularly code-switching (the practice of language alternation in a conversation), we propose code-switching curriculum learning (CSCL) to enhance cross-lingual transfer for LLMs. CSCL mimics the stages of human language learning by progressively training models with a curriculum consisting of 1) token-level code-switching, 2) sentence-level code-switching, and 3) monolingual corpora. Using Qwen 2 as our underlying model, we demonstrate the efficacy of the CSCL in improving language transfer to Korean, achieving significant performance gains compared to monolingual continual pre-training methods. Ablation studies reveal that both token- and sentence-level code-switching significantly enhance cross-lingual transfer and that curriculum learning amplifies these effects. We also extend our findings into various languages, including Japanese (high-resource) and Indonesian (low-resource), and using two additional models (Gemma 2 and Phi 3.5). We further show that CSCL mitigates spurious correlations between language resources and safety alignment, presenting a robust, efficient framework for more equitable language transfer in LLMs. We observe that CSCL is effective for low-resource settings where high-quality, monolingual corpora for language transfer are hardly available.
Authors: Yuxin Xiao, Chaoqun Wan, Yonggang Zhang, Wenxiao Wang, Binbin Lin, Xiaofei He, Xu Shen, Jieping Ye
Abstract: As the development and application of Large Language Models (LLMs) continue to advance rapidly, enhancing their trustworthiness and aligning them with human preferences has become a critical area of research. Traditional methods rely heavily on extensive data for Reinforcement Learning from Human Feedback (RLHF), but representation engineering offers a new, training-free approach. This technique leverages semantic features to control the representation of LLM's intermediate hidden states, enabling the model to meet specific requirements such as increased honesty or heightened safety awareness. However, a significant challenge arises when attempting to fulfill multiple requirements simultaneously. It proves difficult to encode various semantic contents, like honesty and safety, into a singular semantic feature, restricting its practicality. In this work, we address this issue through ``Sparse Activation Control''. By delving into the intrinsic mechanisms of LLMs, we manage to identify and pinpoint components that are closely related to specific tasks within the model, i.e., attention heads. These heads display sparse characteristics that allow for near-independent control over different tasks. Our experiments, conducted on the open-source Llama series models, have yielded encouraging results. The models were able to align with human preferences on issues of safety, factuality, and bias concurrently.
Authors: Sorouralsadat Fatemi, Yuheng Hu, Maryam Mousavi
Abstract: Large Language Models (LLMs) have demonstrated impressive capabilities across diverse Natural Language Processing (NLP) tasks, including language understanding, reasoning, and generation. However, general-domain LLMs often struggle with financial tasks due to the technical and specialized nature of financial texts. This study investigates the efficacy of instruction fine-tuning smaller-scale LLMs, including Mistral-7B, Llama3-8B, and Phi3-mini, to enhance their performance in financial text classification tasks. We fine-tuned both instruction-tuned and base models across four financial classification tasks, achieving significant improvements in task-specific performance. Furthermore, we evaluated the zero-shot capabilities of these fine-tuned models on three unseen complex financial tasks, including argument classification, deal completeness classification, and causal classification. Our results indicate while base model fine-tuning led to greater degradation, instruction-tuned models maintained more robust performance. To address this degradation, we employed model merging techniques, integrating single-task domain-specific fine-tuned models with the base model. Using this merging method resulted in significant enhancements in zero-shot performance, even exceeding the original model's accuracy on certain datasets. Our findings underscore the effectiveness of instruction fine-tuning and model merging for adapting LLMs to specialized financial text classification tasks.
Authors: Guangxuan Xu, Kai Xu, Shivchander Sudalairaj, Hao Wang, Akash Srivastava
Abstract: Preference tuning of large language models (LLMs) relies on high-quality human preference data, which is often expensive and time-consuming to gather. While existing methods can use trained reward models or proprietary model as judges for preference annotation, they have notable drawbacks: training reward models remain dependent on initial human data, and using proprietary model imposes license restrictions that inhibits commercial usage. In this paper, we introduce customized density ratio (CDR) that leverages open-source LLMs for data annotation, offering an accessible and effective solution. Our approach uses the log-density ratio between a well-aligned LLM and a less aligned LLM as a reward signal. We explores 221 different LLMs pairs and empirically demonstrate that increasing the performance gap between paired LLMs correlates with better reward generalization. Furthermore, we show that tailoring the density ratio reward function with specific criteria and preference exemplars enhances performance across domains and within target areas. In our experiment using density ratio from a pair of Mistral-7B models, CDR achieves a RewardBench score of 82.6, outperforming the best in-class trained reward functions and demonstrating competitive performance against SoTA models in Safety (91.0) and Reasoning (88.0) domains. We use CDR to annotate an on-policy preference dataset with which we preference tune Llama-3-8B-Instruct with SimPO. The final model achieves a 37.4% (+15.1%) win rate on ArenaHard and a 40.7% (+17.8%) win rate on Length-Controlled AlpacaEval 2.0, along with a score of 8.0 on MT-Bench.
Authors: Balu Bhasuran, Qiao Jin, Yuzhang Xie, Carl Yang, Karim Hanna, Jennifer Costa, Cindy Shavor, Zhiyong Lu, Zhe He
Abstract: Differential diagnosis is crucial for medicine as it helps healthcare providers systematically distinguish between conditions that share similar symptoms. This study assesses the impact of lab test results on differential diagnoses (DDx) made by large language models (LLMs). Clinical vignettes from 50 case reports from PubMed Central were created incorporating patient demographics, symptoms, and lab results. Five LLMs GPT-4, GPT-3.5, Llama-2-70b, Claude-2, and Mixtral-8x7B were tested to generate Top 10, Top 5, and Top 1 DDx with and without lab data. A comprehensive evaluation involving GPT-4, a knowledge graph, and clinicians was conducted. GPT-4 performed best, achieving 55% accuracy for Top 1 diagnoses and 60% for Top 10 with lab data, with lenient accuracy up to 80%. Lab results significantly improved accuracy, with GPT-4 and Mixtral excelling, though exact match rates were low. Lab tests, including liver function, metabolic/toxicology panels, and serology/immune tests, were generally interpreted correctly by LLMs for differential diagnosis.
Authors: Lindia Tjuatja, Graham Neubig, Tal Linzen, Sophie Hao
Abstract: When comparing the linguistic capabilities of language models (LMs) with humans using LM probabilities, factors such as the length of the sequence and the unigram frequency of lexical items have a significant effect on LM probabilities in ways that humans are largely robust to. Prior works in comparing LM and human acceptability judgments treat these effects uniformly across models, making a strong assumption that models require the same degree of adjustment to control for length and unigram frequency effects. We propose MORCELA, a new linking theory between LM scores and acceptability judgments where the optimal level of adjustment for these effects is estimated from data via learned parameters for length and unigram frequency. We first show that MORCELA outperforms a commonly used linking theory for acceptability--SLOR (Pauls and Klein, 2012; Lau et al. 2017)--across two families of transformer LMs (Pythia and OPT). Furthermore, we demonstrate that the assumed degrees of adjustment in SLOR for length and unigram frequency overcorrect for these confounds, and that larger models require a lower relative degree of adjustment for unigram frequency, though a significant amount of adjustment is still necessary for all models. Finally, our subsequent analysis shows that larger LMs' lower susceptibility to frequency effects can be explained by an ability to better predict rarer words in context.
Authors: Mowafak Allaham, Kimon Kieslich, Nicholas Diakopoulos
Abstract: Expert-driven frameworks for impact assessments (IAs) may inadvertently overlook the effects of AI technologies on the public's social behavior, policy, and the cultural and geographical contexts shaping the perception of AI and the impacts around its use. This research explores the potentials of fine-tuning LLMs on negative impacts of AI reported in a diverse sample of articles from 266 news domains spanning 30 countries around the world to incorporate more diversity into IAs. Our findings highlight (1) the potential of fine-tuned open-source LLMs in supporting IA of AI technologies by generating high-quality negative impacts across four qualitative dimensions: coherence, structure, relevance, and plausibility, and (2) the efficacy of small open-source LLM (Mistral-7B) fine-tuned on impacts from news media in capturing a wider range of categories of impacts that GPT-4 had gaps in covering.
Authors: Sshubam Verma, Mohammed Safi Ur Rahman Khan, Vishwajeet Kumar, Rudra Murthy, Jaydeep Sen
Abstract: Evaluating Large Language Models (LLMs) in low-resource and linguistically diverse languages remains a significant challenge in NLP, particularly for languages using non-Latin scripts like those spoken in India. Existing benchmarks predominantly focus on English, leaving substantial gaps in assessing LLM capabilities in these languages. We introduce MILU, a Multi task Indic Language Understanding Benchmark, a comprehensive evaluation benchmark designed to address this gap. MILU spans 8 domains and 42 subjects across 11 Indic languages, reflecting both general and culturally specific knowledge. With an India-centric design, incorporates material from regional and state-level examinations, covering topics such as local history, arts, festivals, and laws, alongside standard subjects like science and mathematics. We evaluate over 42 LLMs, and find that current LLMs struggle with MILU, with GPT-4o achieving the highest average accuracy at 72 percent. Open multilingual models outperform language-specific fine-tuned models, which perform only slightly better than random baselines. Models also perform better in high resource languages as compared to low resource ones. Domain-wise analysis indicates that models perform poorly in culturally relevant areas like Arts and Humanities, Law and Governance compared to general fields like STEM. To the best of our knowledge, MILU is the first of its kind benchmark focused on Indic languages, serving as a crucial step towards comprehensive cultural evaluation. All code, benchmarks, and artifacts will be made publicly available to foster open research.
Authors: Khalid Alnajjar, Mika H\"am\"al\"ainen, Jack Rueter
Abstract: This paper presents a methodology for training a transformer-based model to classify lexical and morphosyntactic features of Skolt Sami, an endangered Uralic language characterized by complex morphology. The goal of our approach is to create an effective system for understanding and analyzing Skolt Sami, given the limited data availability and linguistic intricacies inherent to the language. Our end-to-end pipeline includes data extraction, augmentation, and training a transformer-based model capable of predicting inflection classes. The motivation behind this work is to support language preservation and revitalization efforts for minority languages like Skolt Sami. Accurate classification not only helps improve the state of Finite-State Transducers (FSTs) by providing greater lexical coverage but also contributes to systematic linguistic documentation for researchers working with newly discovered words from literature and native speakers. Our model achieves an average weighted F1 score of 1.00 for POS classification and 0.81 for inflection class classification. The trained model and code will be released publicly to facilitate future research in endangered NLP.
Authors: Sheng-Chieh Lin, Chankyu Lee, Mohammad Shoeybi, Jimmy Lin, Bryan Catanzaro, Wei Ping
Abstract: State-of-the-art retrieval models typically address a straightforward search scenario, where retrieval tasks are fixed (e.g., finding a passage to answer a specific question) and only a single modality is supported for both queries and retrieved results. This paper introduces techniques for advancing information retrieval with multimodal large language models (MLLMs), enabling a broader search scenario, termed universal multimodal retrieval, where multiple modalities and diverse retrieval tasks are accommodated. To this end, we first study fine-tuning an MLLM as a bi-encoder retriever on 10 datasets with 16 retrieval tasks. Our empirical results show that the fine-tuned MLLM retriever is capable of understanding challenging queries, composed of both text and image, but underperforms a smaller CLIP retriever in cross-modal retrieval tasks due to modality bias from MLLMs. To address the issue, we propose modality-aware hard negative mining to mitigate the modality bias exhibited by MLLM retrievers. Second, we propose to continually fine-tune the universal multimodal retriever to enhance its text retrieval capability while maintaining multimodal retrieval capability. As a result, our model, MM-Embed, achieves state-of-the-art performance on the multimodal retrieval benchmark M-BEIR, which spans multiple domains and tasks, while also surpassing the state-of-the-art text retrieval model, NV-Embed-v1, on MTEB retrieval benchmark. Finally, we explore to prompt the off-the-shelf MLLMs as the zero-shot rerankers to refine the ranking of the candidates from the multimodal retriever. We find that through prompt-and-reranking, MLLMs can further improve multimodal retrieval when the user queries (e.g., text-image composed queries) are more complex and challenging to understand. These findings also pave the way to advance universal multimodal retrieval in the future.
Authors: Zahra Ahani, Moein Shahiki Tash, Fazlourrahman Balouchzahi, Luis Ramos, Grigori Sidorov, Alexander Gelbukh
Abstract: Social support, conveyed through a multitude of interactions and platforms such as social media, plays a pivotal role in fostering a sense of belonging, aiding resilience in the face of challenges, and enhancing overall well-being. This paper introduces Social Support Detection (SSD) as a Natural language processing (NLP) task aimed at identifying supportive interactions within online communities. The study presents the task of Social Support Detection (SSD) in three subtasks: two binary classification tasks and one multiclass task, with labels detailed in the dataset section. We conducted experiments on a dataset comprising 10,000 YouTube comments. Traditional machine learning models were employed, utilizing various feature combinations that encompass linguistic, psycholinguistic, emotional, and sentiment information. Additionally, we experimented with neural network-based models using various word embeddings to enhance the performance of our models across these subtasks.The results reveal a prevalence of group-oriented support in online dialogues, reflecting broader societal patterns. The findings demonstrate the effectiveness of integrating psycholinguistic, emotional, and sentiment features with n-grams in detecting social support and distinguishing whether it is directed toward an individual or a group. The best results for different subtasks across all experiments range from 0.72 to 0.82.
Authors: An Nghiep Huynh, Thanh Dat Do, Trong Hop Do
Abstract: Regional discrimination is a persistent social issue in Vietnam. While existing research has explored hate speech in the Vietnamese language, the specific issue of regional discrimination remains under-addressed. Previous studies primarily focused on model development without considering practical system implementation. In this work, we propose a task called Detection of Regional Discriminatory Comments on Vietnamese Social Media, leveraging the power of machine learning and transfer learning models. We have built the ViRDC (Vietnamese Regional Discrimination Comments) dataset, which contains comments from social media platforms, providing a valuable resource for further research and development. Our approach integrates streaming capabilities to process real-time data from social media networks, ensuring the system's scalability and responsiveness. We developed the system on the Apache Spark framework to efficiently handle increasing data inputs during streaming. Our system offers a comprehensive solution for the real-time detection of regional discrimination in Vietnam.
Authors: Philip Lippmann, Konrad Skublicki, Joshua Tanner, Shonosuke Ishiwatari, Jie Yang
Abstract: Due to the significant time and effort required for handcrafting translations, most manga never leave the domestic Japanese market. Automatic manga translation is a promising potential solution. However, it is a budding and underdeveloped field and presents complexities even greater than those found in standard translation due to the need to effectively incorporate visual elements into the translation process to resolve ambiguities. In this work, we investigate to what extent multimodal large language models (LLMs) can provide effective manga translation, thereby assisting manga authors and publishers in reaching wider audiences. Specifically, we propose a methodology that leverages the vision component of multimodal LLMs to improve translation quality and evaluate the impact of translation unit size, context length, and propose a token efficient approach for manga translation. Moreover, we introduce a new evaluation dataset -- the first parallel Japanese-Polish manga translation dataset -- as part of a benchmark to be used in future research. Finally, we contribute an open-source software suite, enabling others to benchmark LLMs for manga translation. Our findings demonstrate that our proposed methods achieve state-of-the-art results for Japanese-English translation and set a new standard for Japanese-Polish.
Authors: Harshavardhana T. Gowda, Zachary D. McNaughton, Lee M. Miller
Abstract: Each year, millions of individuals lose the ability to speak intelligibly due to causes such as neuromuscular disease, stroke, trauma, and head/neck cancer surgery (e.g. laryngectomy) or treatment (e.g. radiotherapy toxicity to the speech articulators). Effective communication is crucial for daily activities, and losing the ability to speak leads to isolation, depression, anxiety, and a host of detrimental sequelae. Noninvasive surface electromyography (sEMG) has shown promise to restore speech output in these individuals. The goal is to collect sEMG signals from multiple articulatory sites as people silently produce speech and then decode the signals to enable fluent and natural communication. Currently, many fundamental properties of orofacial neuromuscular signals relating to speech articulation remain unanswered. They include questions relating to 1) the data structure of the orofacial sEMG signals, 2)the signal distribution shift of sEMG across individuals, 3) ability of sEMG signals to span the entire English language phonetic space during silent speech articulations, and 4) the generalization capability of non-invasive sEMG based silent speech interfaces. We address these questions through a series of experiments involving healthy human subjects. We show that sEMG signals evince graph data structure and that the signal distribution shift is given by a change of basis. Furthermore, we show that silently voiced articulations spanning the entire English language phonetic space can be decoded using small neural networks which can be trained with little data and that such architectures work well across individuals. To ensure transparency and reproducibility, we open-source all the data and codes used in this study.
Authors: Fan Nie, Xiaotian Hou, Shuhang Lin, James Zou, Huaxiu Yao, Linjun Zhang
Abstract: The propensity of Large Language Models (LLMs) to generate hallucinations and non-factual content undermines their reliability in high-stakes domains, where rigorous control over Type I errors (the conditional probability of incorrectly classifying hallucinations as truthful content) is essential. Despite its importance, formal verification of LLM factuality with such guarantees remains largely unexplored. In this paper, we introduce FactTest, a novel framework that statistically assesses whether an LLM can confidently provide correct answers to given questions with high-probability correctness guarantees. We formulate factuality testing as hypothesis testing problem to enforce an upper bound of Type I errors at user-specified significance levels. Notably, we prove that our framework also ensures strong Type II error control under mild conditions and can be extended to maintain its effectiveness when covariate shifts exist. %These analyses are amenable to the principled NP framework. Our approach is distribution-free and works for any number of human-annotated samples. It is model-agnostic and applies to any black-box or white-box LM. Extensive experiments on question-answering (QA) and multiple-choice benchmarks demonstrate that \approach effectively detects hallucinations and improves the model's ability to abstain from answering unknown questions, leading to an over 40% accuracy improvement.
Authors: Wei He, Tiago Kramer Vieira, Marcos Garcia, Carolina Scarton, Marco Idiart, Aline Villavicencio
Abstract: Idiomatic expressions are an integral part of human languages, often used to express complex ideas in compressed or conventional ways (e.g. eager beaver as a keen and enthusiastic person). However, their interpretations may not be straightforwardly linked to the meanings of their individual components in isolation and this may have an impact for compositional approaches. In this paper, we investigate to what extent word representation models are able to go beyond compositional word combinations and capture multiword expression idiomaticity and some of the expected properties related to idiomatic meanings. We focus on noun compounds of varying levels of idiomaticity in two languages (English and Portuguese), presenting a dataset of minimal pairs containing human idiomaticity judgments for each noun compound at both type and token levels, their paraphrases and their occurrences in naturalistic and sense-neutral contexts, totalling 32,200 sentences. We propose this set of minimal pairs for evaluating how well a model captures idiomatic meanings, and define a set of fine-grained metrics of Affinity and Scaled Similarity, to determine how sensitive the models are to perturbations that may lead to changes in idiomaticity. The results obtained with a variety of representative and widely used models indicate that, despite superficial indications to the contrary in the form of high similarities, idiomaticity is not yet accurately represented in current models. Moreover, the performance of models with different levels of contextualisation suggests that their ability to capture context is not yet able to go beyond more superficial lexical clues provided by the words and to actually incorporate the relevant semantic clues needed for idiomaticity.
Authors: Nouf Alabbasi, Omar Erak, Omar Alhussein, Ismail Lotfi, Sami Muhaidat, Merouane Debbah
Abstract: The telecommunications industry's rapid evolution demands intelligent systems capable of managing complex networks and adapting to emerging technologies. While large language models (LLMs) show promise in addressing these challenges, their deployment in telecom environments faces significant constraints due to edge device limitations and inconsistent documentation. To bridge this gap, we present TeleOracle, a telecom-specialized retrieval-augmented generation (RAG) system built on the Phi-2 small language model (SLM). To improve context retrieval, TeleOracle employs a two-stage retriever that incorporates semantic chunking and hybrid keyword and semantic search. Additionally, we expand the context window during inference to enhance the model's performance on open-ended queries. We also employ low-rank adaption for efficient fine-tuning. A thorough analysis of the model's performance indicates that our RAG framework is effective in aligning Phi-2 to the telecom domain in a downstream question and answer (QnA) task, achieving a 30% improvement in accuracy over the base Phi-2 model, reaching an overall accuracy of 81.20%. Notably, we show that our model not only performs on par with the much larger LLMs but also achieves a higher faithfulness score, indicating higher adherence to the retrieved context.
Authors: Atakan Seyito\u{g}lu, Aleksei Kuvshinov, Leo Schwinn, Stephan G\"unnemann
Abstract: An unintended consequence of the vast pretraining of Large Language Models (LLMs) is the verbatim memorization of fragments of their training data, which may contain sensitive or copyrighted information. In recent years, unlearning has emerged as a solution to effectively remove sensitive knowledge from models after training. Yet, recent work has shown that supposedly deleted information can still be extracted by malicious actors through various attacks. Still, current attacks retrieve sets of possible candidate generations and are unable to pinpoint the output that contains the actual target information. We propose activation steering as a method for exact information retrieval from unlearned LLMs. We introduce a novel approach to generating steering vectors, named Anonymized Activation Steering. Additionally, we develop a simple word frequency method to pinpoint the correct answer among a set of candidates when retrieving unlearned information. Our evaluation across multiple unlearning techniques and datasets demonstrates that activation steering successfully recovers general knowledge (e.g., widely known fictional characters) while revealing limitations in retrieving specific information (e.g., details about non-public individuals). Overall, our results demonstrate that exact information retrieval from unlearned models is possible, highlighting a severe vulnerability of current unlearning techniques.
Authors: Stephen McAleese, Mark Keane
Abstract: Counterfactual explanations can be used to interpret and debug text classifiers by producing minimally altered text inputs that change a classifier's output. In this work, we evaluate five methods for generating counterfactual explanations for a BERT text classifier on two datasets using three evaluation metrics. The results of our experiments suggest that established white-box substitution-based methods are effective at generating valid counterfactuals that change the classifier's output. In contrast, newer methods based on large language models (LLMs) excel at producing natural and linguistically plausible text counterfactuals but often fail to generate valid counterfactuals that alter the classifier's output. Based on these results, we recommend developing new counterfactual explanation methods that combine the strengths of established gradient-based approaches and newer LLM-based techniques to generate high-quality, valid, and plausible text counterfactual explanations.
Authors: Karthik Soman, Andrew Langdon, Catalina Villouta, Chinmay Agrawal, Lashaw Salta, Braian Peetoom, Gianmarco Bellucci, Orion J Buske
Abstract: Rare diseases present unique challenges in healthcare, often suffering from delayed diagnosis and fragmented information landscapes. The scarcity of reliable knowledge in these conditions poses a distinct challenge for Large Language Models (LLMs) in supporting clinical management and delivering precise patient information underscoring the need for focused training on these 'zebra' cases. We present Zebra-Llama, a specialized context-aware language model with high precision Retrieval Augmented Generation (RAG) capability, focusing on Ehlers-Danlos Syndrome (EDS) as our case study. EDS, affecting 1 in 5,000 individuals, exemplifies the complexities of rare diseases with its diverse symptoms, multiple subtypes, and evolving diagnostic criteria. By implementing a novel context-aware fine-tuning methodology trained on questions derived from medical literature, patient experiences, and clinical resources, along with expertly curated responses, Zebra-Llama demonstrates unprecedented capabilities in handling EDS-related queries. On a test set of real-world questions collected from EDS patients and clinicians, medical experts evaluated the responses generated by both models, revealing Zebra-Llama's substantial improvements over base model (Llama 3.1-8B-Instruct) in thoroughness (77.5% vs. 70.1%), accuracy (83.0% vs. 78.8%), clarity (74.7% vs. 72.0%) and citation reliability (70.6% vs. 52.3%). Released as an open-source resource, Zebra-Llama not only provides more accessible and reliable EDS information but also establishes a framework for developing specialized AI solutions for other rare conditions. This work represents a crucial step towards democratizing expert-level knowledge in rare disease management, potentially transforming how healthcare providers and patients navigate the complex landscape of rare diseases.
Authors: Xin Zhang, Victor S. Sheng
Abstract: We propose an innovative token representation and update method in a new ultra-small language model: the Wave network. Specifically, we use a \textbf{complex vector} to represent each token, encoding both global and local semantics of the input text. A \textbf{complex vector} consists of two components: a magnitude vector representing the \textit{global semantics} of the input text, and a phase vector capturing the \textit{relationships between individual tokens and global semantics}. Experiments on the AG News text classification task demonstrate that, when generating complex vectors from randomly initialized token embeddings, our single-layer Wave Network achieves 90.91\% accuracy with wave interference and 91.66\% with wave modulation -- outperforming a single Transformer layer using BERT pre-trained embeddings by 19.23\% and 19.98\%, respectively, and approaching the accuracy of the pre-trained and fine-tuned BERT base model (94.64\%). Additionally, compared to BERT base, the Wave Network reduces video memory usage and training time by 77.34\% and 85.62\% during wave modulation. In summary, we used a 2.4-million-parameter small language model to achieve accuracy comparable to a 100-million-parameter BERT model in text classification.
Authors: Yihan Wang, Andrew Bai, Nanyun Peng, Cho-Jui Hsieh
Abstract: Pretrained Large Language Models (LLMs) require post-training methods such as supervised fine-tuning (SFT) on instruction-response pairs to enable instruction following. However, this process can potentially harm existing capabilities learned during pretraining. In this paper, we investigate the loss of context awareness after SFT, defined as the capability to extract and understand information from the user-provided context and respond accordingly. We are the first to identify and show that the loss of context-awareness appears on instruction-finetuned LLMs when the chat template is applied to the input prompts. We identify the performance decline is partially caused by the bias embedded into the chat template to focus less on the user-provided context. Based on these observations, we propose two methods to mitigate the loss of context awareness in instruct models: post-hoc attention steering on user prompts and conditional instruction fine-tuning with a context-dependency indicator. Empirical experiments on 4 context-dependent downstream tasks and 3 pretrained LLMs of different sizes show that our methods effectively mitigates the loss of context awareness without compromising the general ability to follow instructions. Our findings also strongly advocate the necessity to carefully benchmark context awareness after instruction fine-tuning.
Authors: Seyed Hossein Alavi, Weijia Xu, Nebojsa Jojic, Daniel Kennett, Raymond T. Ng, Sudha Rao, Haiyan Zhang, Bill Dolan, Vered Shwartz
Abstract: We introduce GamePlot, an LLM-powered assistant that supports game designers in crafting immersive narratives for turn-based games, and allows them to test these games through a collaborative game play and refine the plot throughout the process. Our user study with 14 game designers shows high levels of both satisfaction with the generated game plots and sense of ownership over the narratives, but also reconfirms that LLM are limited in their ability to generate complex and truly innovative content. We also show that diverse user populations have different expectations from AI assistants, and encourage researchers to study how tailoring assistants to diverse user groups could potentially lead to increased job satisfaction and greater creativity and innovation over time.
Authors: Shuo Yang, Siwen Luo, Soyeon Caren Han
Abstract: Existing Multimodal Large Language Models (MLLMs) and Visual Language Pretrained Models (VLPMs) have shown remarkable performances in the general Visual Question Answering (VQA). However, these models struggle with VQA questions that require external commonsense knowledge due to the challenges in generating high-quality prompts and the high computational costs of fine-tuning. In this work, we propose a novel graph-based multimodal commonsense knowledge distillation framework that constructs a unified relational graph over commonsense knowledge, visual objects and questions through a Graph Convolutional Network (GCN) following a teacher-student environment. This proposed framework is flexible with any type of teacher and student models without further fine-tuning, and has achieved competitive performances on the ScienceQA dataset.
Authors: Zexu Li, Suraj P. Prabhu, Zachary T. Popp, Shubhi S. Jain, Vijetha Balakundi, Ting Fang Alvin Ang, Rhoda Au, Jinying Chen
Abstract: Biomedical research requires large, diverse samples to produce unbiased results. Automated methods for matching variables across datasets can accelerate this process. Research in this area has been limited, primarily focusing on lexical matching and ontology based semantic matching. We aimed to develop new methods, leveraging large language models (LLM) and ensemble learning, to automate variable matching. Methods: We utilized data from two GERAS cohort (European and Japan) studies to develop variable matching methods. We first manually created a dataset by matching 352 EU variables with 1322 candidate JP variables, where matched variable pairs were positive and unmatched pairs were negative instances. Using this dataset, we developed and evaluated two types of natural language processing (NLP) methods, which matched variables based on variable labels and definitions from data dictionaries: (1) LLM-based and (2) fuzzy matching. We then developed an ensemble-learning method, using the Random Forest model, to integrate individual NLP methods. RF was trained and evaluated on 50 trials. Each trial had a random split (4:1) of training and test sets, with the model's hyperparameters optimized through cross-validation on the training set. For each EU variable, 1322 candidate JP variables were ranked based on NLP-derived similarity scores or RF's probability scores, denoting their likelihood to match the EU variable. Ranking performance was measured by top-n hit ratio (HRn) and mean reciprocal rank (MRR). Results:E5 performed best among individual methods, achieving 0.90 HR-30 and 0.70 MRR. RF performed better than E5 on all metrics over 50 trials (P less than 0.001) and achieved an average HR 30 of 0.98 and MRR of 0.73. LLM-derived features contributed most to RF's performance. One major cause of errors in automatic variable matching was ambiguous variable definitions within data dictionaries.
Authors: Jaewoong Choi
Abstract: While numerous studies have explored the field of research and development (R&D) landscaping, the preponderance of these investigations has emphasized predictive analysis based on R&D outcomes, specifically patents, and academic literature. However, the value of research proposals and novelty analysis has seldom been addressed. This study proposes a systematic approach to constructing and navigating the R&D landscape that can be utilized to guide organizations to respond in a reproducible and timely manner to the challenges presented by increasing number of research proposals. At the heart of the proposed approach is the composite use of the transformer-based language model and the local outlier factor (LOF). The semantic meaning of the research proposals is captured with our further-trained transformers, thereby constructing a comprehensive R&D landscape. Subsequently, the novelty of the newly selected research proposals within the annual landscape is quantified on a numerical scale utilizing the LOF by assessing the dissimilarity of each proposal to others preceding and within the same year. A case study examining research proposals in the energy and resource sector in South Korea is presented. The systematic process and quantitative outcomes are expected to be useful decision-support tools, providing future insights regarding R&D planning and roadmapping.
Authors: Jianqiao Wangni
Abstract: This paper introduces a novel framework that leverages large language models (LLMs) for machine translation (MT). We start with one conjecture: an ideal translation should contain complete and accurate information for a strong enough LLM to recover the original sentence. We generate multiple translation candidates from a source language A to a target language B, and subsequently translate these candidates back to the original language A. By evaluating the cycle consistency between the original and back-translated sentences using metrics such as token-level precision and accuracy, we implicitly estimate the translation quality in language B, without knowing its ground-truth. This also helps to evaluate the LLM translation capability, only with monolingual corpora. For each source sentence, we identify the translation candidate with optimal cycle consistency with the original sentence as the final answer. Our experiments demonstrate that larger LLMs, or the same LLM with more forward passes during inference, exhibit increased cycle consistency, aligning with the LLM model size scaling law and test-time computation scaling law. This work provide methods for, 1) to implicitly evaluate translation quality of a sentence in the target language, 2), to evaluate capability of LLM for any-to-any-language translation, and 3), how to generate a better translation for a specific LLM.
Authors: Mingcheng Li, Dingkang Yang, Yang Liu, Shunli Wang, Jiawei Chen, Shuaibing Wang, Jinjie Wei, Yue Jiang, Qingyao Xu, Xiaolu Hou, Mingyang Sun, Ziyun Qian, Dongliang Kou, Lihua Zhang
Abstract: Multimodal Sentiment Analysis (MSA) is an important research area that aims to understand and recognize human sentiment through multiple modalities. The complementary information provided by multimodal fusion promotes better sentiment analysis compared to utilizing only a single modality. Nevertheless, in real-world applications, many unavoidable factors may lead to situations of uncertain modality missing, thus hindering the effectiveness of multimodal modeling and degrading the model's performance. To this end, we propose a Hierarchical Representation Learning Framework (HRLF) for the MSA task under uncertain missing modalities. Specifically, we propose a fine-grained representation factorization module that sufficiently extracts valuable sentiment information by factorizing modality into sentiment-relevant and modality-specific representations through crossmodal translation and sentiment semantic reconstruction. Moreover, a hierarchical mutual information maximization mechanism is introduced to incrementally maximize the mutual information between multi-scale representations to align and reconstruct the high-level semantics in the representations. Ultimately, we propose a hierarchical adversarial learning mechanism that further aligns and adapts the latent distribution of sentiment-relevant representations to produce robust joint multimodal representations. Comprehensive experiments on three datasets demonstrate that HRLF significantly improves MSA performance under uncertain modality missing cases.
Authors: Akul Datta
Abstract: This paper reviews the development of the Receptance Weighted Key Value (RWKV) architecture, emphasizing its advancements in efficient language modeling. RWKV combines the training efficiency of Transformers with the inference efficiency of RNNs through a novel linear attention mechanism. We examine its core innovations, adaptations across various domains, and performance advantages over traditional models. The paper also discusses challenges and future directions for RWKV as a versatile architecture in deep learning.
Authors: Giwon Hong, Emile van Krieken, Edoardo Ponti, Nikolay Malkin, Pasquale Minervini
Abstract: In-context learning (ICL) adapts LLMs by providing demonstrations without fine-tuning the model parameters; however, it does not differentiate between demonstrations and quadratically increases the complexity of Transformer LLMs, exhausting the memory. As a solution, we propose Mixtures of In-Context Learners (MoICL), a novel approach to treat subsets of demonstrations as experts and learn a weighting function to merge their output distributions based on a training set. In our experiments, we show performance improvements on 5 out of 7 classification datasets compared to a set of strong baselines (up to +13\% compared to ICL and LENS). Moreover, we enhance the Pareto frontier of ICL by reducing the inference time needed to achieve the same performance with fewer demonstrations. Finally, MoICL is more robust to out-of-domain (up to +11\%), imbalanced (up to +49\%), or noisy demonstrations (up to +38\%) or can filter these out from datasets. Overall, MoICL is a more expressive approach to learning from demonstrations without exhausting the context window or memory.
Authors: Hossein Hosseini, Mohammad Siobhan Zare, Amir Hossein Mohammadi, Arefeh Kazemi, Zahra Zojaji, Mohammad Ali Nematbakhsh
Abstract: Retrieval augmented generation (RAG) models, which integrate large-scale pre-trained generative models with external retrieval mechanisms, have shown significant success in various natural language processing (NLP) tasks. However, applying RAG models in Persian language as a low-resource language, poses distinct challenges. These challenges primarily involve the preprocessing, embedding, retrieval, prompt construction, language modeling, and response evaluation of the system. In this paper, we address the challenges towards implementing a real-world RAG system for Persian language called PersianRAG. We propose novel solutions to overcome these obstacles and evaluate our approach using several Persian benchmark datasets. Our experimental results demonstrate the capability of the PersianRAG framework to enhance question answering task in Persian.
Authors: Tao Zhang, Ning Yan, Masood Mortazavi, Hoang H. Nguyen, Zhongfen Deng, Philip S. Yu
Abstract: Large language models (LLMs) pre-trained on massive corpora have demonstrated impressive few-shot learning capability on many NLP tasks. Recasting an NLP task into a text-to-text generation task is a common practice so that generative LLMs can be prompted to resolve it. However, performing document-level relation extraction (DocRE) tasks with generative LLM models is still challenging due to the structured output format of DocRE, which complicates the conversion to plain text. Limited information available in few-shot samples and prompt instructions induce further difficulties and challenges in relation extraction for mentioned entities in a document. In this paper, we represent the structured output as a graph-style triplet rather than natural language expressions and leverage generative LLMs for the DocRE task. Our approach, the Graph-DPEP framework is grounded in the reasoning behind triplet explanation thoughts presented in natural language. In this framework, we first introduce a ``decomposed-plug" method for performing the generation from LLMs over prompts with type-space decomposition to alleviate the burden of distinguishing all relation types. Second, we employ a verifier for calibrating the generation and identifying overlooked query entity pairs. Third, we develop "ensemble-play", reapplying generation on the entire type list by leveraging the reasoning thoughts embedded in a sub-graph associated with the missing query pair to address the missingness issue. Through extensive comparisons with existing prompt techniques and alternative Language Models (LLMs), our framework demonstrates superior performance on publicly available benchmarks in experiments.
Authors: Wei Wu, Zhuoshi Pan, Chao Wang, Liyi Chen, Yunchu Bai, Kun Fu, Zheng Wang, Hui Xiong
Abstract: With the development of large language models (LLMs), the ability to handle longer contexts has become a key capability for Web applications such as cross-document understanding and LLM-powered search systems. However, this progress faces two major challenges: performance degradation due to sequence lengths out-of-distribution, and excessively long inference times caused by the quadratic computational complexity of attention. These issues hinder the application of LLMs in long-context scenarios. In this paper, we propose Dynamic Token-Level KV Cache Selection (TokenSelect), a model-agnostic, training-free method for efficient and accurate long-context inference. TokenSelect builds upon the observation of non-contiguous attention sparsity, using Query-Key dot products to measure per-head KV Cache criticality at token-level. By per-head soft voting mechanism, TokenSelect selectively involves a small number of critical KV cache tokens in the attention calculation without sacrificing accuracy. To further accelerate TokenSelect, we designed the Selection Cache based on observations of consecutive Query similarity and implemented efficient dot product kernel, significantly reducing the overhead of token selection. A comprehensive evaluation of TokenSelect demonstrates up to 23.84x speedup in attention computation and up to 2.28x acceleration in end-to-end latency, while providing superior performance compared to state-of-the-art long-context inference methods.
Authors: Dalal Waadallah Shehab
Abstract: This study investigates the translation of circumlocution from Arabic to English in a corpus of short stories by renowned Arabic authors. By analyzing the source and target texts, the study aims to identify and categorize circumlocution instances in Arabic and their corresponding renditions in English. The study employs Nida's (1964) translation theory as a framework to assess the appropriateness of the translation strategies employed. It examines the extent to which translators successfully rendered Arabic circumlocution into English, identifying potential challenges and limitations in the translation process. The findings reveal significant similarities between Arabic circumlocution categories and English metadiscourse categories, particularly in terms of textual and interpersonal functions. However, the study also highlights instances where translators encountered difficulties in accurately conveying the nuances of circumlocution, often resorting to strategies like addition, subtraction, and alteration.
Authors: Lingjie Jiang, Shaohan Huang, Xun Wu, Furu Wei
Abstract: Image aesthetics is a crucial metric in the field of image generation. However, textual aesthetics has not been sufficiently explored. With the widespread application of large language models (LLMs), previous work has primarily focused on the correctness of content and the helpfulness of responses. Nonetheless, providing responses with textual aesthetics is also an important factor for LLMs, which can offer a cleaner layout and ensure greater consistency and coherence in content. In this work, we introduce a pipeline for aesthetics polishing and help construct a textual aesthetics dataset named TexAes. We propose a textual aesthetics-powered fine-tuning method based on direct preference optimization, termed TAPO, which leverages textual aesthetics without compromising content correctness. Additionally, we develop two evaluation methods for textual aesthetics based on text and image analysis, respectively. Our experiments demonstrate that using textual aesthetics data and employing the TAPO fine-tuning method not only improves aesthetic scores but also enhances performance on general evaluation datasets such as AlpacalEval and Anera-hard.
Authors: Yangning Li, Yinghui Li, Xingyu Wang, Yong Jiang, Zhen Zhang, Xinran Zheng, Hui Wang, Hai-Tao Zheng, Philip S. Yu, Fei Huang, Jingren Zhou
Abstract: Multimodal Retrieval Augmented Generation (mRAG) plays an important role in mitigating the "hallucination" issue inherent in multimodal large language models (MLLMs). Although promising, existing heuristic mRAGs typically predefined fixed retrieval processes, which causes two issues: (1) Non-adaptive Retrieval Queries. (2) Overloaded Retrieval Queries. However, these flaws cannot be adequately reflected by current knowledge-seeking visual question answering (VQA) datasets, since the most required knowledge can be readily obtained with a standard two-step retrieval. To bridge the dataset gap, we first construct Dyn-VQA dataset, consisting of three types of "dynamic" questions, which require complex knowledge retrieval strategies variable in query, tool, and time: (1) Questions with rapidly changing answers. (2) Questions requiring multi-modal knowledge. (3) Multi-hop questions. Experiments on Dyn-VQA reveal that existing heuristic mRAGs struggle to provide sufficient and precisely relevant knowledge for dynamic questions due to their rigid retrieval processes. Hence, we further propose the first self-adaptive planning agent for multimodal retrieval, OmniSearch. The underlying idea is to emulate the human behavior in question solution which dynamically decomposes complex multimodal questions into sub-question chains with retrieval action. Extensive experiments prove the effectiveness of our OmniSearch, also provide direction for advancing mRAG. The code and dataset will be open-sourced at https://github.com/Alibaba-NLP/OmniSearch.
Authors: Keke Zhai
Abstract: This paper delves into the continuous post-training optimization methods for small language models, and proposes a continuous post-training alignment data construction method for small language models. The core of this method is based on the data guidance of large models, optimizing the diversity and accuracy of alignment data. In addition, to verify the effectiveness of the methods in this paper, we used Qwen2-0.5B-Instruct model as the baseline model for small language models, using the alignment dataset constructed by our proposed method, we trained and compared several groups of experiments, including SFT (Supervised Fine Tuning) post-training experiment and KTO (Kahneman Tversky optimization) post-training experiment, as well as SFT-KTO two-stage post-training experiment and model weight fusion experiment. Finally, we evaluated and analyzed the performance of post-training models, and confirmed that the continuous post-training optimization method proposed by us can significantly improve the performance of small language models.
Authors: Francesco Invernici, Francesca Curati, Jelena Jakimov, Amirhossein Samavi, Anna Bernasconi
Abstract: The world is facing a multitude of challenges that hinder the development of human civilization and the well-being of humanity on the planet. The Sustainable Development Goals (SDGs) were formulated by the United Nations in 2015 to address these global challenges by 2030. Natural language processing techniques can help uncover discussions on SDGs within research literature. We propose a completely automated pipeline to 1) fetch content from the Scopus database and prepare datasets dedicated to five groups of SDGs; 2) perform topic modeling, a statistical technique used to identify topics in large collections of textual data; and 3) enable topic exploration through keywords-based search and topic frequency time series extraction. For topic modeling, we leverage the stack of BERTopic scaled up to be applied on large corpora of textual documents (we find hundreds of topics on hundreds of thousands of documents), introducing i) a novel LLM-based embeddings computation for representing scientific abstracts in the continuous space and ii) a hyperparameter optimizer to efficiently find the best configuration for any new big datasets. We additionally produce the visualization of results on interactive dashboards reporting topics' temporal evolution. Results are made inspectable and explorable, contributing to the interpretability of the topic modeling process. Our proposed LLM-based topic modeling pipeline for big-text datasets allows users to capture insights on the evolution of the attitude toward SDGs within scientific abstracts in the 2006-2023 time span. All the results are reproducible by using our system; the workflow can be generalized to be applied at any point in time to any big corpus of textual documents.
Authors: Maren Pielka, Tobias Schneider, Jan Terheyden, Rafet Sifa
Abstract: We present an outline of the first large language model (LLM) based chatbot application in the context of patient-reported outcome measures (PROMs) for diabetic retinopathy. By utilizing the capabilities of current LLMs, we enable patients to provide feedback about their quality of life and treatment progress via an interactive application. The proposed framework offers significant advantages over the current approach, which encompasses only qualitative collection of survey data or a static survey with limited answer options. Using the PROBot LLM-PROM application, patients will be asked tailored questions about their individual challenges, and can give more detailed feedback on the progress of their treatment. Based on this input, we will use machine learning to infer conventional PROM scores, which can be used by clinicians to evaluate the treatment status. The goal of the application is to improve adherence to the healthcare system and treatments, and thus ultimately reduce cases of subsequent vision impairment. The approach needs to be further validated using a survey and a clinical study.
Authors: Michal Shur-Ofry, Bar Horowitz-Amsalem, Adir Rahamim, Yonatan Belinkov
Abstract: How diverse are the outputs of large language models when diversity is desired? We examine the diversity of responses of various models to questions with multiple possible answers, comparing them with human responses. Our findings suggest that models' outputs are highly concentrated, reflecting a narrow, mainstream 'worldview', in comparison to humans, whose responses exhibit a much longer-tail. We examine three ways to increase models' output diversity: 1) increasing generation randomness via temperature sampling; 2) prompting models to answer from diverse perspectives; 3) aggregating outputs from several models. A combination of these measures significantly increases models' output diversity, reaching that of humans. We discuss implications of these findings for AI policy that wishes to preserve cultural diversity, an essential building block of a democratic social fabric.
Authors: Georgy Andryushchenko, Vladimir Ivanov, Vladimir Makharev, Elizaveta Tukhtina, Aidar Valeev
Abstract: Question answering over source code provides software engineers and project managers with helpful information about the implemented features of a software product. This paper presents a work devoted to using large language models for question answering over source code in Python. The proposed method for implementing a source code question answering system involves fine-tuning a large language model on a unified dataset of questions and answers for Python code. To achieve the highest quality answers, we tested various models trained on datasets preprocessed in different ways: a dataset without grammar correction, a dataset with grammar correction, and a dataset augmented with the generated summaries. The model answers were also analyzed for errors manually. We report BLEU-4, BERTScore F1, BLEURT, and Exact Match metric values, along with the conclusions from the manual error analysis. The obtained experimental results highlight the current problems of the research area, such as poor quality of the public genuine question-answering datasets. In addition, the findings include the positive effect of the grammar correction of the training data on the testing metric values. The addressed findings and issues could be important for other researchers who attempt to improve the quality of source code question answering solutions. The training and evaluation code is publicly available at https://github.com/IU-AES-AI4Code/CodeQuestionAnswering.
URLs: https://github.com/IU-AES-AI4Code/CodeQuestionAnswering.
Authors: Mael Houbre, Florian Boudin, Beatrice Daille, Akiko Aizawa
Abstract: State-of-the-art models for keyphrase generation require large amounts of training data to achieve good performance. However, obtaining keyphrase-labeled documents can be challenging and costly. To address this issue, we present a self-compositional data augmentation method. More specifically, we measure the relatedness of training documents based on their shared keyphrases, and combine similar documents to generate synthetic samples. The advantage of our method lies in its ability to create additional training samples that keep domain coherence, without relying on external data or resources. Our results on multiple datasets spanning three different domains, demonstrate that our method consistently improves keyphrase generation. A qualitative analysis of the generated keyphrases for the Computer Science domain confirms this improvement towards their representativity property.
Authors: Bei Li, Tong Zheng, Rui Wang, Jiahao Liu, Qingyan Guo, Junliang Guo, Xu Tan, Tong Xiao, Jingbo Zhu, Jingang Wang, Xunliang Cai
Abstract: Residual networks, as discrete approximations of Ordinary Differential Equations (ODEs), have inspired significant advancements in neural network design, including multistep methods, high-order methods, and multi-particle dynamical systems. The precision of the solution to ODEs significantly affects parameter optimization, thereby impacting model performance. In this work, we present a series of advanced explorations of Transformer architecture design to minimize the error compared to the true ``solution.'' First, we introduce a predictor-corrector learning framework to minimize truncation errors, which consists of a high-order predictor and a multistep corrector. Second, we propose an exponential moving average-based coefficient learning method to strengthen our higher-order predictor. Extensive experiments on large-scale machine translation, abstractive summarization, language modeling, and natural language understanding benchmarks demonstrate the superiority of our approach. On the WMT'14 English-German and English-French tasks, our model achieved BLEU scores of 30.95 and 44.27, respectively. Furthermore, on the OPUS multilingual machine translation task, our model surpasses a robust 3.8B DeepNet by an average of 2.9 SacreBLEU, using only 1/3 parameters. Notably, it also beats LLama models by 5.7 accuracy points on the LM Harness Evaluation.
Authors: Rajkumar Ramamurthy, Meghana Arakkal Rajeev, Oliver Molenschot, James Zou, Nazneen Rajani
Abstract: Large language models (LLMs) often fail to synthesize information from their context to generate an accurate response. This renders them unreliable in knowledge intensive settings where reliability of the output is key. A critical component for reliable LLMs is the integration of a robust fact-checking system that can detect hallucinations across various formats. While several open-access fact-checking models are available, their functionality is often limited to specific tasks, such as grounded question-answering or entailment verification, and they perform less effectively in conversational settings. On the other hand, closed-access models like GPT-4 and Claude offer greater flexibility across different contexts, including grounded dialogue verification, but are hindered by high costs and latency. In this work, we introduce VERITAS, a family of hallucination detection models designed to operate flexibly across diverse contexts while minimizing latency and costs. VERITAS achieves state-of-the-art results considering average performance on all major hallucination detection benchmarks, with $10\%$ increase in average performance when compared to similar-sized models and get close to the performance of GPT4 turbo with LLM-as-a-judge setting.
Authors: Reynier Leyva La O, Carlos A. Catania, Tatiana Parlanti
Abstract: This work analyzes the use of large language models (LLMs) for detecting domain generation algorithms (DGAs). We perform a detailed evaluation of two important techniques: In-Context Learning (ICL) and Supervised Fine-Tuning (SFT), showing how they can improve detection. SFT increases performance by using domain-specific data, whereas ICL helps the detection model to quickly adapt to new threats without requiring much retraining. We use Meta's Llama3 8B model, on a custom dataset with 68 malware families and normal domains, covering several hard-to-detect schemes, including recent word-based DGAs. Results proved that LLM-based methods can achieve competitive results in DGA detection. In particular, the SFT-based LLM DGA detector outperforms state-of-the-art models using attention layers, achieving 94% accuracy with a 4% false positive rate (FPR) and excelling at detecting word-based DGA domains.
Authors: Qisheng Hu, Quanyu Long, Wenya Wang
Abstract: Fact-checking pipelines increasingly adopt the Decompose-Then-Verify paradigm, where texts are broken down into smaller claims for individual verification and subsequently combined for a veracity decision. While decomposition is widely-adopted in such pipelines, its effects on final fact-checking performance remain underexplored. Some studies have reported improvements from decompostition, while others have observed performance declines, indicating its inconsistent impact. To date, no comprehensive analysis has been conducted to understand this variability. To address this gap, we present an in-depth analysis that explicitly examines the impact of decomposition on downstream verification performance. Through error case inspection and experiments, we introduce a categorization of decomposition errors and reveal a trade-off between accuracy gains and the noise introduced through decomposition. Our analysis provides new insights into understanding current system's instability and offers guidance for future studies toward improving claim decomposition in fact-checking pipelines.
Authors: Hansa Meghwani
Abstract: Ranking consistently emerges as a primary focus in information retrieval research. Retrieval and ranking models serve as the foundation for numerous applications, including web search, open domain QA, enterprise domain QA, and text-based recommender systems. Typically, these models undergo training on triplets consisting of binary relevance assignments, comprising one positive and one negative passage. However, their utilization involves a context where a significantly more nuanced understanding of relevance is necessary, especially when re-ranking a large pool of potentially relevant passages. Although collecting positive examples through user feedback like impressions or clicks is straightforward, identifying suitable negative pairs from a vast pool of possibly millions or even billions of documents possess a greater challenge. Generating a substantial number of negative pairs is often necessary to maintain the high quality of the model. Several approaches have been suggested in literature to tackle the issue of selecting suitable negative pairs from an extensive corpus. This study focuses on explaining the crucial role of hard negatives in the training process of cross-encoder models, specifically aiming to explain the performance gains observed with hard negative sampling compared to random sampling. We have developed a robust hard negative mining technique for efficient training of cross-encoder re-rank models on an enterprise dataset which has domain specific context. We provide a novel perspective to enhance retrieval models, ultimately influencing the performance of advanced LLM systems like Retrieval-Augmented Generation (RAG) and Reasoning and Action Agents (ReAct). The proposed approach demonstrates that learning both similarity and dissimilarity simultaneously with cross-encoders improves performance of retrieval systems.
Authors: Vedant Das Swain, Qiuyue "Joy" Zhong, Jash Rajesh Parekh, Yechan Jeon, Roy Zimmerman, Mary Czerwinski, Jina Suh, Varun Mishra, Koustuv Saha, Javier Hernandez
Abstract: Client-Service Representatives (CSRs) are vital to organizations. Frequent interactions with disgruntled clients, however, disrupt their mental well-being. To help CSRs regulate their emotions while interacting with uncivil clients, we designed Pro-Pilot, an LLM-powered assistant, and evaluated its efficacy, perception, and use. Our comparative analyses between 665 human and Pro-Pilot-generated support messages demonstrate Pro-Pilot's ability to adapt to and demonstrate empathy in various incivility incidents. Additionally, 143 CSRs assessed Pro-Pilot's empathy as more sincere and actionable than human messages. Finally, we interviewed 20 CSRs who interacted with Pro-Pilot in a simulation exercise. They reported that Pro-Pilot helped them avoid negative thinking, recenter thoughts, and humanize clients; showing potential for bridging gaps in coworker support. Yet, they also noted deployment challenges and emphasized the irreplaceability of shared experiences. We discuss future designs and societal implications of AI-mediated emotional labor, underscoring empathy as a critical function for AI assistants in front-office roles.
Authors: Jiedong Lang, Zhehao Guo, Shuyu Huang
Abstract: Large Language Models (LLMs) have been extensively researched and used in both academia and industry since the rise in popularity of the Transformer model, which demonstrates excellent performance in AI. However, the computational demands of LLMs are immense, and the energy resources required to run them are often limited. For instance, popular models like GPT-3, with 175 billion parameters and a storage requirement of 350 GB, present significant challenges for deployment on resource-constrained IoT devices and embedded systems. These systems often lack the computational capacity to handle such large models. Quantization, a technique that reduces the precision of model values to a smaller set of discrete values, offers a promising solution by reducing the size of LLMs and accelerating inference. In this research, we provide a comprehensive analysis of quantization techniques within the machine learning field, with a particular focus on their application to LLMs. We begin by exploring the mathematical theory of quantization, followed by a review of common quantization methods and how they are implemented. Furthermore, we examine several prominent quantization methods applied to LLMs, detailing their algorithms and performance outcomes.
Authors: Edward Vendrow, Omiros Pantazis, Alexander Shepard, Gabriel Brostow, Kate E. Jones, Oisin Mac Aodha, Sara Beery, Grant Van Horn
Abstract: We introduce INQUIRE, a text-to-image retrieval benchmark designed to challenge multimodal vision-language models on expert-level queries. INQUIRE includes iNaturalist 2024 (iNat24), a new dataset of five million natural world images, along with 250 expert-level retrieval queries. These queries are paired with all relevant images comprehensively labeled within iNat24, comprising 33,000 total matches. Queries span categories such as species identification, context, behavior, and appearance, emphasizing tasks that require nuanced image understanding and domain expertise. Our benchmark evaluates two core retrieval tasks: (1) INQUIRE-Fullrank, a full dataset ranking task, and (2) INQUIRE-Rerank, a reranking task for refining top-100 retrievals. Detailed evaluation of a range of recent multimodal models demonstrates that INQUIRE poses a significant challenge, with the best models failing to achieve an mAP@50 above 50%. In addition, we show that reranking with more powerful multimodal models can enhance retrieval performance, yet there remains a significant margin for improvement. By focusing on scientifically-motivated ecological challenges, INQUIRE aims to bridge the gap between AI capabilities and the needs of real-world scientific inquiry, encouraging the development of retrieval systems that can assist with accelerating ecological and biodiversity research. Our dataset and code are available at https://inquire-benchmark.github.io
Authors: Maitreya Patel, Abhiram Kusumba, Sheng Cheng, Changhoon Kim, Tejas Gokhale, Chitta Baral, Yezhou Yang
Abstract: Contrastive Language-Image Pretraining (CLIP) models maximize the mutual information between text and visual modalities to learn representations. This makes the nature of the training data a significant factor in the efficacy of CLIP for downstream tasks. However, the lack of compositional diversity in contemporary image-text datasets limits the compositional reasoning ability of CLIP. We show that generating ``hard'' negative captions via in-context learning and synthesizing corresponding negative images with text-to-image generators offers a solution. We introduce a novel contrastive pre-training strategy that leverages these hard negative captions and images in an alternating fashion to train CLIP. We demonstrate that our method, named TripletCLIP, when applied to existing datasets such as CC3M and CC12M, enhances the compositional capabilities of CLIP, resulting in an absolute improvement of over 9% on the SugarCrepe benchmark on an equal computational budget, as well as improvements in zero-shot image classification and image retrieval. Our code, models, and data are available at: https://tripletclip.github.io
Authors: Jinghan Zhang, Henry Xie, Xinhao Zhang, Kunpeng Liu
Abstract: In the financial field, precise risk assessment tools are essential for decision-making. Recent studies have challenged the notion that traditional network loss functions like Mean Square Error (MSE) are adequate, especially under extreme risk conditions that can lead to significant losses during market upheavals. Transformers and Transformer-based models are now widely used in financial forecasting according to their outstanding performance in time-series-related predictions. However, these models typically lack sensitivity to extreme risks and often underestimate great financial losses. To address this problem, we introduce a novel loss function, the Loss-at-Risk, which incorporates Value at Risk (VaR) and Conditional Value at Risk (CVaR) into Transformer models. This integration allows Transformer models to recognize potential extreme losses and further improves their capability to handle high-stakes financial decisions. Moreover, we conduct a series of experiments with highly volatile financial datasets to demonstrate that our Loss-at-Risk function improves the Transformers' risk prediction and management capabilities without compromising their decision-making accuracy or efficiency. The results demonstrate that integrating risk-aware metrics during training enhances the Transformers' risk assessment capabilities while preserving their core strengths in decision-making and reasoning across diverse scenarios.
Authors: Jiawei Zhou, Amy Z. Chen, Darshi Shah, Laura Schwab Reese, Munmun De Choudhury
Abstract: Recent breakthroughs in large language models (LLMs) have generated both interest and concern about their potential adoption as accessible information sources or communication tools across different domains. In public health -- where stakes are high and impacts extend across populations -- adopting LLMs poses unique challenges that require thorough evaluation. However, structured approaches for assessing potential risks in public health remain under-explored. To address this gap, we conducted focus groups with health professionals and health issue experiencers to unpack their concerns, situated across three distinct and critical public health issues that demand high-quality information: vaccines, opioid use disorder, and intimate partner violence. We synthesize participants' perspectives into a risk taxonomy, distinguishing and contextualizing the potential harms LLMs may introduce when positioned alongside traditional health communication. This taxonomy highlights four dimensions of risk in individual behaviors, human-centered care, information ecosystem, and technology accountability. For each dimension, we discuss specific risks and example reflection questions to help practitioners adopt a risk-reflexive approach. This work offers a shared vocabulary and reflection tool for experts in both computing and public health to collaboratively anticipate, evaluate, and mitigate risks in deciding when to employ LLM capabilities (or not) and how to mitigate harm when they are used.
Authors: Jennifer Grannen, Siddharth Karamcheti, Suvir Mirchandani, Percy Liang, Dorsa Sadigh
Abstract: We introduce Vocal Sandbox, a framework for enabling seamless human-robot collaboration in situated environments. Systems in our framework are characterized by their ability to adapt and continually learn at multiple levels of abstraction from diverse teaching modalities such as spoken dialogue, object keypoints, and kinesthetic demonstrations. To enable such adaptation, we design lightweight and interpretable learning algorithms that allow users to build an understanding and co-adapt to a robot's capabilities in real-time, as they teach new behaviors. For example, after demonstrating a new low-level skill for "tracking around" an object, users are provided with trajectory visualizations of the robot's intended motion when asked to track a new object. Similarly, users teach high-level planning behaviors through spoken dialogue, using pretrained language models to synthesize behaviors such as "packing an object away" as compositions of low-level skills $-$ concepts that can be reused and built upon. We evaluate Vocal Sandbox in two settings: collaborative gift bag assembly and LEGO stop-motion animation. In the first setting, we run systematic ablations and user studies with 8 non-expert participants, highlighting the impact of multi-level teaching. Across 23 hours of total robot interaction time, users teach 17 new high-level behaviors with an average of 16 novel low-level skills, requiring 22.1% less active supervision compared to baselines and yielding more complex autonomous performance (+19.7%) with fewer failures (-67.1%). Qualitatively, users strongly prefer Vocal Sandbox systems due to their ease of use (+20.6%) and overall performance (+13.9%). Finally, we pair an experienced system-user with a robot to film a stop-motion animation; over two hours of continuous collaboration, the user teaches progressively more complex motion skills to shoot a 52 second (232 frame) movie.
Authors: Wanying Ding, Vinay K. Chaudhri, Naren Chittar, Krishna Konakanchi
Abstract: Knowledge Graphs have emerged as a compelling abstraction for capturing key relationship among the entities of interest to enterprises and for integrating data from heterogeneous sources. JPMorgan Chase (JPMC) is leading this trend by leveraging knowledge graphs across the organization for multiple mission critical applications such as risk assessment, fraud detection, investment advice, etc. A core problem in leveraging a knowledge graph is to link mentions (e.g., company names) that are encountered in textual sources to entities in the knowledge graph. Although several techniques exist for entity linking, they are tuned for entities that exist in Wikipedia, and fail to generalize for the entities that are of interest to an enterprise. In this paper, we propose a novel end-to-end neural entity linking model (JEL) that uses minimal context information and a margin loss to generate entity embeddings, and a Wide & Deep Learning model to match character and semantic information respectively. We show that JEL achieves the state-of-the-art performance to link mentions of company names in financial news with entities in our knowledge graph. We report on our efforts to deploy this model in the company-wide system to generate alerts in response to financial news. The methodology used for JEL is directly applicable and usable by other enterprises who need entity linking solutions for data that are unique to their respective situations.
Authors: Soroush Nasiriany, Sean Kirmani, Tianli Ding, Laura Smith, Yuke Zhu, Danny Driess, Dorsa Sadigh, Ted Xiao
Abstract: We explore how intermediate policy representations can facilitate generalization by providing guidance on how to perform manipulation tasks. Existing representations such as language, goal images, and trajectory sketches have been shown to be helpful, but these representations either do not provide enough context or provide over-specified context that yields less robust policies. We propose conditioning policies on affordances, which capture the pose of the robot at key stages of the task. Affordances offer expressive yet lightweight abstractions, are easy for users to specify, and facilitate efficient learning by transferring knowledge from large internet datasets. Our method, RT-Affordance, is a hierarchical model that first proposes an affordance plan given the task language, and then conditions the policy on this affordance plan to perform manipulation. Our model can flexibly bridge heterogeneous sources of supervision including large web datasets and robot trajectories. We additionally train our model on cheap-to-collect in-domain affordance images, allowing us to learn new tasks without collecting any additional costly robot trajectories. We show on a diverse set of novel tasks how RT-Affordance exceeds the performance of existing methods by over 50%, and we empirically demonstrate that affordances are robust to novel settings. Videos available at https://snasiriany.me/rt-affordance
Authors: Yunkai Dang, Mengxi Gao, Yibo Yan, Xin Zou, Yanggan Gu, Aiwei Liu, Xuming Hu
Abstract: Ensuring that Multimodal Large Language Models (MLLMs) maintain consistency in their responses is essential for developing trustworthy multimodal intelligence. However, existing benchmarks include many samples where all MLLMs \textit{exhibit high response uncertainty when encountering misleading information}, requiring even 5-15 response attempts per sample to effectively assess uncertainty. Therefore, we propose a two-stage pipeline: first, we collect MLLMs' responses without misleading information, and then gather misleading ones via specific misleading instructions. By calculating the misleading rate, and capturing both correct-to-incorrect and incorrect-to-correct shifts between the two sets of responses, we can effectively metric the model's response uncertainty. Eventually, we establish a \textbf{\underline{M}}ultimodal \textbf{\underline{U}}ncertainty \textbf{\underline{B}}enchmark (\textbf{MUB}) that employs both explicit and implicit misleading instructions to comprehensively assess the vulnerability of MLLMs across diverse domains. Our experiments reveal that all open-source and close-source MLLMs are highly susceptible to misleading instructions, with an average misleading rate exceeding 86\%. To enhance the robustness of MLLMs, we further fine-tune all open-source MLLMs by incorporating explicit and implicit misleading data, which demonstrates a significant reduction in misleading rates. Our code is available at: \href{https://github.com/Yunkai696/MUB}{https://github.com/Yunkai696/MUB}
URLs: https://github.com/Yunkai696/MUB, https://github.com/Yunkai696/MUB
Authors: Sheshera Mysore, Garima Dhanania, Kishor Patil, Surya Kallumadi, Andrew McCallum, Hamed Zamani
Abstract: Personalized search represents a problem where retrieval models condition on historical user interaction data in order to improve retrieval results. However, personalization is commonly perceived as opaque and not amenable to control by users. Further, personalization necessarily limits the space of items that users are exposed to. Therefore, prior work notes a tension between personalization and users' ability for discovering novel items. While discovery of novel items in personalization setups may be resolved through search result diversification, these approaches do little to allow user control over personalization. Therefore, in this paper, we introduce an approach for controllable personalized search. Our model, CtrlCE presents a novel cross-encoder model augmented with an editable memory constructed from users historical items. Our proposed memory augmentation allows cross-encoder models to condition on large amounts of historical user data and supports interaction from users permitting control over personalization. Further, controllable personalization for search must account for queries which don't require personalization, and in turn user control. For this, we introduce a calibrated mixing model which determines when personalization is necessary. This allows system designers using CtrlCE to only obtain user input for control when necessary. In multiple datasets of personalized search, we show CtrlCE to result in effective personalization as well as fulfill various key goals for controllable personalized search.
Authors: Zhibin Wen, Bin Li
Abstract: The goal of Multilingual Visual Answer Localization (MVAL) is to locate a video segment that answers a given multilingual question. Existing methods either focus solely on visual modality or integrate visual and subtitle modalities. However, these methods neglect the audio modality in videos, consequently leading to incomplete input information and poor performance in the MVAL task. In this paper, we propose a unified Audio-Visual-Textual Span Localization (AVTSL) method that incorporates audio modality to augment both visual and textual representations for the MVAL task. Specifically, we integrate features from three modalities and develop three predictors, each tailored to the unique contributions of the fused modalities: an audio-visual predictor, a visual predictor, and a textual predictor. Each predictor generates predictions based on its respective modality. To maintain consistency across the predicted results, we introduce an Audio-Visual-Textual Consistency module. This module utilizes a Dynamic Triangular Loss (DTL) function, allowing each modality's predictor to dynamically learn from the others. This collaborative learning ensures that the model generates consistent and comprehensive answers. Extensive experiments show that our proposed method outperforms several state-of-the-art (SOTA) methods, which demonstrates the effectiveness of the audio modality.
Authors: Zhan Li, Yongtao Wu, Yihang Chen, Francesco Tonin, Elias Abad Rocamora, Volkan Cevher
Abstract: Large vision-language models (VLLMs) exhibit promising capabilities for processing multi-modal tasks across various application scenarios. However, their emergence also raises significant data security concerns, given the potential inclusion of sensitive information, such as private photos and medical records, in their training datasets. Detecting inappropriately used data in VLLMs remains a critical and unresolved issue, mainly due to the lack of standardized datasets and suitable methodologies. In this study, we introduce the first membership inference attack (MIA) benchmark tailored for various VLLMs to facilitate training data detection. Then, we propose a novel MIA pipeline specifically designed for token-level image detection. Lastly, we present a new metric called MaxR\'enyi-K%, which is based on the confidence of the model output and applies to both text and image data. We believe that our work can deepen the understanding and methodology of MIAs in the context of VLLMs. Our code and datasets are available at https://github.com/LIONS-EPFL/VL-MIA.
Authors: Yuankai Fan, Tonghui Ren, Can Huang, Zhenying He, X. Sean Wang
Abstract: Natural Language Interfaces for Databases empower non-technical users to interact with data using natural language (NL). Advanced approaches, utilizing either neural sequence-to-sequence or more recent sophisticated large-scale language models, typically implement NL to SQL (NL2SQL) translation in an end-to-end fashion. However, like humans, these end-to-end translation models may not always generate the best SQL output on their first try. In this paper, we propose CycleSQL, an iterative framework designed for end-to-end translation models to autonomously generate the best output through self-evaluation. The main idea of CycleSQL is to introduce data-grounded NL explanations of query results as self-provided feedback, and use the feedback to validate the correctness of the translation iteratively, hence improving the overall translation accuracy. Extensive experiments, including quantitative and qualitative evaluations, are conducted to study CycleSQL by applying it to seven existing translation models on five widely used benchmarks. The results show that 1) the feedback loop introduced in CycleSQL can consistently improve the performance of existing models, and in particular, by applying CycleSQL to RESDSQL, obtains a translation accuracy of 82.0% (+2.6%) on the validation set, and 81.6% (+3.2%) on the test set of Spider benchmark; 2) the generated NL explanations can also provide insightful information for users, aiding in the comprehension of translation results and consequently enhancing the interpretability of NL2SQL translation.
Authors: Ying Zhou, Xinyao Wang, Yulei Niu, Yaojie Shen, Lexin Tang, Fan Chen, Ben He, Le Sun, Longyin Wen
Abstract: Recent advancements in large language models (LLMs) have significantly enhanced their knowledge and generative capabilities, leading to a surge of interest in leveraging LLMs for high-quality data synthesis. However, synthetic data generation via prompting LLMs remains challenging due to LLMs' limited understanding of target data distributions and the complexity of prompt engineering, especially for structured formatted data. To address these issues, we introduce DiffLM, a controllable data synthesis framework based on variational autoencoder (VAE), which further (1) leverages diffusion models to reserve more information of original distribution and format structure in the learned latent distribution and (2) decouples the learning of target distribution knowledge from the LLM's generative objectives via a plug-and-play latent feature injection module. As we observed significant discrepancies between the VAE's latent representations and the real data distribution, the latent diffusion module is introduced into our framework to learn a fully expressive latent distribution. Evaluations on seven real-world datasets with structured formatted data (i.e., Tabular, Code and Tool data) demonstrate that DiffLM generates high-quality data, with performance on downstream tasks surpassing that of real data by 2-7 percent in certain cases. The data and code will be publicly available upon completion of internal review.
Authors: Ziliang Gan, Yu Lu, Dong Zhang, Haohan Li, Che Liu, Jian Liu, Ji Liu, Haipang Wu, Chaoyou Fu, Zenglin Xu, Rongjunchen Zhang, Yong Dai
Abstract: In recent years, multimodal benchmarks for general domains have guided the rapid development of multimodal models on general tasks. However, the financial field has its peculiarities. It features unique graphical images (e.g., candlestick charts, technical indicator charts) and possesses a wealth of specialized financial knowledge (e.g., futures, turnover rate). Therefore, benchmarks from general fields often fail to measure the performance of multimodal models in the financial domain, and thus cannot effectively guide the rapid development of large financial models. To promote the development of large financial multimodal models, we propose MME-Finance, an bilingual open-ended and practical usage-oriented Visual Question Answering (VQA) benchmark. The characteristics of our benchmark are finance and expertise, which include constructing charts that reflect the actual usage needs of users (e.g., computer screenshots and mobile photography), creating questions according to the preferences in financial domain inquiries, and annotating questions by experts with 10+ years of experience in the financial industry. Additionally, we have developed a custom-designed financial evaluation system in which visual information is first introduced in the multi-modal evaluation process. Extensive experimental evaluations of 19 mainstream MLLMs are conducted to test their perception, reasoning, and cognition capabilities. The results indicate that models performing well on general benchmarks cannot do well on MME-Finance; for instance, the top-performing open-source and closed-source models obtain 65.69 (Qwen2VL-72B) and 63.18 (GPT-4o), respectively. Their performance is particularly poor in categories most relevant to finance, such as candlestick charts and technical indicator charts. In addition, we propose a Chinese version, which helps compare performance of MLLMs under a Chinese context.
Authors: Jo\~ao A. Leite, Olesya Razuvayevskaya, Kalina Bontcheva, Carolina Scarton
Abstract: Credibility signals represent a wide range of heuristics typically used by journalists and fact-checkers to assess the veracity of online content. Automating the extraction of credibility signals presents significant challenges due to the necessity of training high-accuracy, signal-specific extractors, coupled with the lack of sufficiently large annotated datasets. This paper introduces Pastel (Prompted weAk Supervision wiTh crEdibility signaLs), a weakly supervised approach that leverages large language models (LLMs) to extract credibility signals from web content, and subsequently combines them to predict the veracity of content without relying on human supervision. We validate our approach using four article-level misinformation detection datasets, demonstrating that Pastel outperforms zero-shot veracity detection by 38.3% and achieves 86.7% of the performance of the state-of-the-art system trained with human supervision. Moreover, in cross-domain settings where training and testing datasets originate from different domains, Pastel significantly outperforms the state-of-the-art supervised model by 63%. We further study the association between credibility signals and veracity, and perform an ablation study showing the impact of each signal on model performance. Our findings reveal that 12 out of the 19 proposed signals exhibit strong associations with veracity across all datasets, while some signals show domain-specific strengths.
Authors: Sheshera Mysore, Zhuoran Lu, Mengting Wan, Longqi Yang, Bahareh Sarrafzadeh, Steve Menezes, Tina Baghaee, Emmanuel Barajas Gonzalez, Jennifer Neville, Tara Safavi
Abstract: Powerful large language models have facilitated the development of writing assistants that promise to significantly improve the quality and efficiency of composition and communication. However, a barrier to effective assistance is the lack of personalization in LLM outputs to the author's communication style, specialized knowledge, and values. In this paper, we address this challenge by proposing Pearl, a LLM writing assistant personalized with a retriever that is trained to be generation-calibrated for personalization. Generation calibration ensures that our retriever selects historic user authored documents to augment an LLM prompt such that they are likely to help an LLM generation better adhere to a users' preferences. We propose two key novelties for training such a retriever: (1) A training data selection method that identifies user requests likely to benefit from personalization and documents that provide that benefit; and (2) A scale-calibrating KL-divergence objective that ensures that our retriever scores remain proportional to the downstream generation quality from using the document for personalized generation. In a series of holistic evaluations, we demonstrate the effectiveness of Pearl in generating long-form texts on multiple social media datasets. Finally, we demonstrate how a generation-calibrated retriever can double as a performance predictor -- detecting low quality retrieval, and improving potentially under-performing outputs via revision with LLMs.
Authors: Md Abrar Jahin, Subrata Talapatra
Abstract: This research delves into Musculoskeletal Disorder (MSD) risk factors, using a blend of Natural Language Processing (NLP) and mode-based ranking. The aim is to refine understanding, classification, and prioritization for focused prevention and treatment. Eight NLP models are evaluated, combining pre-trained transformers, cosine similarity, and distance metrics to categorize factors into personal, biomechanical, workplace, psychological, and organizational classes. BERT with cosine similarity achieves 28% accuracy; sentence transformer with Euclidean, Bray-Curtis, and Minkowski distances scores 100%. With 10-fold cross-validation, statistical tests ensure robust results. Survey data and mode-based ranking determine severity hierarchy, aligning with the literature. "Working posture" is the most severe, highlighting posture's role. Survey insights emphasize "Job insecurity," "Effort reward imbalance," and "Poor employee facility" as significant contributors. Rankings offer actionable insights for MSD prevention. The study suggests targeted interventions, workplace improvements, and future research directions. This integrated NLP and ranking approach enhances MSD comprehension and informs occupational health strategies.
Authors: Tianyu Zheng, Shuyue Guo, Xingwei Qu, Jiawei Guo, Xinrun Du, Qi Jia, Chenghua Lin, Wenhao Huang, Jie Fu, Ge Zhang
Abstract: In this paper, we introduce Kun, a novel approach for creating high-quality instruction-tuning datasets for large language models (LLMs) without relying on manual annotations. Adapting a self-training algorithm based on instruction back-translation and answer polishment, Kun leverages unlabelled data from diverse sources such as Wudao, Wanjuan, and SkyPile to generate a substantial dataset of over a million Chinese instructional data points. This approach significantly deviates from traditional methods by using a self-curation process to refine and select the most effective instruction-output pairs. Our experiments with the 6B-parameter Yi model across various benchmarks demonstrate Kun's robustness and scalability. Our method's core contributions lie in its algorithmic advancement, which enhances data retention and clarity, and its innovative data generation approach that substantially reduces the reliance on costly and time-consuming manual annotations. This methodology presents a scalable and efficient solution for improving the instruction-following capabilities of LLMs, with significant implications for their application across diverse fields. The code and dataset can be found at https://github.com/Zheng0428/COIG-Kun
Authors: Shengzhi Li, Rongyu Lin, Shichao Pei
Abstract: Multi-modal large language models (MLLMs) are expected to support multi-turn queries of interchanging image and text modalities in production. However, the current MLLMs trained with visual-question-answering (VQA) datasets could suffer from degradation, as VQA datasets lack the diversity and complexity of the original text instruction datasets with which the underlying language model was trained. To address this degradation, we first collect a lightweight, 5k-sample VQA preference dataset where answers were annotated by Gemini for five quality metrics in a granular fashion and investigate standard Supervised Fine-tuning, rejection sampling, Direct Preference Optimization (DPO) and SteerLM algorithms. Our findings indicate that with DPO, we can surpass the instruction-following capabilities of the language model, achieving a 6.73 score on MT-Bench, compared to Vicuna's 6.57 and LLaVA's 5.99. This enhancement in textual instruction-following capability correlates with boosted visual instruction performance (+4.9\% on MM-Vet, +6\% on LLaVA-Bench), with minimal alignment tax on visual knowledge benchmarks compared to the previous RLHF approach. In conclusion, we propose a distillation-based multi-modal alignment model with fine-grained annotations on a small dataset that restores and boosts MLLM's language capability after visual instruction tuning.
Authors: Ge Bai, Jie Liu, Xingyuan Bu, Yancheng He, Jiaheng Liu, Zhanhui Zhou, Zhuoran Lin, Wenbo Su, Tiezheng Ge, Bo Zheng, Wanli Ouyang
Abstract: The advent of Large Language Models (LLMs) has drastically enhanced dialogue systems. However, comprehensively evaluating the dialogue abilities of LLMs remains a challenge. Previous benchmarks have primarily focused on single-turn dialogues or provided coarse-grained and incomplete assessments of multi-turn dialogues, overlooking the complexity and fine-grained nuances of real-life dialogues. To address this issue, we introduce MT-Bench-101, specifically designed to evaluate the fine-grained abilities of LLMs in multi-turn dialogues. By conducting a detailed analysis of real multi-turn dialogue data, we construct a three-tier hierarchical ability taxonomy comprising 4208 turns across 1388 multi-turn dialogues in 13 distinct tasks. We then evaluate 21 popular LLMs based on MT-Bench-101, conducting comprehensive analyses from both ability and task perspectives and observing differing trends in LLMs performance across dialogue turns within various tasks. Further analysis indicates that neither utilizing common alignment techniques nor chat-specific designs has led to obvious enhancements in the multi-turn abilities of LLMs. Extensive case studies suggest that our designed tasks accurately assess the corresponding multi-turn abilities. The data and code are available at \url{https://github.com/mtbench101/mt-bench-101}.
Authors: Zhexin Zhang, Yida Lu, Jingyuan Ma, Di Zhang, Rui Li, Pei Ke, Hao Sun, Lei Sha, Zhifang Sui, Hongning Wang, Minlie Huang
Abstract: The safety of Large Language Models (LLMs) has gained increasing attention in recent years, but there still lacks a comprehensive approach for detecting safety issues within LLMs' responses in an aligned, customizable and explainable manner. In this paper, we propose ShieldLM, an LLM-based safety detector, which aligns with common safety standards, supports customizable detection rules, and provides explanations for its decisions. To train ShieldLM, we compile a large bilingual dataset comprising 14,387 query-response pairs, annotating the safety of responses based on various safety standards. Through extensive experiments, we demonstrate that ShieldLM surpasses strong baselines across four test sets, showcasing remarkable customizability and explainability. Besides performing well on standard detection datasets, ShieldLM has also been shown to be effective as a safety evaluator for advanced LLMs. ShieldLM is released at \url{https://github.com/thu-coai/ShieldLM} to support accurate and explainable safety detection under various safety standards.
Authors: Hui Huang, Yingqi Qu, Xingyuan Bu, Hongli Zhou, Jing Liu, Muyun Yang, Bing Xu, Tiejun Zhao
Abstract: Recently, there has been a growing trend of utilizing Large Language Model (LLM) to evaluate the quality of other LLMs. Many studies have employed proprietary close-sourced models, especially GPT-4, as the evaluator. Alternatively, other works have fine-tuned judge models based on open-source LLMs as the evaluator. While the fine-tuned judge models are claimed to achieve comparable evaluation capability with GPT-4, in this work, we conduct an empirical study of judge models. Our findings indicate that although the fine-tuned judge models achieve high performance on in-domain test sets, even surpassing GPT-4, they underperform GPT-4 across several dimensions, including generalizability, fairness, aspect-specific evaluation, and scalability. We also reveal that the fine-tuned judge model inherently operates as a task-specific classifier, consequently imposing the limitations. Finally, we introduce a integrated method, leveraging GPT-4 to compensate for the limitations and improve the fine-tuned judges. Experiment results show our method achieves accuracy on par with GPT-4 with only 50% of the API expense.
Authors: Emad A. Alghamdi, Reem I. Masoud, Deema Alnuhait, Afnan Y. Alomairi, Ahmed Ashraf, Mohamed Zaytoon
Abstract: The swift progress and widespread acceptance of artificial intelligence (AI) systems highlight a pressing requirement to comprehend both the capabilities and potential risks associated with AI. Given the linguistic complexity, cultural richness, and underrepresented status of Arabic in AI research, there is a pressing need to focus on Large Language Models (LLMs) performance and safety for Arabic-related tasks. Despite some progress in their development, there is a lack of comprehensive trustworthiness evaluation benchmarks, which presents a major challenge in accurately assessing and improving the safety of LLMs when prompted in Arabic. In this paper, we introduce AraTrust, the first comprehensive trustworthiness benchmark for LLMs in Arabic. AraTrust comprises 522 human-written multiple-choice questions addressing diverse dimensions related to truthfulness, ethics, safety, physical health, mental health, unfairness, illegal activities, privacy, and offensive language. We evaluated a set of LLMs against our benchmark to assess their trustworthiness. GPT-4 was the most trustworthy LLM, while open-source models, particularly AceGPT 7B and Jais 13B, struggled to achieve a score of 60% in our benchmark.
Authors: Jiaxuan Wu, Zhengxian Wu, Yiming Xue, Juan Wen, Wanli Peng
Abstract: Recent advances in large language models (LLMs) have blurred the boundary of high-quality text generation between humans and machines, which is favorable for generative text steganography. While, current advanced steganographic mapping is not suitable for LLMs since most users are restricted to accessing only the black-box API or user interface of the LLMs, thereby lacking access to the training vocabulary and its sampling probabilities. In this paper, we explore a black-box generative text steganographic method based on the user interfaces of large language models, which is called LLM-Stega. The main goal of LLM-Stega is that the secure covert communication between Alice (sender) and Bob (receiver) is conducted by using the user interfaces of LLMs. Specifically, We first construct a keyword set and design a new encrypted steganographic mapping to embed secret messages. Furthermore, to guarantee accurate extraction of secret messages and rich semantics of generated stego texts, an optimization mechanism based on reject sampling is proposed. Comprehensive experiments demonstrate that the proposed LLM-Stega outperforms current state-of-the-art methods.
Authors: Seongyun Lee, Sue Hyun Park, Seungone Kim, Minjoon Seo
Abstract: Although humans inherently have diverse values, current large language model (LLM) alignment methods often assume that aligning LLMs with the general public's preferences is optimal. A major challenge in adopting a more individualized approach to LLM alignment is its lack of scalability, as it involves repeatedly acquiring preference data and training new reward models and LLMs for each individual's preferences. To address these challenges, we propose a new paradigm where users specify what they value most within the system message, steering the LLM's generation behavior to better align with the user's intentions. However, a naive application of such an approach is non-trivial since LLMs are typically trained on a uniform system message (e.g., "You are a helpful assistant") which limits their ability to generalize to diverse, unseen system messages. To improve this generalization, we create the Multifaceted Collection, a preference dataset with 192k combinations of values beyond generic helpfulness and harmlessness, spanning 65k user instructions. Using this dataset, we train a 7B LLM called Janus and test it on 921 prompts from 5 benchmarks (AlpacaEval 2.0, FLASK, Koala, MT-Bench, and Self-Instruct) by adding various unseen system messages that reflect user preferences. Janus achieves tie+win rate of 75.2%, 72.4%, and 66.4% against Mistral 7B Instruct v0.2, GPT-3.5 Turbo, and GPT-4, respectively. Unexpectedly, on three benchmarks focused on response helpfulness (AlpacaEval 2.0, MT-Bench, Arena Hard Auto v0.1), Janus also outperforms LLaMA 3 8B Instruct by a +4.0%, +0.1%, +3.0% margin, underscoring that training with a vast array of system messages could also enhance alignment to the general public's preference as well. Our code, dataset, benchmark, and models are available at https://github.com/kaistAI/Janus.
Authors: Jinqi Luo, Tianjiao Ding, Kwan Ho Ryan Chan, Darshan Thaker, Aditya Chattopadhyay, Chris Callison-Burch, Ren\'e Vidal
Abstract: Large Language Models (LLMs) are being used for a wide variety of tasks. While they are capable of generating human-like responses, they can also produce undesirable output including potentially harmful information, racist or sexist language, and hallucinations. Alignment methods are designed to reduce such undesirable outputs via techniques such as fine-tuning, prompt engineering, and representation engineering. However, existing methods face several challenges: some require costly fine-tuning for every alignment task; some do not adequately remove undesirable concepts, failing alignment; some remove benign concepts, lowering the linguistic capabilities of LLMs. To address these issues, we propose Parsimonious Concept Engineering (PaCE), a novel activation engineering framework for alignment. First, to sufficiently model the concepts, we construct a large-scale concept dictionary in the activation space, in which each atom corresponds to a semantic concept. Given any alignment task, we instruct a concept partitioner to efficiently annotate the concepts as benign or undesirable. Then, at inference time, we decompose the LLM activations along the concept dictionary via sparse coding, to accurately represent the activations as linear combinations of benign and undesirable components. By removing the latter ones from the activations, we reorient the behavior of the LLM towards the alignment goal. We conduct experiments on tasks such as response detoxification, faithfulness enhancement, and sentiment revising, and show that PaCE achieves state-of-the-art alignment performance while maintaining linguistic capabilities.
Authors: Joao Monteiro, Pierre-Andre Noel, Etienne Marcotte, Sai Rajeswar, Valentina Zantedeschi, David Vazquez, Nicolas Chapados, Christopher Pal, Perouz Taslakian
Abstract: Large Language Models (LLMs) are trained on vast amounts of data, most of which is automatically scraped from the internet. This data includes encyclopedic documents that harbor a vast amount of general knowledge (e.g., Wikipedia) but also potentially overlap with benchmark datasets used for evaluating LLMs. Consequently, evaluating models on test splits that might have leaked into the training set is prone to misleading conclusions. To foster sound evaluation of language models, we introduce a new test dataset named RepLiQA, suited for question-answering and topic retrieval tasks. RepLiQA is a collection of five splits of test sets, four of which have not been released to the internet or exposed to LLM APIs prior to this publication. Each sample in RepLiQA comprises (1) a reference document crafted by a human annotator and depicting an imaginary scenario (e.g., a news article) absent from the internet; (2) a question about the document's topic; (3) a ground-truth answer derived directly from the information in the document; and (4) the paragraph extracted from the reference document containing the answer. As such, accurate answers can only be generated if a model can find relevant content within the provided document. We run a large-scale benchmark comprising several state-of-the-art LLMs to uncover differences in performance across models of various types and sizes in a context-conditional language modeling setting. Released splits of RepLiQA can be found here: https://huggingface.co/datasets/ServiceNow/repliqa.
Authors: Yixia Li, Boya Xiong, Guanhua Chen, Yun Chen
Abstract: Out-of-distribution (OOD) detection is crucial for the safe deployment of neural networks. Existing CLIP-based approaches perform OOD detection by devising novel scoring functions or sophisticated fine-tuning methods. In this work, we propose SeTAR, a novel, training-free OOD detection method that leverages selective low-rank approximation of weight matrices in vision-language and vision-only models. SeTAR enhances OOD detection via post-hoc modification of the model's weight matrices using a simple greedy search algorithm. Based on SeTAR, we further propose SeTAR+FT, a fine-tuning extension optimizing model performance for OOD detection tasks. Extensive evaluations on ImageNet1K and Pascal-VOC benchmarks show SeTAR's superior performance, reducing the relatively false positive rate by up to 18.95% and 36.80% compared to zero-shot and fine-tuning baselines. Ablation studies further validate SeTAR's effectiveness, robustness, and generalizability across different model backbones. Our work offers a scalable, efficient solution for OOD detection, setting a new state-of-the-art in this area.
Authors: Xiangfeng Wang, Zaiyi Chen, Zheyong Xie, Tong Xu, Yongyi He, Enhong Chen
Abstract: With the rising popularity of Transformer-based large language models (LLMs), reducing their high inference costs has become a significant research focus. One effective approach is to compress the long input contexts. Existing methods typically leverage the self-attention mechanism of the LLM itself for context compression. While these methods have achieved notable results, the compression process still involves quadratic time complexity, which limits their applicability. To mitigate this limitation, we propose the In-Context Former (IC-Former). Unlike previous methods, IC-Former does not depend on the target LLMs. Instead, it leverages the cross-attention mechanism and a small number of learnable digest tokens to directly condense information from the contextual word embeddings. This approach significantly reduces inference time, which achieves linear growth in time complexity within the compression range. Experimental results indicate that our method requires only 1/32 of the floating-point operations of the baseline during compression and improves processing speed by 68 to 112 times while achieving over 90% of the baseline performance on evaluation metrics. Overall, our model effectively reduces compression costs and makes real-time compression scenarios feasible.
Authors: Shilong Li, Yancheng He, Hangyu Guo, Xingyuan Bu, Ge Bai, Jie Liu, Jiaheng Liu, Xingwei Qu, Yangguang Li, Wanli Ouyang, Wenbo Su, Bo Zheng
Abstract: Long-context capabilities are essential for large language models (LLMs) to tackle complex and long-input tasks. Despite numerous efforts made to optimize LLMs for long contexts, challenges persist in robustly processing long inputs. In this paper, we introduce GraphReader, a graph-based agent system designed to handle long texts by structuring them into a graph and employing an agent to explore this graph autonomously. Upon receiving a question, the agent first undertakes a step-by-step analysis and devises a rational plan. It then invokes a set of predefined functions to read node content and neighbors, facilitating a coarse-to-fine exploration of the graph. Throughout the exploration, the agent continuously records new insights and reflects on current circumstances to optimize the process until it has gathered sufficient information to generate an answer. Experimental results on the LV-Eval dataset reveal that GraphReader, using a 4k context window, consistently outperforms GPT-4-128k across context lengths from 16k to 256k by a large margin. Additionally, our approach demonstrates superior performance on four challenging single-hop and multi-hop benchmarks.
Authors: Dhananjay Ram, Aditya Rawal, Momchil Hardalov, Nikolaos Pappas, Sheng Zha
Abstract: Training with mixed data distributions is a common and important part of creating multi-task and instruction-following models. The diversity of the data distributions and cost of joint training makes the optimization procedure extremely challenging. Data mixing methods partially address this problem, albeit having a sub-optimal performance across data sources and require multiple expensive training runs. In this paper, we propose a simple and efficient alternative for better optimization of the data sources by combining models individually trained on each data source with the base model using basic element-wise vector operations. The resulting model, namely Distribution Edited Model (DEM), is 11x cheaper than standard data mixing and outperforms strong baselines on a variety of benchmarks, yielding upto 6.2% improvement on MMLU, 11.5% on BBH, 16.1% on DROP, 6% on MathQA, and 9.3% on HELM with models of size 3B to 13B. Notably, DEM does not require full re-training when modifying a single data-source, thus making it very flexible and scalable for training with diverse data sources.
Authors: Fateme Hashemi Chaleshtori, Atreya Ghosal, Alexander Gill, Purbid Bambroo, Ana Marasovi\'c
Abstract: Is explainability a false promise? This debate has emerged from the insufficient evidence that explanations help people in situations they are introduced for. More human-centered, application-grounded evaluations of explanations are needed to settle this. Yet, with no established guidelines for such studies in NLP, researchers accustomed to standardized proxy evaluations must discover appropriate measurements, tasks, datasets, and sensible models for human-AI teams in their studies. To aid with this, we first review existing metrics suitable for application-grounded evaluation. We then establish criteria to select appropriate datasets, and using them, we find that only 4 out of over 50 datasets available for explainability research in NLP meet them. We then demonstrate the importance of reassessing the state of the art to form and study human-AI teams: teaming people with models for certain tasks might only now start to make sense, and for others, it remains unsound. Finally, we present the exemplar studies of human-AI decision-making for one of the identified tasks -- verifying the correctness of a legal claim given a contract. Our results show that providing AI predictions, with or without explanations, does not cause decision makers to speed up their work without compromising performance. We argue for revisiting the setup of human-AI teams and improving automatic deferral of instances to AI, where explanations could play a useful role.
Authors: Juan Luis Gastaldi, John Terilla, Luca Malagutti, Brian DuSell, Tim Vieira, Ryan Cotterell
Abstract: Tokenization - the practice of converting strings of characters from an alphabet into sequences of tokens over a vocabulary - is a critical step in the NLP pipeline. The use of token representations is widely credited with increased model performance but is also the source of many undesirable behaviors, such as spurious ambiguity or inconsistency. Despite its recognized importance as a standard representation method in NLP, the theoretical underpinnings of tokenization are not yet fully understood. In particular, the impact of tokenization on statistical estimation has been investigated mostly through empirical means. The present paper contributes to addressing this theoretical gap by proposing a unified formal framework for representing and analyzing tokenizer models. Based on the category of stochastic maps, this framework enables us to establish general conditions for a principled use of tokenizers, and most importantly, the necessary and sufficient conditions for a tokenizer model to preserve the consistency of statistical estimators. Additionally, we discuss statistical and computational concerns crucial for designing and implementing tokenizer models, such as inconsistency, ambiguity, tractability, and boundedness. The framework and results advanced in this paper contribute to building robust theoretical foundations for representations in neural language modeling that can inform future empirical research.
Authors: Timothy Nguyen
Abstract: Transformer based large-language models (LLMs) display extreme proficiency with language yet a precise understanding of how they work remains elusive. One way of demystifying transformer predictions would be to describe how they depend on their context in terms of simple template functions. This paper takes a first step in this direction by considering families of functions (i.e. rules) formed out of simple N-gram based statistics of the training data. By studying how well these rulesets approximate transformer predictions, we obtain a variety of novel discoveries: a simple method to detect overfitting during training without using a holdout set, a quantitative measure of how transformers progress from learning simple to more complex statistical rules over the course of training, a model-variance criterion governing when transformer predictions tend to be described by N-gram rules, and insights into how well transformers can be approximated by N-gram rulesets in the limit where these rulesets become increasingly complex. In this latter direction, we find that for 79% and 68% of LLM next-token distributions on TinyStories and Wikipedia, respectively, their top-1 predictions agree with those provided by our N-gram rulesets.
Authors: Andrew Zhu, Liam Dugan, Chris Callison-Burch
Abstract: Recently, there has been increasing interest in using Large Language Models (LLMs) to construct complex multi-agent systems to perform tasks such as compiling literature reviews, drafting consumer reports, and planning vacations. Many tools and libraries exist for helping create such systems, however none support recursive multi-agent systems -- where the models themselves flexibly decide when to delegate tasks and how to organize their delegation structure. In this work, we introduce ReDel: a toolkit for recursive multi-agent systems that supports custom tool-use, delegation schemes, event-based logging, and interactive replay in an easy-to-use web interface. We show that, using ReDel, we are able to easily identify potential areas of improvements through the visualization and debugging tools. Our code, documentation, and PyPI package are open-source and free to use under the MIT license at https://github.com/zhudotexe/redel.
Authors: Niklas Wretblad, Oskar Holmstr\"om, Erik Larsson, Axel Wiks\"ater, Oscar S\"oderlund, Hjalmar \"Ohman, Ture Pont\'en, Martin Forsberg, Martin S\"orme, Fredrik Heintz
Abstract: Relational databases often suffer from uninformative descriptors of table contents, such as ambiguous columns and hard-to-interpret values, impacting both human users and text-to-SQL models. In this paper, we explore the use of large language models (LLMs) to automatically generate detailed natural language descriptions for SQL database columns, aiming to improve text-to-SQL performance and automate metadata creation. We create a dataset of gold column descriptions based on the BIRD-Bench benchmark, manually refining its column descriptions and creating a taxonomy for categorizing column difficulty. We then evaluate several different LLMs in generating column descriptions across the columns and different difficulties in the dataset, finding that models unsurprisingly struggle with columns that exhibit inherent ambiguity, highlighting the need for manual expert input. We also find that incorporating such generated column descriptions consistently enhances text-to-SQL model performance, particularly for larger models like GPT-4o, Qwen2 72B and Mixtral 22Bx8. Notably, Qwen2-generated descriptions, containing by annotators deemed superfluous information, outperform manually curated gold descriptions, suggesting that models benefit from more detailed metadata than humans expect. Future work will investigate the specific features of these high-performing descriptions and explore other types of metadata, such as numerical reasoning and synonyms, to further improve text-to-SQL systems. The dataset, annotations and code will all be made available.
Authors: Yushi Yang, Andrew M. Bean, Robert McCraith, Adam Mahdi
Abstract: Fine-tuning Large Language Models (LLMs) incurs considerable training costs, driving the need for data-efficient training with optimised data ordering. Human-inspired strategies offer a solution by organising data based on human learning practices. This study evaluates the fine-tuning efficiency of five human-inspired strategies across four language models, three datasets, and both human- and LLM-labelled data in the context of medical question answering. These strategies achieve the best accuracy gain of 1.81% and an average gain of 1.02% across datasets, with interleaved strategies delivering the best average results. However, the best strategy varies across model-dataset combinations, limiting the generalisability of the effects of any single strategy. Additionally, LLM-defined question difficulty outperforms human-defined labels in curriculum-based learning, showing the potential of model-generated data as a cost-effective alternative for optimising fine-tuning.
Authors: Yunmeng Li, Jun Suzuki, Makoto Morishita, Kaori Abe, Kentaro Inui
Abstract: Machine translation models are still inappropriate for translating chats, despite the popularity of translation software and plug-in applications. The complexity of dialogues poses significant challenges and can hinder crosslingual communication. Instead of pursuing a flawless translation system, a more practical approach would be to issue warning messages about potential mistranslations to reduce confusion. However, it is still unclear how individuals perceive these warning messages and whether they benefit the crowd. This paper tackles to investigate this question and demonstrates the warning messages' contribution to making chat translation systems effective.
Authors: Kai Yao, Penglei Gao, Lichun Li, Yuan Zhao, Xiaofeng Wang, Wei Wang, Jianke Zhu
Abstract: Parameter-Efficient Fine-Tuning (PEFT) methods have gained significant popularity for adapting pre-trained Large Language Models (LLMs) to downstream tasks, primarily due to their potential to significantly reduce memory and computational overheads. However, a common limitation in most PEFT approaches is their application of a uniform architectural design across all layers. This uniformity involves identical trainable modules and ignores the varying importance of each layer, leading to sub-optimal fine-tuning results. To overcome the above limitation and obtain better performance, we develop a novel approach, Importance-aware Sparse Tuning (IST), to fully utilize the inherent sparsity and select the most important subset of full layers with effective layer-wise importance scoring. The proposed IST is a versatile and plug-and-play technique compatible with various PEFT methods that operate on a per-layer basis. By leveraging the estimated importance scores, IST dynamically updates these selected layers in PEFT modules, leading to reduced memory demands. We further provide theoretical proof of convergence and empirical evidence of superior performance to demonstrate the advantages of IST over uniform updating strategies. Extensive experiments on a range of LLMs, PEFTs, and downstream tasks substantiate the effectiveness of our proposed method, showcasing IST's capacity to enhance existing layer-based PEFT methods. Our code is available at https://github.com/Kaiseem/IST.
Authors: Virgile Rennard, Christos Xypolopoulos, Michalis Vazirgiannis
Abstract: Large language models (LLMs) inherit biases from their training data and alignment processes, influencing their responses in subtle ways. While many studies have examined these biases, little work has explored their robustness during interactions. In this paper, we introduce a novel approach where two instances of an LLM engage in self-debate, arguing opposing viewpoints to persuade a neutral version of the model. Through this, we evaluate how firmly biases hold and whether models are susceptible to reinforcing misinformation or shifting to harmful viewpoints. Our experiments span multiple LLMs of varying sizes, origins, and languages, providing deeper insights into bias persistence and flexibility across linguistic and cultural contexts.
Authors: Samar M. Magdy, Fakhraddin Alwajih, Sang Yun Kwon, Reem Abdel-Salam, Muhammad Abdul-Mageed
Abstract: Writing has long been considered a hallmark of human intelligence and remains a pinnacle task for artificial intelligence (AI) due to the intricate cognitive processes involved. Recently, rapid advancements in generative AI, particularly through the development of Large Language Models (LLMs), have significantly transformed the landscape of writing assistance. However, underrepresented languages like Arabic encounter significant challenges in the development of advanced AI writing tools, largely due to the limited availability of data. This scarcity constrains the training of effective models, impeding the creation of sophisticated writing assistance technologies. To address these issues, we present Gazelle, a comprehensive dataset for Arabic writing assistance. In addition, we offer an evaluation framework designed to enhance Arabic writing assistance tools. Our human evaluation of leading LLMs, including GPT-4, GPT-4o, Cohere Command R+, and Gemini 1.5 Pro, highlights their respective strengths and limitations in addressing the challenges of Arabic writing. Our findings underscore the need for continuous model training and dataset enrichment to manage the complexities of Arabic language processing, paving the way for more effective AI-powered Arabic writing tools.
Authors: Sergio Burdisso, Srikanth Madikeri, Petr Motlicek
Abstract: Efficiently deriving structured workflows from unannotated dialogs remains an underexplored and formidable challenge in computational linguistics. Automating this process could significantly accelerate the manual design of workflows in new domains and enable the grounding of large language models in domain-specific flowcharts, enhancing transparency and controllability. In this paper, we introduce Dialog2Flow (D2F) embeddings, which differ from conventional sentence embeddings by mapping utterances to a latent space where they are grouped according to their communicative and informative functions (i.e., the actions they represent). D2F allows for modeling dialogs as continuous trajectories in a latent space with distinct action-related regions. By clustering D2F embeddings, the latent space is quantized, and dialogs can be converted into sequences of region/action IDs, facilitating the extraction of the underlying workflow. To pre-train D2F, we build a comprehensive dataset by unifying twenty task-oriented dialog datasets with normalized per-turn action annotations. We also introduce a novel soft contrastive loss that leverages the semantic information of these actions to guide the representation learning process, showing superior performance compared to standard supervised contrastive loss. Evaluation against various sentence embeddings, including dialog-specific ones, demonstrates that D2F yields superior qualitative and quantitative results across diverse domains.
Authors: E. Zhixuan Zeng, Yuhao Chen, Alexander Wong
Abstract: Recent advances in image generation have made diffusion models powerful tools for creating high-quality images. However, their iterative denoising process makes understanding and interpreting their semantic latent spaces more challenging than other generative models, such as GANs. Recent methods have attempted to address this issue by identifying semantically meaningful directions within the latent space. However, they often need manual interpretation or are limited in the number of vectors that can be trained, restricting their scope and utility. This paper proposes a novel framework for unsupervised exploration of diffusion latent spaces. We directly leverage natural language prompts and image captions to map latent directions. This method allows for the automatic understanding of hidden features and supports a broader range of analysis without the need to train specific vectors. Our method provides a more scalable and interpretable understanding of the semantic knowledge encoded within diffusion models, facilitating comprehensive analysis of latent biases and the nuanced representations these models learn. Experimental results show that our framework can uncover hidden patterns and associations in various domains, offering new insights into the interpretability of diffusion model latent spaces.
Authors: Yujin Potter, Shiyang Lai, Junsol Kim, James Evans, Dawn Song
Abstract: How could LLMs influence our democracy? We investigate LLMs' political leanings and the potential influence of LLMs on voters by conducting multiple experiments in a U.S. presidential election context. Through a voting simulation, we first demonstrate 18 open- and closed-weight LLMs' political preference for a Democratic nominee over a Republican nominee. We show how this leaning towards the Democratic nominee becomes more pronounced in instruction-tuned models compared to their base versions by analyzing their responses to candidate-policy related questions. We further explore the potential impact of LLMs on voter choice by conducting an experiment with 935 U.S. registered voters. During the experiments, participants interacted with LLMs (Claude-3, Llama-3, and GPT-4) over five exchanges. The experiment results show a shift in voter choices towards the Democratic nominee following LLM interaction, widening the voting margin from 0.7% to 4.6%, even though LLMs were not asked to persuade users to support the Democratic nominee during the discourse. This effect is larger than many previous studies on the persuasiveness of political campaigns, which have shown minimal effects in presidential elections. Many users also expressed a desire for further political interaction with LLMs. Which aspects of LLM interactions drove these shifts in voter choice requires further study. Lastly, we explore how a safety method can make LLMs more politically neutral, while leaving some open questions.
Authors: Sam Blouir, Jimmy T. H. Smith, Antonios Anastasopoulos, Amarda Shehu
Abstract: Efficient state space models (SSMs), such as linear recurrent neural networks and linear attention variants, offer computational advantages over Transformers but struggle with tasks requiring long-range in-context retrieval-like text copying, associative recall, and question answering over long contexts. Previous efforts to address these challenges have focused on architectural modifications, often reintroducing computational inefficiencies. In this paper, we propose a novel training procedure, Birdie, that significantly enhances the in-context retrieval capabilities of SSMs without altering their architecture. Our approach combines bidirectional input processing with dynamic mixtures of specialized pre-training objectives, optimized via reinforcement learning. We introduce a new bidirectional SSM architecture that seamlessly transitions from bidirectional context processing to causal generation. Experimental evaluations demonstrate that Birdie markedly improves performance on retrieval-intensive tasks such as multi-number phone book lookup, long paragraph question-answering, and infilling. This narrows the performance gap with Transformers, while retaining computational efficiency. Our findings highlight the importance of training procedures in leveraging the fixed-state capacity of SSMs, offering a new direction to advance their capabilities. All code and pre-trained models are available at https://www.github.com/samblouir/birdie, with support for JAX and PyTorch.
Authors: Jiankun Wei, Abdulrahman Abdulrazzag, Tianchen Zhang, Adel Muursepp, Gururaj Saileshwar
Abstract: Speculative decoding in large language models (LLMs) accelerates token generation by speculatively predicting multiple tokens cheaply and verifying them in parallel, and has been widely deployed. In this paper, we provide the first study demonstrating the privacy risks of speculative decoding. We observe that input-dependent patterns of correct and incorrect predictions can be leaked out to an adversary monitoring token generation times and packet sizes, leading to privacy breaches. By observing the pattern of correctly and incorrectly speculated tokens, we show that a malicious adversary can fingerprint queries and learn private user inputs with more than $90\%$ accuracy across three different speculative decoding techniques - REST (almost $100\%$ accuracy), LADE (up to $92\%$ accuracy), and BiLD (up to $95\%$ accuracy). We show that an adversary can also leak out confidential intellectual property used to design these techniques, such as data from data-stores used for prediction (in REST) at a rate of more than $25$ tokens per second, or even hyper-parameters used for prediction (in LADE). We also discuss mitigation strategies, such as aggregating tokens across multiple iterations and padding packets with additional bytes, to avoid such privacy or confidentiality breaches.
Authors: Liat Bezalel, Eyal Orgad, Amir Globerson
Abstract: Large Language Models (LLMs) often struggle when prompted to generate content under specific constraints. However, in such cases it is often easy to check whether these constraints are satisfied or violated. Recent works have shown that LLMs can benefit from such ``corrective feedback''. Here we claim that this skill of LLMs can be significantly enhanced via training. We introduce an RL framework for teaching models to use such rewards, by simulating interaction sessions, and rewarding the model according to its ability to satisfy the constraints. We refer to our method as CORGI (Controlled Generation with RL for Guided Interaction), and evaluate it on a variety of controlled generation tasks using unlabeled training data. We find that CORGI consistently outperforms the baseline reinforcement learning method that does not incorporate conversational feedback. Furthermore, CORGI's interactive framework enables meta-learning, allowing the LLM to generalize better to guided interaction in new tasks. Our results clearly show that conversational optimization, when combined with reinforcement learning, significantly improves the effectiveness of LLMs in controlled generation contexts.
Authors: William F. Bradley
Abstract: As large language models (LLMs) become increasingly powerful, traditional evaluation metrics tend to saturate, making it challenging to distinguish between models based on their performance. We propose a general method to transform existing LLM evaluations into a series of progressively more difficult tasks. These enhanced evaluations emphasize reasoning capabilities and can reveal relative performance differences that are not apparent in the original assessments. To demonstrate the effectiveness of our approach, we create a new multiple-choice test corpus, extend it into a family of evaluations, and assess a collection of LLMs. Our results offer insights into the comparative reasoning abilities of these models, particularly highlighting distinctions between OpenAI's o1-preview and Google's gemini-pro-1.5-002.
Authors: William F. Bradley
Abstract: We investigate the patterns of incorrect answers produced by large language models (LLMs) during evaluation. These errors exhibit highly non-intuitive behaviors unique to each model. By analyzing these patterns, we measure the similarities between LLMs and construct a taxonomy that categorizes them based on their error correlations. Our findings reveal that the incorrect responses are not randomly distributed but systematically correlated across models, providing new insights into the underlying structures and relationships among LLMs.
Authors: Makoto Fukushima, Shusuke Eshita, Hiroshige Fukuhara
Abstract: Color-word associations play a fundamental role in human cognition and design applications. Large Language Models (LLMs) have become widely available and demonstrated intelligent behaviors in various benchmarks with natural conversation skills. However, their ability to replicate human color-word associations remains understudied. We compared multiple generations of LLMs (from GPT-3 to GPT-4o) against human color-word associations using data collected from over 10,000 Japanese participants, involving 17 colors and words from eight categories in Japanese. Our findings reveal a clear progression in LLM performance across generations, with GPT-4o achieving the highest accuracy in predicting the best voted word for each color and category. However, the highest median performance was approximately 50% even for GPT-4o with visual inputs (chance level is 10%), and the performance levels varied significantly across word categories and colors, indicating a failure to fully replicate human color-word associations. On the other hand, color discrimination ability estimated from our color-word association data showed that LLMs demonstrated high correlation with human color discrimination patterns, similarly to previous studies. Our study highlights both the advancements in LLM capabilities and their persistent limitations, suggesting differences in semantic memory structures between humans and LLMs in representing color-word associations.
Authors: Xingwu Sun, Yanfeng Chen, Yiqing Huang, Ruobing Xie, Jiaqi Zhu, Kai Zhang, Shuaipeng Li, Zhen Yang, Jonny Han, Xiaobo Shu, Jiahao Bu, Zhongzhi Chen, Xuemeng Huang, Fengzong Lian, Saiyong Yang, Jianfeng Yan, Yuyuan Zeng, Xiaoqin Ren, Chao Yu, Lulu Wu, Yue Mao, Jun Xia, Tao Yang, Suncong Zheng, Kan Wu, Dian Jiao, Jinbao Xue, Xipeng Zhang, Decheng Wu, Kai Liu, Dengpeng Wu, Guanghui Xu, Shaohua Chen, Shuang Chen, Xiao Feng, Yigeng Hong, Junqiang Zheng, Chengcheng Xu, Zongwei Li, Xiong Kuang, Jianglu Hu, Yiqi Chen, Yuchi Deng, Guiyang Li, Ao Liu, Chenchen Zhang, Shihui Hu, Zilong Zhao, Zifan Wu, Yao Ding, Weichao Wang, Han Liu, Roberts Wang, Hao Fei, Peijie She, Ze Zhao, Xun Cao, Hai Wang, Fusheng Xiang, Mengyuan Huang, Zhiyuan Xiong, Bin Hu, Xuebin Hou, Lei Jiang, Jiajia Wu, Yaping Deng, Yi Shen, Qian Wang, Weijie Liu, Jie Liu, Meng Chen, Liang Dong, Weiwen Jia, Hu Chen, Feifei Liu, Rui Yuan, Huilin Xu, Zhenxiang Yan, Tengfei Cao, Zhichao Hu, Xinhua Feng, Dong Du, Tinghao She, Yangyu Tao, Feng Zhang, Jianchen Zhu, Chengzhong Xu, Xirui Li, Chong Zha, Wen Ouyang, Yinben Xia, Xiang Li, Zekun He, Rongpeng Chen, Jiawei Song, Ruibin Chen, Fan Jiang, Chongqing Zhao, Bo Wang, Hao Gong, Rong Gan, Winston Hu, Zhanhui Kang, Yong Yang, Yuhong Liu, Di Wang, Jie Jiang
Abstract: In this paper, we introduce Hunyuan-Large, which is currently the largest open-source Transformer-based mixture of experts model, with a total of 389 billion parameters and 52 billion activation parameters, capable of handling up to 256K tokens. We conduct a thorough evaluation of Hunyuan-Large's superior performance across various benchmarks including language understanding and generation, logical reasoning, mathematical problem-solving, coding, long-context, and aggregated tasks, where it outperforms LLama3.1-70B and exhibits comparable performance when compared to the significantly larger LLama3.1-405B model. Key practice of Hunyuan-Large include large-scale synthetic data that is orders larger than in previous literature, a mixed expert routing strategy, a key-value cache compression technique, and an expert-specific learning rate strategy. Additionally, we also investigate the scaling laws and learning rate schedule of mixture of experts models, providing valuable insights and guidances for future model development and optimization. The code and checkpoints of Hunyuan-Large are released to facilitate future innovations and applications. Codes: https://github.com/Tencent/Hunyuan-Large Models: https://huggingface.co/tencent/Tencent-Hunyuan-Large
URLs: https://github.com/Tencent/Hunyuan-Large, https://huggingface.co/tencent/Tencent-Hunyuan-Large
Authors: Jinqi Luo, Kwan Ho Ryan Chan, Dimitris Dimos, Ren\'e Vidal
Abstract: Modern text-to-vision generative models often hallucinate when the prompt describing the scene to be generated is underspecified. In large language models (LLMs), a prevalent strategy to reduce hallucinations is to retrieve factual knowledge from an external database. While such retrieval augmentation strategies have great potential to enhance text-to-vision generators, existing static top-K retrieval methods explore the knowledge pool once, missing the broader context necessary for high-quality generation. Furthermore, LLMs internally possess rich world knowledge learned during large-scale training (parametric knowledge) that could mitigate the need for external data retrieval. This paper proposes Contextual Knowledge Pursuit (CKPT), a framework that leverages the complementary strengths of external and parametric knowledge to help generators produce reliable visual content. Instead of the one-time retrieval of facts from an external database to improve a given prompt, CKPT uses (1) an LLM to decide whether to seek external knowledge or to self-elicit descriptions from LLM parametric knowledge, (2) a knowledge pursuit process to contextually seek and sequentially gather most relevant facts, (3) a knowledge aggregator for prompt enhancement with the gathered fact context, and (4) a filtered fine-tuning objective to improve visual synthesis with richer prompts. We evaluate CKPT across multiple text-driven generative tasks (image, 3D rendering, and video) on datasets of rare objects and daily scenarios. Our results show that CKPT is capable of generating faithful and semantically rich content across diverse visual domains, offering a promising data source for zero-shot synthesis and filtered fine-tuning of text-to-vision generative models.
Authors: Jonathan Thomm, Giacomo Camposampiero, Aleksandar Terzic, Michael Hersche, Bernhard Sch\"olkopf, Abbas Rahimi
Abstract: We analyze the capabilities of Transformer language models in learning compositional discrete tasks. To this end, we evaluate training LLaMA models and prompting GPT-4 and Gemini on four tasks demanding to learn a composition of several discrete sub-tasks. In particular, we measure how well these models can reuse primitives observable in the sub-tasks to learn the composition task. Our results indicate that compositional learning in state-of-the-art Transformer language models is highly sample inefficient: LLaMA requires more data samples than relearning all sub-tasks from scratch to learn the compositional task; in-context prompting with few samples is unreliable and fails at executing the sub-tasks or correcting the errors in multi-round code generation. Further, by leveraging complexity theory, we support these findings with a theoretical analysis focused on the sample inefficiency of gradient descent in memorizing feedforward models. We open source our code at https://github.com/IBM/limitations-lm-algorithmic-compositional-learning.
URLs: https://github.com/IBM/limitations-lm-algorithmic-compositional-learning.
Authors: Qiang Hu, Jin Wen, Maxime Cordy, Yuheng Huang, Wei Ma, Xiaofei Xie, Lei Ma
Abstract: Large language models (LLMs) have recently achieved significant success across various application domains, garnering substantial attention from different communities. Unfortunately, even for the best LLM, many \textit{faults} still exist that LLM cannot properly predict. Such faults will harm the usability of LLMs in general and could introduce safety issues in reliability-critical systems such as autonomous driving systems. How to quickly reveal these faults in real-world datasets that LLM could face is important, but challenging. The major reason is that the ground truth is necessary but the data labeling process is heavy considering the time and human effort. To handle this problem, in the conventional deep learning testing field, test selection methods have been proposed for efficiently evaluating deep learning models by prioritizing faults. However, despite their importance, the usefulness of these methods on LLMs is unclear, and lack of exploration. In this paper, we conduct the first empirical study to investigate the effectiveness of existing fault detection methods for LLMs. Experimental results on four different tasks~(including both code tasks and natural language processing tasks) and four LLMs~(e.g., LLaMA3 and GPT4) demonstrated that simple methods such as Margin perform well on LLMs but there is still a big room for improvement. Based on the study, we further propose \textbf{MuCS}, a prompt \textbf{Mu}tation-based prediction \textbf{C}onfidence \textbf{S}moothing framework to boost the fault detection capability of existing methods. Concretely, multiple prompt mutation techniques have been proposed to help collect more diverse outputs for confidence smoothing. The results show that our proposed framework significantly enhances existing methods with the improvement of test relative coverage by up to 70.53\%.
Authors: Xuelong Geng, Tianyi Xu, Kun Wei, Bingshen Mu, Hongfei Xue, He Wang, Yangze Li, Pengcheng Guo, Yuhang Dai, Longhao Li, Mingchen Shao, Lei Xie
Abstract: Large Language Models (LLMs) have demonstrated unparalleled effectiveness in various NLP tasks, and integrating LLMs with automatic speech recognition (ASR) is becoming a mainstream paradigm. Building upon this momentum, our research delves into an in-depth examination of this paradigm on a large open-source Chinese dataset. Specifically, our research aims to evaluate the impact of various configurations of speech encoders, LLMs, and projector modules in the context of the speech foundation encoder-LLM ASR paradigm. Furthermore, we introduce a three-stage training approach, expressly developed to enhance the model's ability to align auditory and textual information. The implementation of this approach, alongside the strategic integration of ASR components, enabled us to achieve the SOTA performance on the AISHELL-1, Test_Net, and Test_Meeting test sets. Our analysis presents an empirical foundation for future research in LLM-based ASR systems and offers insights into optimizing performance using Chinese datasets. We will publicly release all scripts used for data preparation, training, inference, and scoring, as well as pre-trained models and training logs to promote reproducible research.
Authors: Xin Xiao, Bohong Wu, Jiacong Wang, Chunyuan Li, Xun Zhou, Haoyuan Guo
Abstract: Existing image-text modality alignment in Vision Language Models (VLMs) treats each text token equally in an autoregressive manner. Despite being simple and effective, this method results in sub-optimal cross-modal alignment by over-emphasizing the text tokens that are less correlated with or even contradictory with the input images. In this paper, we advocate for assigning distinct contributions for each text token based on its visual correlation. Specifically, we present by contrasting image inputs, the difference in prediction logits on each text token provides strong guidance of visual correlation. We therefore introduce Contrastive ALignment (CAL), a simple yet effective re-weighting strategy that prioritizes training visually correlated tokens. Our experimental results demonstrate that CAL consistently improves different types of VLMs across different resolutions and model sizes on various benchmark datasets. Importantly, our method incurs minimal additional computational overhead, rendering it highly efficient compared to alternative data scaling strategies. Codes are available at https://github.com/foundation-multimodal-models/CAL.
Authors: Kanghee Park, Jiayu Wang, Taylor Berg-Kirkpatrick, Nadia Polikarpova, Loris D'Antoni
Abstract: Large Language Models (LLMs) struggle with reliably generating highly structured outputs, such as program code, mathematical formulas, or well-formed markup. Constrained decoding approaches mitigate this problem by greedily restricting what tokens an LLM can output at each step to guarantee that the output matches a given constraint. Specifically, in grammar-constrained decoding (GCD), the LLM's output must follow a given grammar. In this paper, we demonstrate that GCD techniques (and in general constrained decoding techniques) can distort the LLM's distribution, leading to outputs that are grammatical but appear with likelihoods that are not proportional to the ones given by the LLM, and so ultimately are low-quality. We call the problem of aligning sampling with a grammar constraint, grammar-aligned decoding (GAD), and propose adaptive sampling with approximate expected futures (ASAp), a decoding algorithm that guarantees the output to be grammatical while provably producing outputs that match the conditional probability of the LLM's distribution conditioned on the given grammar constraint. Our algorithm uses prior sample outputs to soundly overapproximate the future grammaticality of different output prefixes. Our evaluation on code generation and structured NLP tasks shows how ASAp often produces outputs with higher likelihood (according to the LLM's distribution) than existing GCD techniques, while still enforcing the desired grammatical constraints.
Authors: Rafael Rafailov, Yaswanth Chittepu, Ryan Park, Harshit Sikchi, Joey Hejna, Bradley Knox, Chelsea Finn, Scott Niekum
Abstract: Reinforcement Learning from Human Feedback (RLHF) has been crucial to the recent success of Large Language Models (LLMs), however, it is often a complex and brittle process. In the classical RLHF framework, a reward model is first trained to represent human preferences, which is in turn used by an online reinforcement learning (RL) algorithm to optimize the LLM. A prominent issue with such methods is reward over-optimization or reward hacking, where performance as measured by the learned proxy reward model increases, but true quality plateaus or even deteriorates. Direct Alignment Algorithms (DDAs) like Direct Preference Optimization have emerged as alternatives to the classical RLHF pipeline by circumventing the reward modeling phase. However, although DAAs do not use a separate proxy reward model, they still commonly deteriorate from over-optimization. While the so-called reward hacking phenomenon is not well-defined for DAAs, we still uncover similar trends: at higher KL budgets, DAA algorithms exhibit similar degradation patterns to their classic RLHF counterparts. In particular, we find that DAA methods deteriorate not only across a wide range of KL budgets but also often before even a single epoch of the dataset is completed. Through extensive empirical experimentation, this work formulates and formalizes the reward over-optimization or hacking problem for DAAs and explores its consequences across objectives, training regimes, and model scales.
Authors: Songlin Yang, Bailin Wang, Yu Zhang, Yikang Shen, Yoon Kim
Abstract: Transformers with linear attention (i.e., linear transformers) and state-space models have recently been suggested as a viable linear-time alternative to transformers with softmax attention. However, these models still underperform transformers especially on tasks that require in-context retrieval. While more expressive variants of linear transformers which replace the additive update in linear transformers with the delta rule (DeltaNet) have been found to be more effective at associative recall, existing algorithms for training such models do not parallelize over sequence length and are thus inefficient to train on modern hardware. This work describes a hardware-efficient algorithm for training linear transformers with the delta rule, which exploits a memory-efficient representation for computing products of Householder matrices. This algorithm allows us to scale up DeltaNet to standard language modeling settings. We train a 1.3B model for 100B tokens and find that it outperforms recent linear-time baselines such as Mamba and GLA in terms of perplexity and zero-shot performance on downstream tasks. We also experiment with two hybrid models which combine DeltaNet layers with (1) sliding-window attention layers every other layer or (2) two global attention layers, and find that these hybrids outperform strong transformer baselines.
Authors: Zhexin Zhang, Junxiao Yang, Pei Ke, Shiyao Cui, Chujie Zheng, Hongning Wang, Minlie Huang
Abstract: LLMs are known to be vulnerable to jailbreak attacks, even after safety alignment. An important observation is that, while different types of jailbreak attacks can generate significantly different queries, they mostly result in similar responses that are rooted in the same harmful knowledge (e.g., detailed steps to make a bomb). Therefore, we conjecture that directly unlearn the harmful knowledge in the LLM can be a more effective way to defend against jailbreak attacks than the mainstream supervised fine-tuning (SFT) approaches. Our extensive experiments demonstrate the surprising generalizability of our unlearning-based approach: using only 20 raw harmful questions without any jailbreak prompt during training, our solution reduced the Attack Success Rate (ASR) in Vicuna-7B from 82.6% to 7.7% on out-of-distribution (OOD) harmful questions wrapped with various complex jailbreak prompts . This significantly outperforms Llama2-7B-Chat, which is fine-tuned on about 0.1M safety alignment samples but still has an ASR of 21.9% even under the help of an additional safety system prompt. Further analysis reveals that the generalization ability of our solution may stem from the intrinsic relatedness among harmful responses across harmful questions (e.g., response patterns, shared steps and actions in response, and similarity among their learned representations in the LLM). Our code is available at \url{https://github.com/thu-coai/SafeUnlearning}.
Authors: Siddhartha Chatterjee, Bina Ramamurthy
Abstract: This study analyzes the application of code-generating Large Language Models in the creation of immutable Solidity smart contracts on the Ethereum Blockchain. Other works have previously analyzed Artificial Intelligence code generation abilities. This paper aims to expand this to a larger scope to include programs where security and efficiency are of utmost priority such as smart contracts. The hypothesis leading into the study was that LLMs in general would have difficulty in rigorously implementing security details in the code, which was shown through our results, but surprisingly generally succeeded in many common types of contracts. We also discovered a novel way of generating smart contracts through new prompting strategies.
Authors: Abhishek Dutta, Yen-Che Hsiao
Abstract: We propose a novel in-context learning algorithm for building autonomous decision-making language agents. The language agent continuously attempts to solve the same task by self-correcting each time the task fails. Our selected language agent demonstrates the ability to solve tasks in a text-based game environment. Our results show that the gemma-2-9b-it language model, using our proposed method, can successfully complete two of six tasks that failed in the first attempt. This highlights the effectiveness of our approach in enhancing the problem-solving capabilities of a single language model through self-correction, paving the way for more advanced autonomous agents. The code is publicly available at https://github.com/YenCheHsiao/AutonomousLLMAgentwithAdaptingPlanning.
URLs: https://github.com/YenCheHsiao/AutonomousLLMAgentwithAdaptingPlanning.
Authors: Zhifei Xie, Changqiao Wu
Abstract: Recent advances in language models have achieved significant progress. GPT-4o, as a new milestone, has enabled real-time conversations with humans, demonstrating near-human natural fluency. Such human-computer interaction necessitates models with the capability to perform reasoning directly with the audio modality and generate output in streaming. However, this remains beyond the reach of current academic models, as they typically depend on extra TTS systems for speech synthesis, resulting in undesirable latency. This paper introduces the Mini-Omni, an audio-based end-to-end conversational model, capable of real-time speech interaction. To achieve this capability, we propose a text-instructed speech generation method, along with batch-parallel strategies during inference to further boost the performance. Our method also helps to retain the original model's language capabilities with minimal degradation, enabling other works to establish real-time interaction capabilities. We call this training method "Any Model Can Talk". We also introduce the VoiceAssistant-400K dataset to fine-tune models optimized for speech output. To our best knowledge, Mini-Omni is the first fully end-to-end, open-source model for real-time speech interaction, offering valuable potential for future research.
Authors: Emet Bethany, Mazal Bethany, Juan Arturo Nolazco Flores, Sumit Kumar Jha, Peyman Najafirad
Abstract: Recent advancements in AI safety have led to increased efforts in training and red-teaming large language models (LLMs) to mitigate unsafe content generation. However, these safety mechanisms may not be comprehensive, leaving potential vulnerabilities unexplored. This paper introduces MathPrompt, a novel jailbreaking technique that exploits LLMs' advanced capabilities in symbolic mathematics to bypass their safety mechanisms. By encoding harmful natural language prompts into mathematical problems, we demonstrate a critical vulnerability in current AI safety measures. Our experiments across 13 state-of-the-art LLMs reveal an average attack success rate of 73.6\%, highlighting the inability of existing safety training mechanisms to generalize to mathematically encoded inputs. Analysis of embedding vectors shows a substantial semantic shift between original and encoded prompts, helping explain the attack's success. This work emphasizes the importance of a holistic approach to AI safety, calling for expanded red-teaming efforts to develop robust safeguards across all potential input types and their associated risks.
Authors: Ezra Karger, Houtan Bastani, Chen Yueh-Han, Zachary Jacobs, Danny Halawi, Fred Zhang, Philip E. Tetlock
Abstract: Forecasts of future events are essential inputs into informed decision-making. Machine learning (ML) systems have the potential to deliver forecasts at scale, but there is no framework for evaluating the accuracy of ML systems on a standardized set of forecasting questions. To address this gap, we introduce ForecastBench: a dynamic benchmark that evaluates the accuracy of ML systems on an automatically generated and regularly updated set of 1,000 forecasting questions. To avoid any possibility of data leakage, ForecastBench is comprised solely of questions about future events that have no known answer at the time of submission. We quantify the capabilities of current ML systems by collecting forecasts from expert (human) forecasters, the general public, and LLMs on a random subset of questions from the benchmark ($N=200$). While LLMs have achieved super-human performance on many benchmarks, they perform less well here: expert forecasters outperform the top-performing LLM (p-value $=0.01$). We display system and human scores in a public leaderboard at www.forecastbench.org.
Authors: Yadong Li, Haoze Sun, Mingan Lin, Tianpeng Li, Guosheng Dong, Tao Zhang, Bowen Ding, Wei Song, Zhenglin Cheng, Yuqi Huo, Song Chen, Xu Li, Da Pan, Shusen Zhang, Xin Wu, Zheng Liang, Jun Liu, Tao Zhang, Keer Lu, Yaqi Zhao, Yanjun Shen, Fan Yang, Kaicheng Yu, Tao Lin, Jianhua Xu, Zenan Zhou, Weipeng Chen
Abstract: The salient multimodal capabilities and interactive experience of GPT-4o highlight its critical role in practical applications, yet it lacks a high-performing open-source counterpart. In this paper, we introduce Ocean-omni, the first open-source 7B Multimodal Large Language Model (MLLM) adept at concurrently processing and analyzing modalities of image, video, audio, and text, while delivering an advanced multimodal interactive experience and strong performance. We propose an effective multimodal training schema starting with 7B model and proceeding through two stages of multimodal alignment and multitask fine-tuning across audio, image, video, and text modal. This approach equips the language model with the ability to handle visual and audio data effectively. Demonstrating strong performance across various omni-modal and multimodal benchmarks, we aim for this contribution to serve as a competitive baseline for the open-source community in advancing multimodal understanding and real-time interaction.
Authors: Yanan Zhang, Jiangmeng Li, Lixiang Liu, Wenwen Qiang
Abstract: Foundational Vision-Language models such as CLIP have exhibited impressive generalization in downstream tasks. However, CLIP suffers from a two-level misalignment issue, i.e., task misalignment and data misalignment, when adapting to specific tasks. Soft prompt tuning has mitigated the task misalignment, yet the data misalignment remains a challenge. To analyze the impacts of the data misalignment, we revisit the pre-training and adaptation processes of CLIP and develop a structural causal model. We discover that while we expect to capture task-relevant information for downstream tasks accurately, the task-irrelevant knowledge impacts the prediction results and hampers the modeling of the true relationships between the images and the predicted classes. As task-irrelevant knowledge is unobservable, we leverage the front-door adjustment and propose Causality-Guided Semantic Decoupling and Classification (CDC) to mitigate the interference of task-irrelevant knowledge. Specifically, we decouple semantics contained in the data of downstream tasks and perform classification based on each semantic. Furthermore, we employ the Dempster-Shafer evidence theory to evaluate the uncertainty of each prediction generated by diverse semantics. Experiments conducted in multiple different settings have consistently demonstrated the effectiveness of CDC.
Authors: Ryan Liu, Jiayi Geng, Addison J. Wu, Ilia Sucholutsky, Tania Lombrozo, Thomas L. Griffiths
Abstract: Chain-of-thought (CoT) prompting has become a widely used strategy for working with large language and multimodal models. While CoT has been shown to improve performance across many tasks, determining the settings in which it is effective remains an ongoing effort. In particular, it is still an open question in what settings CoT systematically reduces model performance. In this paper, we seek to identify the characteristics of tasks where CoT reduces performance by drawing inspiration from cognitive psychology, looking at cases where (i) verbal thinking or deliberation hurts performance in humans, and (ii) the constraints governing human performance generalize to language models. Three such cases are implicit statistical learning, visual recognition, and classifying with patterns containing exceptions. In extensive experiments across all three settings, we find that a diverse collection of state-of-the-art models exhibit significant drop-offs in performance (e.g., up to 36.3% absolute accuracy for OpenAI o1-preview compared to GPT-4o) when using inference-time reasoning compared to zero-shot counterparts. We also identify three tasks that satisfy condition (i) but not (ii), and find that while verbal thinking reduces human performance in these tasks, CoT retains or increases model performance. Overall, our results show that while there is not an exact parallel between the cognitive processes of models and those of humans, considering cases where thinking has negative consequences for human performance can help us identify settings where it negatively impacts models. By connecting the literature on human deliberation with evaluations of CoT, we offer a new tool that can be used in understanding the impact of prompt choices and inference-time reasoning.
Authors: Shihan Dou, Jiazheng Zhang, Jianxiang Zang, Yunbo Tao, Weikang Zhou, Haoxiang Jia, Shichun Liu, Yuming Yang, Zhiheng Xi, Shenxi Wu, Shaoqing Zhang, Muling Wu, Changze Lv, Limao Xiong, Wenyu Zhan, Lin Zhang, Rongxiang Weng, Jingang Wang, Xunliang Cai, Yueming Wu, Ming Wen, Rui Zheng, Tao Ji, Yixin Cao, Tao Gui, Xipeng Qiu, Qi Zhang, Xuanjing Huang
Abstract: We introduce MPLSandbox, an out-of-the-box multi-programming language sandbox designed to provide unified and comprehensive feedback from compiler and analysis tools for Large Language Models (LLMs). It can automatically identify the programming language of the code, compiling and executing it within an isolated sub-sandbox to ensure safety and stability. In addition, MPLSandbox also integrates both traditional and LLM-based code analysis tools, providing a comprehensive analysis of generated code. MPLSandbox can be effortlessly integrated into the training and deployment of LLMs to improve the quality and correctness of their generated code. It also helps researchers streamline their workflows for various LLM-based code-related tasks, reducing the development cost. To validate the effectiveness of MPLSandbox, we integrate it into training and deployment approaches, and also employ it to optimize workflows for a wide range of real-world code-related tasks. Our goal is to enhance researcher productivity on LLM-based code-related tasks by simplifying and automating workflows through delegation to MPLSandbox.
Authors: Ognjen (Oggi), Rudovic, Pranay Dighe, Yi Su, Vineet Garg, Sameer Dharur, Xiaochuan Niu, Ahmed H. Abdelaziz, Saurabh Adya, Ahmed Tewfik
Abstract: Follow-up conversations with virtual assistants (VAs) enable a user to seamlessly interact with a VA without the need to repeatedly invoke it using a keyword (after the first query). Therefore, accurate Device-directed Speech Detection (DDSD) from the follow-up queries is critical for enabling naturalistic user experience. To this end, we explore the notion of Large Language Models (LLMs) and model the first query when making inference about the follow-ups (based on the ASR-decoded text), via prompting of a pretrained LLM, or by adapting a binary classifier on top of the LLM. In doing so, we also exploit the ASR uncertainty when designing the LLM prompts. We show on the real-world dataset of follow-up conversations that this approach yields large gains (20-40% reduction in false alarms at 10% fixed false rejects) due to the joint modeling of the previous speech context and ASR uncertainty, compared to when follow-ups are modeled alone.