Authors: Kwanho Kim, Soojong Kim
Abstract: Sentiment analysis of alternative tobacco products on social media is important for tobacco control research. Large Language Models (LLMs) can help streamline the labor-intensive human sentiment analysis process. This study examined the accuracy of LLMs in replicating human sentiment evaluation of social media messages about heated tobacco products (HTPs). The research used GPT-3.5 and GPT-4 Turbo to classify 500 Facebook and 500 Twitter messages, including anti-HTPs, pro-HTPs, and neutral messages. The models evaluated each message up to 20 times, and their majority label was compared to human evaluators. Results showed that GPT-3.5 accurately replicated human sentiment 61.2% of the time for Facebook messages and 57.0% for Twitter messages. GPT-4 Turbo performed better, with 81.7% accuracy for Facebook and 77.0% for Twitter. Using three response instances, GPT-4 Turbo achieved 99% of the accuracy of twenty instances. GPT-4 Turbo also had higher accuracy for anti- and pro-HTPs messages compared to neutral ones. Misclassifications by GPT-3.5 often involved anti- or pro-HTPs messages being labeled as neutral or irrelevant, while GPT-4 Turbo showed improvements across all categories. In conclusion, LLMs can be used for sentiment analysis of HTP-related social media messages, with GPT-4 Turbo reaching around 80% accuracy compared to human experts. However, there's a risk of misrepresenting overall sentiment due to differences in accuracy across sentiment categories.
Authors: Jiale Fu, Yuchu Jiang, Junkai Chen, Jiaming Fan, Xin Geng, Xu Yang
Abstract: Ensemble methods enhance Large Language Models (LLMs) by combining multiple models but suffer from high computational costs. In this paper, we introduce Speculative Ensemble, a novel framework that accelerates LLM ensembles without sacrificing performance, inspired by Speculative Decoding-where a small proposal model generates tokens sequentially, and a larger target model verifies them in parallel. Our approach builds on two key insights: (1) the verification distribution can be the ensemble distribution of both the proposal and target models, and (2) alternating each model as the proposer and verifier can further enhance efficiency. We generalize this method to ensembles with n models and theoretically prove that SE is never slower than a standard ensemble, typically achieving faster speed. Extensive experiments demonstrate speed improvements of 1.11x-2.23x over standard ensemble techniques without compromising generation quality. Our code is available at https://github.com/Kamichanw/Speculative-Ensemble/
Authors: Md. Shahriar Hossain Apu, Md Saiful Islam, Tanjim Taharat Aurpa
Abstract: In 2024, the outbreak of Human Metapneumovirus (HMPV) in China, which later spread to the UK and other countries, raised significant public concern. While HMPV typically causes mild symptoms, its effects on vulnerable individuals prompted health authorities to emphasize preventive measures. This paper explores how sentiment analysis can enhance our understanding of public reactions to HMPV by analyzing social media data. We apply transformer models, particularly XLNet, achieving 93.50% accuracy in sentiment classification. Additionally, we use explainable AI (XAI) through SHAP to improve model transparency.
Authors: Arpita Vats, Rahul Raja, Mrinal Mathur, Vinija Jain, Aman Chadha
Abstract: The diversity and complexity of Indic languages present unique challenges for natural language processing (NLP) tasks, particularly in the domain of question answering (QA).To address these challenges, this paper explores the application of State Space Models (SSMs),to build efficient and contextually aware QA systems tailored for Indic languages. SSMs are particularly suited for this task due to their ability to model long-term and short-term dependencies in sequential data, making them well-equipped to handle the rich morphology, complex syntax, and contextual intricacies characteristic of Indian languages. We evaluated multiple SSM architectures across diverse datasets representing various Indic languages and conducted a comparative analysis of their performance. Our results demonstrate that these models effectively capture linguistic subtleties, leading to significant improvements in question interpretation, context alignment, and answer generation. This work represents the first application of SSMs to question answering tasks in Indic languages, establishing a foundational benchmark for future research in this domain. We propose enhancements to existing SSM frameworks, optimizing their applicability to low-resource settings and multilingual scenarios prevalent in Indic languages.
Authors: Man Luo, Bradley Peterson, Rafael Gan, Hari Ramalingame, Navya Gangrade, Ariadne Dimarogona, Imon Banerjee, Phillip Howard
Abstract: Peer review is crucial for advancing and improving science through constructive criticism. However, toxic feedback can discourage authors and hinder scientific progress. This work explores an important but underexplored area: detecting toxicity in peer reviews. We first define toxicity in peer reviews across four distinct categories and curate a dataset of peer reviews from the OpenReview platform, annotated by human experts according to these definitions. Leveraging this dataset, we benchmark a variety of models, including a dedicated toxicity detection model, a sentiment analysis model, several open-source large language models (LLMs), and two closed-source LLMs. Our experiments explore the impact of different prompt granularities, from coarse to fine-grained instructions, on model performance. Notably, state-of-the-art LLMs like GPT-4 exhibit low alignment with human judgments under simple prompts but achieve improved alignment with detailed instructions. Moreover, the model's confidence score is a good indicator of better alignment with human judgments. For example, GPT-4 achieves a Cohen's Kappa score of 0.56 with human judgments, which increases to 0.63 when using only predictions with a confidence score higher than 95%. Overall, our dataset and benchmarks underscore the need for continued research to enhance toxicity detection capabilities of LLMs. By addressing this issue, our work aims to contribute to a healthy and responsible environment for constructive academic discourse and scientific collaboration.
Authors: Alexander Kilpatrick, Maria Flaksman
Abstract: This meta-study explores the relationships between humor, phonemic bigram surprisal, emotional valence, and memory recall. Prior research indicates that words with higher phonemic surprisal are more readily remembered, suggesting that unpredictable phoneme sequences promote long-term memory recall. Emotional valence is another well-documented factor influencing memory, with negative experiences and stimuli typically being remembered more easily than positive ones. Building on existing findings, this study highlights that words with negative associations often exhibit greater surprisal and are easier to recall. Humor, however, presents an exception: while associated with positive emotions, humorous words also display heightened surprisal and enhanced memorability.
Authors: Peiwen Yuan, Shaoxiong Feng, Yiwei Li, Xinglin Wang, Yueqi Zhang, Jiayi Shi, Chuyi Tan, Boyuan Pan, Yao Hu, Kan Li
Abstract: The rapid advancement of large language models (LLMs) has led to a surge in both model supply and application demands. To facilitate effective matching between them, reliable, generic and efficient benchmark generators are widely needed. However, human annotators are constrained by inefficiency, and current LLM benchmark generators not only lack generalizability but also struggle with limited reliability, as they lack a comprehensive evaluation framework for validation and optimization. To fill this gap, we first propose an automated and unbiased evaluation framework, structured around four dimensions and ten criteria. Under this framework, we carefully analyze the advantages and weaknesses of directly prompting LLMs as generic benchmark generators. To enhance the reliability, we introduce a series of methods to address the identified weaknesses and integrate them as BenchMaker. Experiments across multiple LLMs and tasks confirm that BenchMaker achieves superior or comparable performance to human-annotated benchmarks on all metrics, highlighting its generalizability and reliability. More importantly, it delivers highly consistent evaluation results across 12 LLMs (0.967 Pearson correlation against MMLU-Pro), while taking only $0.005 and 0.38 minutes per sample.
Authors: Hadas Ben-Atya, Naama Gavrielov, Zvi Badash, Gili Focht, Ruth Cytter-Kuint, Talar Hagopian, Dan Turner, Moti Freiman
Abstract: Reliable extraction of structured data from radiology reports using Large Language Models (LLMs) remains challenging, especially for complex, non-English texts like Hebrew. This study introduces an agent-based uncertainty-aware approach to improve the trustworthiness of LLM predictions in medical applications. We analyzed 9,683 Hebrew radiology reports from Crohn's disease patients (from 2010 to 2023) across three medical centers. A subset of 512 reports was manually annotated for six gastrointestinal organs and 15 pathological findings, while the remaining reports were automatically annotated using HSMP-BERT. Structured data extraction was performed using Llama 3.1 (Llama 3-8b-instruct) with Bayesian Prompt Ensembles (BayesPE), which employed six semantically equivalent prompts to estimate uncertainty. An Agent-Based Decision Model integrated multiple prompt outputs into five confidence levels for calibrated uncertainty and was compared against three entropy-based models. Performance was evaluated using accuracy, F1 score, precision, recall, and Cohen's Kappa before and after filtering high-uncertainty cases. The agent-based model outperformed the baseline across all metrics, achieving an F1 score of 0.3967, recall of 0.6437, and Cohen's Kappa of 0.3006. After filtering high-uncertainty cases (greater than or equal to 0.5), the F1 score improved to 0.4787, and Kappa increased to 0.4258. Uncertainty histograms demonstrated clear separation between correct and incorrect predictions, with the agent-based model providing the most well-calibrated uncertainty estimates. By incorporating uncertainty-aware prompt ensembles and an agent-based decision model, this approach enhances the performance and reliability of LLMs in structured data extraction from radiology reports, offering a more interpretable and trustworthy solution for high-stakes medical applications.
Authors: Alan Zhu, Parth Asawa, Jared Quincy Davis, Lingjiao Chen, Ion Stoica, Joseph E. Gonzalez, Matei Zaharia
Abstract: As the demand for high-quality data in model training grows, researchers and developers are increasingly generating synthetic data to tune and train LLMs. A common assumption about synthetic data is that sampling from instruct-tuned models is sufficient; however, these models struggle to produce diverse outputs-a key requirement for generalization. Despite various prompting methods, in this work we show that achieving meaningful diversity from instruct-tuned models remains challenging. In contrast, we find base models without post-training exhibit greater diversity, but are less capable at instruction following and hence of lower quality. Leveraging this insight, we propose Base-Refine (BARE), a synthetic data generation method that combines the diversity of base models with the quality of instruct-tuned models through a two-stage process. With minimal few-shot examples and curation, BARE generates diverse and high-quality datasets, improving downstream task performance. We show that fine-tuning with as few as 1,000 BARE-generated samples can reach performance comparable to the best similarly sized models on LiveCodeBench tasks. Furthermore, fine-tuning with BARE-generated data achieves a 101% improvement over instruct-only data on GSM8K and a 18.4% improvement over SOTA methods on RAFT.
Authors: Alexei Figueroa, Justus Westerhoff, Atefi Golzar, Dennis Fast, Benjamin Winter, Felix Alexader Gers, Alexander L\"oser, Wolfang Nejdl
Abstract: Biologically inspired neural networks offer alternative avenues to model data distributions. FlyVec is a recent example that draws inspiration from the fruit fly's olfactory circuit to tackle the task of learning word embeddings. Surprisingly, this model performs competitively even against deep learning approaches specifically designed to encode text, and it does so with the highest degree of computational efficiency. We pose the question of whether this performance can be improved further. For this, we introduce Comply. By incorporating positional information through complex weights, we enable a single-layer neural network to learn sequence representations. Our experiments show that Comply not only supersedes FlyVec but also performs on par with significantly larger state-of-the-art models. We achieve this without additional parameters. Comply yields sparse contextual representations of sentences that can be interpreted explicitly from the neuron weights.
Authors: Nina Corvelo Benz, Stratis Tsirtsis, Eleni Straitouri, Ivi Chatzi, Ander Artola Velasco, Suhas Thejaswi, Manuel Gomez-Rodriguez
Abstract: State of the art large language models rely on randomization to respond to a prompt. As an immediate consequence, a model may respond differently to the same prompt if asked multiple times. In this work, we argue that the evaluation and ranking of large language models should control for the randomization underpinning their functioning. Our starting point is the development of a causal model for coupled autoregressive generation, which allows different large language models to sample responses with the same source of randomness. Building upon our causal model, we first show that, on evaluations based on benchmark datasets, coupled autoregressive generation leads to the same conclusions as vanilla autoregressive generation but using provably fewer samples. However, we further show that, on evaluations based on (human) pairwise comparisons, coupled and vanilla autoregressive generation can surprisingly lead to different rankings when comparing more than two models, even with an infinite amount of samples. This suggests that the apparent advantage of a model over others in existing evaluation protocols may not be genuine but rather confounded by the randomness inherent to the generation process. To illustrate and complement our theoretical results, we conduct experiments with several large language models from the Llama family. We find that, across multiple knowledge areas from the popular MMLU benchmark dataset, coupled autoregressive generation requires up to 40% fewer samples to reach the same conclusions as vanilla autoregressive generation. Further, using data from the LMSYS Chatbot Arena platform, we find that the win-rates derived from pairwise comparisons by a strong large language model to prompts differ under coupled and vanilla autoregressive generation.
Authors: Prashant Garg
Abstract: Cass Sunstein's essay 'On Bob Dylan' describes Dylan's 'dishabituating' style -- a constant refusal to conform to expectation and a penchant for reinventing his musical and lyrical identity. In this paper, I extend Sunstein's observations through a large-scale computational analysis of Dylan's lyrics from 1962 to 2012. Using o3-mini-high (a large language model), I extract concept-to-concept relationships from the lyrics and construct directed knowledge graphs that capture Dylan's thematic structure. I then quantify shifts in sentiment, metaphorical expression, thematic diversity, and network complexity over time. The results indicate that Dylan's lyrics increasingly rely on metaphor, display an evolving sentiment profile, and exhibit heightened dishabituation -- measured here as a growing variance in the network centrality of key concepts. I also find that references to movement, protest, and mythic imagery fluctuate in ways that align with well-known phases of Dylan's career, reflecting the dynamic and unpredictable quality of his art. These findings not only deepen our empirical understanding of Sunstein's thesis but also introduce a novel computational method for analyzing an artist's evolution-offering broader applicability to the study of cultural and creative change.
Authors: Diyana Muhammed, Gollam Rabby, S\"oren Auer
Abstract: Detecting hallucinations in Large Language Models (LLMs) remains a critical challenge for their reliable deployment in real-world applications. To address this, we introduce SelfCheckAgent, a novel framework integrating three different agents: the Symbolic Agent, the Specialized Detection Agent, and the Contextual Consistency Agent. These agents provide a robust multi-dimensional approach to hallucination detection. Notable results include the Contextual Consistency Agent leveraging Llama 3.1 with Chain-of-Thought (CoT) to achieve outstanding performance on the WikiBio dataset, with NonFactual hallucination detection scoring 93.64%, Factual 70.26%, and Ranking 78.48% respectively. On the AIME dataset, GPT-4o with CoT excels in NonFactual detection with 94.89% but reveals trade-offs in Factual with 30.58% and Ranking with 30.68%, underscoring the complexity of hallucination detection in the complex mathematical domains. The framework also incorporates a triangulation strategy, which increases the strengths of the SelfCheckAgent, yielding significant improvements in real-world hallucination identification. The comparative analysis demonstrates SelfCheckAgent's applicability across diverse domains, positioning it as a crucial advancement for trustworthy LLMs. These findings highlight the potentiality of consistency-driven methodologies in detecting hallucinations in LLMs.
Authors: Ziad Shaker, Brendan Ashdown, Hugo Fitzalan, Alistair Heathcote, Jocasta Huntington
Abstract: Generating semantically coherent text requires a robust internal representation of linguistic structures, which traditional embedding techniques often fail to capture adequately. A novel approach, Latent Lexical Projection (LLP), is introduced to refine lexical representations through a structured transformation into a latent space, thereby enhancing the alignment between input embeddings and their contextual meanings. The method integrates an optimized projection mechanism within an existing language model architecture, enabling more accurate token selection while maintaining syntactic integrity. Evaluations across multiple benchmarks indicate a reduction in perplexity and an increase in BLEU scores, suggesting improvements in predictive accuracy and fluency. The analysis of lexical diversity reveals a more varied vocabulary in generated text, addressing common issues of redundancy and repetitive phrase structures. Further assessments of entropy distributions demonstrate a decline in uncertainty during decoding, reflecting enhanced confidence in word selection. Additionally, long-range dependency retention exhibits measurable gains, with increased classification accuracy at extended token distances. Computational efficiency remains within manageable constraints, despite the added projection mechanism, highlighting the practicality of LLP for integration into existing architectures.
Authors: Oliver Kramer
Abstract: We introduce Conceptual Metaphor Theory (CMT) as a framework for enhancing large language models (LLMs) through cognitive prompting in complex reasoning tasks. CMT leverages metaphorical mappings to structure abstract reasoning, improving models' ability to process and explain intricate concepts. By incorporating CMT-based prompts, we guide LLMs toward more structured and human-like reasoning patterns. To evaluate this approach, we compare four native models (Llama3.2, Phi3, Gemma2, and Mistral) against their CMT-augmented counterparts on benchmark tasks spanning domain-specific reasoning, creative insight, and metaphor interpretation. Responses were automatically evaluated using the Llama3.3 70B model. Experimental results indicate that CMT prompting significantly enhances reasoning accuracy, clarity, and metaphorical coherence, outperforming baseline models across all evaluated tasks.
Authors: Avery Ma, Yangchen Pan, Amir-massoud Farahmand
Abstract: Many-shot jailbreaking circumvents the safety alignment of large language models by exploiting their ability to process long input sequences. To achieve this, the malicious target prompt is prefixed with hundreds of fabricated conversational turns between the user and the model. These fabricated exchanges are randomly sampled from a pool of malicious questions and responses, making it appear as though the model has already complied with harmful instructions. In this paper, we present PANDAS: a hybrid technique that improves many-shot jailbreaking by modifying these fabricated dialogues with positive affirmations, negative demonstrations, and an optimized adaptive sampling method tailored to the target prompt's topic. Extensive experiments on AdvBench and HarmBench, using state-of-the-art LLMs, demonstrate that PANDAS significantly outperforms baseline methods in long-context scenarios. Through an attention analysis, we provide insights on how long-context vulnerabilities are exploited and show how PANDAS further improves upon many-shot jailbreaking.
Authors: Xiang Liu, Zhenheng Tang, Hong Chen, Peijie Dong, Zeyu Li, Xiuze Zhou, Bo Li, Xuming Hu, Xiaowen Chu
Abstract: This paper investigates an under-explored challenge in large language models (LLMs): the impact of KV cache compression methods on LLMs' fundamental capabilities. While existing methods achieve impressive compression ratios on long-context benchmarks, their effects on core model capabilities remain understudied. We present a comprehensive empirical study evaluating prominent KV cache compression methods across diverse tasks, spanning world knowledge, commonsense reasoning, arithmetic reasoning, code generation, safety, and long-context understanding and generation.Our analysis reveals that KV cache compression methods exhibit task-specific performance degradation. Arithmetic reasoning tasks prove particularly sensitive to aggressive compression, with different methods showing performance drops of $17.4\%$-$43.3\%$. Notably, the DeepSeek R1 Distill model exhibits more robust compression tolerance compared to instruction-tuned models, showing only $9.67\%$-$25.53\%$ performance degradation. Based on our analysis of attention patterns and cross-task compression performance, we propose ShotKV, a novel compression approach that distinctly handles prefill and decoding phases while maintaining shot-level semantic coherence. Empirical results show that ShotKV achieves $9\%$-$18\%$ performance improvements on long-context generation tasks under aggressive compression ratios.
Authors: Qingling Li, Wushao Wen, Jinghui Qin
Abstract: The Aspect Sentiment Triplet Extraction (ASTE) task aims to extract aspect terms, opinion terms, and their corresponding sentiment polarity from a given sentence. It remains one of the most prominent subtasks in fine-grained sentiment analysis. Most existing approaches frame triplet extraction as a 2D table-filling process in an end-to-end manner, focusing primarily on word-level interactions while often overlooking sentence-level representations. This limitation hampers the model's ability to capture global contextual information, particularly when dealing with multi-word aspect and opinion terms in complex sentences. To address these issues, we propose boundary-driven table-filling with cross-granularity contrastive learning (BTF-CCL) to enhance the semantic consistency between sentence-level representations and word-level representations. By constructing positive and negative sample pairs, the model is forced to learn the associations at both the sentence level and the word level. Additionally, a multi-scale, multi-granularity convolutional method is proposed to capture rich semantic information better. Our approach can capture sentence-level contextual information more effectively while maintaining sensitivity to local details. Experimental results show that the proposed method achieves state-of-the-art performance on public benchmarks according to the F1 score.
Authors: Jinlong Pang, Na Di, Zhaowei Zhu, Jiaheng Wei, Hao Cheng, Chen Qian, Yang Liu
Abstract: Recent studies show that in supervised fine-tuning (SFT) of large language models (LLMs), data quality matters more than quantity. While most data cleaning methods concentrate on filtering entire samples, the quality of individual tokens within a sample can vary significantly. After pre-training, even in high-quality samples, patterns or phrases that are not task-related can be redundant or uninformative. Continuing to fine-tune on these patterns may offer limited benefit and even degrade downstream task performance. In this paper, we investigate token quality from a noisy-label perspective and propose a generic token cleaning pipeline for SFT tasks. Our method filters out uninformative tokens while preserving those carrying key task-specific information. Specifically, we first evaluate token quality by examining the influence of model updates on each token, then apply a threshold-based separation. The token influence can be measured in a single pass with a fixed reference model or iteratively with self-evolving reference models. The benefits and limitations of both methods are analyzed theoretically by error upper bounds. Extensive experiments show that our framework consistently improves performance across multiple downstream tasks.
Authors: Wenhao Zheng, Yixiao Chen, Weitong Zhang, Souvik Kundu, Yun Li, Zhengzhong Liu, Eric P. Xing, Hongyi Wang, Huaxiu Yao
Abstract: Large language models have achieved remarkable success in various tasks but suffer from high computational costs during inference, limiting their deployment in resource-constrained applications. To address this issue, we propose a novel CITER (\textbf{C}ollaborative \textbf{I}nference with \textbf{T}oken-l\textbf{E}vel \textbf{R}outing) framework that enables efficient collaboration between small and large language models (SLMs & LLMs) through a token-level routing strategy. Specifically, CITER routes non-critical tokens to an SLM for efficiency and routes critical tokens to an LLM for generalization quality. We formulate router training as a policy optimization, where the router receives rewards based on both the quality of predictions and the inference costs of generation. This allows the router to learn to predict token-level routing scores and make routing decisions based on both the current token and the future impact of its decisions. To further accelerate the reward evaluation process, we introduce a shortcut which significantly reduces the costs of the reward estimation and improving the practicality of our approach. Extensive experiments on five benchmark datasets demonstrate that CITER reduces the inference costs while preserving high-quality generation, offering a promising solution for real-time and resource-constrained applications.
Authors: Derek Yotheringhay, Beatrix Nightingale, Maximilian Featherstone, Edmund Worthington, Hugo Ashdown
Abstract: Generating structured textual content requires mechanisms that enforce coherence, stability, and adherence to predefined constraints while maintaining semantic fidelity. Conventional approaches often rely on rule-based heuristics or fine-tuning strategies that lack flexibility and generalizability across diverse tasks. The incorporation of Gradient-Regularized Latent Space Modulation (GRLSM) introduces a novel paradigm for guiding text generation through the application of structured constraints within the latent space. The integration of gradient-based regularization mitigates abrupt variations in latent representations, ensuring a smoother encoding process that enhances structural consistency and logical progression within generated sequences. Comparative evaluations demonstrate that latent space modulation leads to a reduction in perplexity, increased coherence scores, and improved structural alignment across multiple domains. Stability assessments further indicate that the imposition of spectral norm constraints facilitates more controlled variations in generated text, preserving semantic consistency under input perturbations. Empirical results confirm that structured latent space constraints not only refine the organization of generated outputs but also enhance interpretability through more predictable and reliable synthesis patterns. Performance metrics illustrate that the GRLSM framework substantially reduces structural inconsistencies while preserving the generative flexibility inherent in neural models.
Authors: Tunazzina Islam, Dan Goldwasser
Abstract: Nowadays, social media is pivotal in shaping public discourse, especially on polarizing issues like vaccination, where diverse moral perspectives influence individual opinions. In NLP, data scarcity and complexity of psycholinguistic tasks such as identifying morality frames makes relying solely on human annotators costly, time-consuming, and prone to inconsistency due to cognitive load. To address these issues, we leverage large language models (LLMs), which are adept at adapting new tasks through few-shot learning, utilizing a handful of in-context examples coupled with explanations that connect examples to task principles. Our research explores LLMs' potential to assist human annotators in identifying morality frames within vaccination debates on social media. We employ a two-step process: generating concepts and explanations with LLMs, followed by human evaluation using a "think-aloud" tool. Our study shows that integrating LLMs into the annotation process enhances accuracy, reduces task difficulty, lowers cognitive load, suggesting a promising avenue for human-AI collaboration in complex psycholinguistic tasks.
Authors: Yui Oka, Taku Hasegawa, Kyosuke Nishida, Kuniko Saito
Abstract: In the realm of large-scale language models, a significant challenge arises when extrapolating sequences beyond the maximum allowable length. This is because the model's position embedding mechanisms are limited to positions encountered during training, thus preventing effective representation of positions in longer sequences. We analyzed conventional position encoding methods for long contexts and found the following characteristics. (1) When the representation dimension is regarded as the time axis, Rotary Position Embedding (RoPE) can be interpreted as a restricted wavelet transform using Haar-like wavelets. However, because it uses only a fixed scale parameter, it does not fully exploit the advantages of wavelet transforms, which capture the fine movements of non-stationary signals using multiple scales (window sizes). This limitation could explain why RoPE performs poorly in extrapolation. (2) Previous research as well as our own analysis indicates that Attention with Linear Biases (ALiBi) functions similarly to windowed attention, using windows of varying sizes. However, it has limitations in capturing deep dependencies because it restricts the receptive field of the model. From these insights, we propose a new position representation method that captures multiple scales (i.e., window sizes) by leveraging wavelet transforms without limiting the model's attention field. Experimental results show that this new method improves the performance of the model in both short and long contexts. In particular, our method allows extrapolation of position information without limiting the model's attention field.
Authors: Pengxiao Lin, Zhongwang Zhang, Zhi-Qin John Xu
Abstract: Since the inception of Large Language Models (LLMs), the quest to efficiently train them for superior reasoning capabilities has been a pivotal challenge. The dominant training paradigm for LLMs is based on next token prediction (NTP). Alternative methodologies, called Critical Token Prediction (CTP), focused exclusively on specific critical tokens (such as the answer in Q\&A dataset), aiming to reduce the overfitting of extraneous information and noise. Contrary to initial assumptions, our research reveals that despite NTP's exposure to noise during training, it surpasses CTP in reasoning ability. We attribute this counterintuitive outcome to the regularizing influence of noise on the training dynamics. Our empirical analysis shows that NTP-trained models exhibit enhanced generalization and robustness across various benchmark reasoning datasets, demonstrating greater resilience to perturbations and achieving flatter loss minima. These findings illuminate that NTP is instrumental in fostering reasoning abilities during pretraining, whereas CTP is more effective for finetuning, thereby enriching our comprehension of optimal training strategies in LLM development.
Authors: Anneketh Vij, Changhao Liu, Rahul Anil Nair, Theo Ho, Edward Shi, Ayan Bhowmick
Abstract: This research presents an exploration and study of the recipe generation task by fine-tuning various very small language models, with a focus on developing robust evaluation metrics and comparing across different language models the open-ended task of recipe generation. This study presents extensive experiments with multiple model architectures, ranging from T5-small (Raffel et al., 2023) and SmolLM-135M (Allal et al., 2024) to Phi-2 (Research, 2023),implementing both traditional NLP metrics and custom domain-specific evaluation metrics. Our novel evaluation framework incorporates recipe-specific metrics for assessing content quality and introduces an approach to allergen substitution. The results indicate that, while larger models generally perform better on standard metrics, the relationship between model size and recipe quality is more nuanced when considering domain-specific metrics. We find that SmolLM-360M and SmolLM-1.7B demonstrate comparable performance despite their size difference, while Phi-2 shows limitations in recipe generation despite its larger parameter count. Our comprehensive evaluation framework and allergen substitution system provide valuable insights for future work in recipe generation and broader NLG tasks that require domain expertise and safety considerations.
Authors: Nikhil Bhendawade, Mahyar Najibi, Devang Naik, Irina Belousova
Abstract: Residual transformations enhance the representational depth and expressive power of large language models (LLMs). However, applying static residual transformations across all tokens in auto-regressive generation leads to a suboptimal trade-off between inference efficiency and generation fidelity. Existing methods, including Early Exiting, Skip Decoding, and Mixture-of-Depth address this by modulating the residual transformation based on token-level complexity. Nevertheless, these approaches predominantly consider the distance traversed by tokens through the model layers, neglecting the underlying velocity of residual evolution. We introduce Mixture of Multi-rate Residuals (M2R2), a framework that dynamically modulates residual velocity to improve early alignment, enhancing inference efficiency. Evaluations on reasoning oriented tasks such as Koala, Self-Instruct, WizardLM, and MT-Bench show M2R2 surpasses state-of-the-art distance-based strategies, balancing generation quality and speedup. In self-speculative decoding setup, M2R2 achieves up to 2.8x speedups on MT-Bench, outperforming methods like 2-model speculative decoding, Medusa, LookAhead Decoding, and DEED. In Mixture-of-Experts (MoE) architectures, integrating early residual alignment with ahead-of-time expert loading into high-bandwidth memory (HBM) accelerates decoding, reduces expert-switching bottlenecks, and achieves a 2.9x speedup, making it highly effective in resource-constrained environments.
Authors: Frederick Dillon, Gregor Halvorsen, Simon Tattershall, Magnus Rowntree, Gareth Vanderpool
Abstract: Memory retention challenges in deep neural architectures have ongoing limitations in the ability to process and recall extended contextual information. Token dependencies degrade as sequence length increases, leading to a decline in coherence and factual consistency across longer outputs. A structured approach is introduced to mitigate this issue through the reweaving of latent states captured at different processing layers, reinforcing token representations over extended sequences. The proposed Contextual Memory Reweaving framework incorporates a Layered Latent State Reconstruction mechanism to systematically integrate past contextual embeddings without introducing external memory modules. Experimental results demonstrate improvements in recall accuracy across a range of sequence lengths, with notable gains in the retention of rarely occurring tokens and numerical reasoning consistency. Further analysis of computational efficiency indicates that the additional processing overhead remains within acceptable thresholds, enabling scalability across different model sizes. Evaluations in long-form text generation and ambiguous query resolution highlight the capacity of memory reweaving to enhance continuity and reduce inconsistencies over extended outputs. Attention weight distributions reveal more structured allocation patterns, suggesting that reweaved latent states contribute to improved contextual awareness. The findings establish a framework for refining memory retention mechanisms in language models, addressing long-standing challenges in handling complex, multi-step reasoning tasks.
Authors: Nebiyou Daniel Hailemariam, Blessed Guda, Tsegazeab Tefferi
Abstract: This research presents a novel framework for translating extractive question-answering datasets into low-resource languages, as demonstrated by the creation of the AmaSQuAD dataset, a translation of SQuAD 2.0 into Amharic. The methodology addresses challenges related to misalignment between translated questions and answers, as well as the presence of multiple answer instances in the translated context. For this purpose, we used cosine similarity utilizing embeddings from a fine-tuned BERT-based model for Amharic and Longest Common Subsequence (LCS). Additionally, we fine-tune the XLM-R model on the AmaSQuAD synthetic dataset for Amharic Question-Answering. The results show an improvement in baseline performance, with the fine-tuned model achieving an increase in the F1 score from 36.55% to 44.41% and 50.01% to 57.5% on the AmaSQuAD development dataset. Moreover, the model demonstrates improvement on the human-curated AmQA dataset, increasing the F1 score from 67.80% to 68.80% and the exact match score from 52.50% to 52.66%.The AmaSQuAD dataset is publicly available Datasets
Authors: Rajiv Bahl, Venkatesan N, Parimal Aglawe, Aastha Sarasapalli, Bhavya Kancharla, Chaitanya kolukuluri, Harish Mohite, Japneet Hora, Kiran Kakollu, Rahul Diman, Shubham Kapale, Sri Bhagya Kathula, Vamsikrishna Motru, Yogeshwar Reddy
Abstract: The rapid evolution of Large Language Models (LLMs) has transformed natural language processing but raises critical concerns about biases inherent in their deployment and use across diverse linguistic and sociocultural contexts. This paper presents a framework named ASCenD BDS (Adaptable, Stochastic and Context-aware framework for Detection of Bias, Discrimination and Stereotyping). The framework presents approach to detecting bias, discrimination, stereotyping across various categories such as gender, caste, age, disability, socioeconomic status, linguistic variations, etc., using an approach which is Adaptive, Stochastic and Context-Aware. The existing frameworks rely heavily on usage of datasets to generate scenarios for detection of Bias, Discrimination and Stereotyping. Examples include datasets such as Civil Comments, Wino Gender, WinoBias, BOLD, CrowS Pairs and BBQ. However, such an approach provides point solutions. As a result, these datasets provide a finite number of scenarios for assessment. The current framework overcomes this limitation by having features which enable Adaptability, Stochasticity, Context Awareness. Context awareness can be customized for any nation or culture or sub-culture (for example an organization's unique culture). In this paper, context awareness in the Indian context has been established. Content has been leveraged from Indian Census 2011 to have a commonality of categorization. A framework has been developed using Category, Sub-Category, STEM, X-Factor, Synonym to enable the features for Adaptability, Stochasticity and Context awareness. The framework has been described in detail in Section 3. Overall 800 plus STEMs, 10 Categories, 31 unique SubCategories were developed by a team of consultants at Saint Fox Consultancy Private Ltd. The concept has been tested out in SFCLabs as part of product development.
Authors: Prasanta Bhattacharya, Hong Zhang, Yiming Cao, Wei Gao, Brandon Siyuan Loh, Joseph J. P. Simons, Liang Ze Wong
Abstract: Stance detection has emerged as a popular task in natural language processing research, enabled largely by the abundance of target-specific social media data. While there has been considerable research on the development of stance detection models, datasets, and application, we highlight important gaps pertaining to (i) a lack of theoretical conceptualization of stance, and (ii) the treatment of stance at an individual- or user-level, as opposed to message-level. In this paper, we first review the interdisciplinary origins of stance as an individual-level construct to highlight relevant attributes (e.g., psychological features) that might be useful to incorporate in stance detection models. Further, we argue that recent pre-trained and large language models (LLMs) might offer a way to flexibly infer such user-level attributes and/or incorporate them in modelling stance. To better illustrate this, we briefly review and synthesize the emerging corpus of studies on using LLMs for inferring stance, and specifically on incorporating user attributes in such tasks. We conclude by proposing a four-point agenda for pursuing stance detection research that is theoretically informed, inclusive, and practically impactful.
Authors: Bowen Ping, Jiali Zeng, Fandong Meng, Shuo Wang, Jie Zhou, Shanghang Zhang
Abstract: Long-form generation is crucial for academic writing papers and repo-level code generation. Despite this, current models, including GPT-4o, still exhibit unsatisfactory performance. Existing methods that utilize preference learning with outcome supervision often fail to provide detailed feedback for extended contexts. This shortcoming can lead to content that does not fully satisfy query requirements, resulting in issues like length deviations, and diminished quality. In this paper, we propose enhancing long-form generation by incorporating process supervision. We employ Monte Carlo Tree Search to gather stepwise preference pairs, utilizing a global memory pool to maintain consistency. To address the issue of suboptimal candidate selection, we integrate external critiques to refine and improve the quality of the preference pairs. Finally, we apply step-level DPO using the collected stepwise preference pairs. Experimental results show that our method improves length and quality on long-form generation benchmarks, with almost lossless performance on general benchmarks across various model backbones.
Authors: Sanket Shinde, Raviraj Joshi
Abstract: While topic modeling in English has become a prevalent and well-explored area, venturing into topic modeling for Indic languages remains relatively rare. The limited availability of resources, diverse linguistic structures, and unique challenges posed by Indic languages contribute to the scarcity of research and applications in this domain. Despite the growing interest in natural language processing and machine learning, there exists a noticeable gap in the comprehensive exploration of topic modeling methodologies tailored specifically for languages such as Hindi, Marathi, Tamil, and others. In this paper, we examine several topic modeling approaches applied to the Marathi language. Specifically, we compare various BERT and non-BERT approaches, including multilingual and monolingual BERT models, using topic coherence and topic diversity as evaluation metrics. Our analysis provides insights into the performance of these approaches for Marathi language topic modeling. The key finding of the paper is that BERTopic, when combined with BERT models trained on Indic languages, outperforms LDA in terms of topic modeling performance.
Authors: Pavel Bedrin, Maksim Varlamov, Alexander Yatskov
Abstract: This paper addresses the challenge of automatically extracting attributes from news article web pages across multiple languages. Recent neural network models have shown high efficacy in extracting information from semi-structured web pages. However, these models are predominantly applied to domains like e-commerce and are pre-trained using English data, complicating their application to web pages in other languages. We prepared a multilingual dataset comprising 3,172 marked-up news web pages across six languages (English, German, Russian, Chinese, Korean, and Arabic) from 161 websites. The dataset is publicly available on GitHub. We fine-tuned the pre-trained state-of-the-art model, MarkupLM, to extract news attributes from these pages and evaluated the impact of translating pages into English on extraction quality. Additionally, we pre-trained another state-of-the-art model, DOM-LM, on multilingual data and fine-tuned it on our dataset. We compared both fine-tuned models to existing open-source news data extraction tools, achieving superior extraction metrics.
Authors: Daniel Tamayo, Aitor Gonzalez-Agirre, Javier Hernando, Marta Villegas
Abstract: Recent research has explored methods for updating and modifying factual knowledge in large language models, often focusing on specific multi-layer perceptron blocks. This study expands on this work by examining the effectiveness of existing knowledge editing methods across languages and delving into the role of attention mechanisms in this process. Drawing from the insights gained, we propose Mass-Editing Memory with Attention in Transformers (MEMAT), a method that achieves significant improvements in all metrics while requiring minimal parameter modifications. MEMAT delivers a remarkable 10% increase in magnitude metrics, benefits languages not included in the training data and also demonstrates a high degree of portability. Our code and data are at https://github.com/dtamayo-nlp/MEMAT.
Authors: Felix Drinkall, Janet B. Pierrehumbert, Stefan Zohren
Abstract: Large language models (LLMs) have shown remarkable success in language modelling due to scaling laws found in model size and the hidden dimension of the model's text representation. Yet, we demonstrate that compressed representations of text can yield better performance in LLM-based regression tasks. In this paper, we compare the relative performance of embedding compression in three different signal-to-noise contexts: financial return prediction, writing quality assessment and review scoring. Our results show that compressing embeddings, in a minimally supervised manner using an autoencoder's hidden representation, can mitigate overfitting and improve performance on noisy tasks, such as financial return prediction; but that compression reduces performance on tasks that have high causal dependencies between the input and target data. Our results suggest that the success of interpretable compressed representations such as sentiment may be due to a regularising effect.
Authors: Atharva Mangeshkumar Agrawal, Rutika Pandurang Shinde, Vasanth Kumar Bhukya, Ashmita Chakraborty, Sagar Bharat Shah, Tanmay Shukla, Sree Pradeep Kumar Relangi, Nilesh Mutyam
Abstract: Large language models (LLMs) have shown impressive capabilities in natural language processing tasks, including dialogue generation. This research aims to conduct a novel comparative analysis of two prominent techniques, fine-tuning with LoRA (Low-Rank Adaptation) and the Retrieval-Augmented Generation (RAG) framework, in the context of doctor-patient chat conversations with multiple datasets of mixed medical domains. The analysis involves three state-of-the-art models: Llama-2, GPT, and the LSTM model. Employing real-world doctor-patient dialogues, we comprehensively evaluate the performance of models, assessing key metrics such as language quality (perplexity, BLEU score), factual accuracy (fact-checking against medical knowledge bases), adherence to medical guidelines, and overall human judgments (coherence, empathy, safety). The findings provide insights into the strengths and limitations of each approach, shedding light on their suitability for healthcare applications. Furthermore, the research investigates the robustness of the models in handling diverse patient queries, ranging from general health inquiries to specific medical conditions. The impact of domain-specific knowledge integration is also explored, highlighting the potential for enhancing LLM performance through targeted data augmentation and retrieval strategies.
Authors: Bernardo Magnini, Roberto Zanoli, Michele Resta, Martin Cimmino, Paolo Albano, Marco Madeddu, Viviana Patti
Abstract: We describe Evalita-LLM, a new benchmark designed to evaluate Large Language Models (LLMs) on Italian tasks. The distinguishing and innovative features of Evalita-LLM are the following: (i) all tasks are native Italian, avoiding issues of translating from Italian and potential cultural biases; (ii) in addition to well established multiple-choice tasks, the benchmark includes generative tasks, enabling more natural interaction with LLMs; (iii) all tasks are evaluated against multiple prompts, this way mitigating the model sensitivity to specific prompts and allowing a fairer and objective evaluation. We propose an iterative methodology, where candidate tasks and candidate prompts are validated against a set of LLMs used for development. We report experimental results from the benchmark's development phase, and provide performance statistics for several state-of-the-art LLMs.
Authors: Jinyang Wu, Mingkuan Feng, Shuai Zhang, Ruihan Jin, Feihu Che, Zengqi Wen, Jianhua Tao
Abstract: Multimodal large language models (MLLMs) exhibit impressive capabilities but still face challenges in complex visual reasoning. While recent efforts attempt to enhance MLLMs' reasoning by incorporating OpenAI o1-like structured thinking through explicit search structures or teacher-guided distillation, they often struggle to balance performance and efficiency. A critical limitation is their heavy reliance on extensive data and search spaces, resulting in low-efficiency implicit insight extraction and data utilization. To address this, we propose AStar, an Automated Structured thinking paradigm for multimodal reasoning via Monte Carlo Tree Search (MCTS). AStar automatically derives high-level cognitive reasoning patterns from limited data using MCTS-powered hierarchical structures. Building on these explicit patterns, we design a unified reasoning framework that seamlessly integrates models' internal reasoning capabilities and external reasoning guidelines, enabling efficient inference with minimal tree iterations. This novel paradigm strikes a compelling balance between performance and efficiency. Extensive experiments demonstrate AStar's effectiveness, achieving superior accuracy (54.0$\%$) on the MathVerse benchmark with a 7B backbone, surpassing GPT-4o (50.2$\%$) while maintaining substantial data and computational efficiency.
Authors: Sagnik Mukherjee, Abhinav Chinta, Takyoung Kim, Tarun Anoop Sharma, Dilek Hakkani Tur
Abstract: Chain-of-Thought (CoT) prompting enhances mathematical reasoning in large language models (LLMs) by enabling detailed step-by-step solutions. However, due to the verbosity of LLMs, the resulting reasoning chains can be long, making it harder to verify the reasoning steps and trace issues resulting from dependencies between the steps that may be farther away in the sequence of steps. Importantly, mathematical reasoning allows each step to be derived from a small set of premises, which are a subset of the preceding steps in the reasoning chain. In this paper, we present a framework that identifies the premises for each step, to improve the evaluation of reasoning. We restructure conventional linear reasoning chains into Premise Augmented Reasoning Chains (PARC) by introducing premise links, resulting in a directed acyclic graph where the nodes are the steps and the edges are the premise links. Through experiments with a PARC-based dataset that we built, namely PERL (Premises and ERrors identification in LLMs), we demonstrate that LLMs can reliably identify premises within complex reasoning chains. In particular, even open-source LLMs achieve 90% recall in premise identification. We also show that PARC helps to identify errors in reasoning chains more reliably. The accuracy of error identification improves by 6% to 16% absolute when step-by-step verification is carried out in PARC under the premises. Our findings highlight the utility of premise-centric representations in addressing complex problem-solving tasks and open new avenues for improving the reliability of LLM-based reasoning evaluations.
Authors: Yichi Zhang, Siyuan Zhang, Yao Huang, Zeyu Xia, Zhengwei Fang, Xiao Yang, Ranjie Duan, Dong Yan, Yinpeng Dong, Jun Zhu
Abstract: Ensuring the safety and harmlessness of Large Language Models (LLMs) has become equally critical as their performance in applications. However, existing safety alignment methods typically suffer from safety-performance trade-offs and the susceptibility to jailbreak attacks, primarily due to their reliance on direct refusals for malicious queries. In this paper, we propose STAIR, a novel framework that integrates SafeTy Alignment with Itrospective Reasoning. We enable LLMs to identify safety risks through step-by-step analysis by self-improving chain-of-thought (CoT) reasoning with safety awareness. STAIR first equips the model with a structured reasoning capability and then advances safety alignment via iterative preference optimization on step-level reasoning data generated using our newly proposed Safety-Informed Monte Carlo Tree Search (SI-MCTS). We further train a process reward model on this data to guide test-time searches for improved responses. Extensive experiments show that STAIR effectively mitigates harmful outputs while better preserving helpfulness, compared to instinctive alignment strategies. With test-time scaling, STAIR achieves a safety performance comparable to Claude-3.5 against popular jailbreak attacks. Relevant resources in this work are available at https://github.com/thu-ml/STAIR.
Authors: Jianfeng Pan, Senyou Deng, Shaomang Huang
Abstract: Research on LLM technologies is rapidly emerging, with most of them employing a 'fast thinking' approach to inference. Most LLMs generate the final result based solely on a single query and LLM's reasoning capabilities. However, with the advent of OpenAI-o1, 'slow thinking' techniques have garnered increasing attention because its process is closer to the human thought process. Inspired by the human ability to constantly associate and replenish knowledge during thinking, we developed the novel Chain-of-Associated-Thoughts (CoAT) framework, which introduces an innovative synergy between the Monte Carlo Tree Search (MCTS) algorithm and a dynamic mechanism for integrating new key information, termed 'associative memory'. By combining the structured exploration capabilities of MCTS with the adaptive learning capacity of associative memory, CoAT significantly expands the LLM search space, enabling our framework to explore diverse reasoning pathways and dynamically update its knowledge base in real-time. This allows the framework to not only revisit and refine earlier inferences but also adaptively incorporate evolving information, ensuring that the final output is both accurate and comprehensive. To validate the effectiveness of our framework, we conducted extensive experiments across a range of generative and reasoning tasks. These experiments demonstrated that our framework outperforms conventional inference processes on accuracy, coherence, and diversity. The framework's ability to iteratively expand its search space while retaining contextually relevant information results.
Authors: Ibrahim Bouabdallaoui, Fatima Guerouate, Samya Bouhaddour, Chaimae Saadi, Mohammed Sbihi
Abstract: We introduce FewTopNER, a novel framework that integrates few-shot named entity recognition (NER) with topic-aware contextual modeling to address the challenges of cross-lingual and low-resource scenarios. FewTopNER leverages a shared multilingual encoder based on XLM-RoBERTa, augmented with language-specific calibration mechanisms, to generate robust contextual embeddings. The architecture comprises a prototype-based entity recognition branch, employing BiLSTM and Conditional Random Fields for sequence labeling, and a topic modeling branch that extracts document-level semantic features through hybrid probabilistic and neural methods. A cross-task bridge facilitates dynamic bidirectional attention and feature fusion between entity and topic representations, thereby enhancing entity disambiguation by incorporating global semantic context. Empirical evaluations on multilingual benchmarks across English, French, Spanish, German, and Italian demonstrate that FewTopNER significantly outperforms existing state-of-the-art few-shot NER models. In particular, the framework achieves improvements of 2.5-4.0 percentage points in F1 score and exhibits enhanced topic coherence, as measured by normalized pointwise mutual information. Ablation studies further confirm the critical contributions of the shared encoder and cross-task integration mechanisms to the overall performance. These results underscore the efficacy of incorporating topic-aware context into few-shot NER and highlight the potential of FewTopNER for robust cross-lingual applications in low-resource settings.
Authors: Amin Heyrani Nobari, Kaveh Alimohammadi, Ali ArjomandBigdeli, Akash Srivastava, Faez Ahmed, Navid Azizan
Abstract: Model merging, a method that combines the parameters and embeddings of multiple fine-tuned large language models (LLMs), offers a promising approach to enhance model performance across various tasks while maintaining computational efficiency. This paper introduces Activation-Informed Merging (AIM), a technique that integrates the information from the activation space of LLMs into the merging process to improve performance and robustness. AIM is designed as a flexible, complementary solution that is applicable to any existing merging method. It aims to preserve critical weights from the base model, drawing on principles from continual learning~(CL) and model compression. Utilizing a task-agnostic calibration set, AIM selectively prioritizes essential weights during merging. We empirically demonstrate that AIM significantly enhances the performance of merged models across multiple benchmarks. Our findings suggest that considering the activation-space information can provide substantial advancements in the model merging strategies for LLMs with up to 40\% increase in benchmark performance.
Authors: Haoran Ye, Tianze Zhang, Yuhang Xie, Liyuan Zhang, Yuanyi Ren, Xin Zhang, Guojie Song
Abstract: Values are core drivers of individual and collective perception, cognition, and behavior. Value systems, such as Schwartz's Theory of Basic Human Values, delineate the hierarchy and interplay among these values, enabling cross-disciplinary investigations into decision-making and societal dynamics. Recently, the rise of Large Language Models (LLMs) has raised concerns regarding their elusive intrinsic values. Despite growing efforts in evaluating, understanding, and aligning LLM values, a psychologically grounded LLM value system remains underexplored. This study addresses the gap by introducing the Generative Psycho-Lexical Approach (GPLA), a scalable, adaptable, and theoretically informed method for constructing value systems. Leveraging GPLA, we propose a psychologically grounded five-factor value system tailored for LLMs. For systematic validation, we present three benchmarking tasks that integrate psychological principles with cutting-edge AI priorities. Our results reveal that the proposed value system meets standard psychological criteria, better captures LLM values, improves LLM safety prediction, and enhances LLM alignment, when compared to the canonical Schwartz's values.
Authors: Calvin Yixiang Cheng, Scott A Hale
Abstract: This study explores computational approaches for measuring moral foundations (MFs) in non-English corpora. Since most resources are developed primarily for English, cross-linguistic applications of moral foundation theory remain limited. Using Chinese as a case study, this paper evaluates the effectiveness of applying English resources to machine translated text, local language lexicons, multilingual language models, and large language models (LLMs) in measuring MFs in non-English texts. The results indicate that machine translation and local lexicon approaches are insufficient for complex moral assessments, frequently resulting in a substantial loss of cultural information. In contrast, multilingual models and LLMs demonstrate reliable cross-language performance with transfer learning, with LLMs excelling in terms of data efficiency. Importantly, this study also underscores the need for human-in-the-loop validation of automated MF assessment, as the most advanced models may overlook cultural nuances in cross-language measurements. The findings highlight the potential of LLMs for cross-language MF measurements and other complex multilingual deductive coding tasks.
Authors: Qianhao Yuan, Yanjiang Liu, Yaojie Lu, Hongyu Lin, Ben He, Xianpei Han, Le Sun
Abstract: Multimodal Large Language Models (MLLMs) mainly fall into two architectures, each involving a trade-off between training and inference efficiency: embedding space alignment (e.g., LLaVA-1.5) is inefficient during inference, while cross-attention space alignment (e.g., Flamingo) is inefficient in training. In this paper, we compare these two architectures and identify the key factors for building efficient MLLMs. A primary difference between them lies in how attention is applied to visual tokens, particularly in their interactions with each other. To investigate whether attention among visual tokens is necessary, we propose a new self-attention mechanism, NAAViT (\textbf{N}o \textbf{A}ttention \textbf{A}mong \textbf{Vi}sual \textbf{T}okens), which eliminates this type of attention. Our pilot experiment on LLaVA-1.5 shows that attention among visual tokens is highly redundant. Based on these insights, we introduce SAISA (\textbf{S}elf-\textbf{A}ttention \textbf{I}nput \textbf{S}pace \textbf{A}lignment), a novel architecture that enhance both training and inference efficiency. SAISA directly aligns visual features with the input spaces of NAAViT self-attention blocks, reducing computational overhead in both self-attention blocks and feed-forward networks (FFNs). Using the same configuration as LLaVA-1.5, SAISA reduces inference FLOPs by 66\% and training budget by 26\%, while achieving superior performance in terms of accuracy. Comprehensive ablation studies further validate the effectiveness of SAISA across various LLMs and visual encoders. The code and model will be publicly available at https://github.com/icip-cas/SAISA.
Authors: Menglong Cui, Pengzhi Gao, Wei Liu, Jian Luan, BinWang
Abstract: Large language models (LLMs) have shown continuously improving multilingual capabilities, and even small-scale open-source models have demonstrated rapid performance enhancement. In this paper, we systematically explore the abilities of open LLMs with less than ten billion parameters to handle multilingual machine translation (MT) tasks. We conduct comprehensive evaluations on six popular LLMs and find that models like Gemma2-9B exhibit impressive multilingual translation capabilities. We then introduce the Parallel-First Monolingual-Second (PFMS) data mixing strategy in the continual pretraining stage to further enhance the MT performance and present GemmaX2-28, a 9B model achieving top-tier multilingual translation performance across 28 languages. Specifically, GemmaX2-28 consistently outperforms the state-of-the-art (SOTA) models such as TowerInstruct and XALMA and achieves competitive performance with Google Translate and GPT-4-turbo.
Authors: Maohao Shen, Guangtao Zeng, Zhenting Qi, Zhang-Wei Hong, Zhenfang Chen, Wei Lu, Gregory Wornell, Subhro Das, David Cox, Chuang Gan
Abstract: Large language models (LLMs) have demonstrated remarkable reasoning capabilities across diverse domains. Recent studies have shown that increasing test-time computation enhances LLMs' reasoning capabilities. This typically involves extensive sampling at inference time guided by an external LLM verifier, resulting in a two-player system. Despite external guidance, the effectiveness of this system demonstrates the potential of a single LLM to tackle complex tasks. Thus, we pose a new research problem: Can we internalize the searching capabilities to fundamentally enhance the reasoning abilities of a single LLM? This work explores an orthogonal direction focusing on post-training LLMs for autoregressive searching (i.e., an extended reasoning process with self-reflection and self-exploration of new strategies). To achieve this, we propose the Chain-of-Action-Thought (COAT) reasoning and a two-stage training paradigm: 1) a small-scale format tuning stage to internalize the COAT reasoning format and 2) a large-scale self-improvement stage leveraging reinforcement learning. Our approach results in Satori, a 7B LLM trained on open-source models and data. Extensive empirical evaluations demonstrate that Satori achieves state-of-the-art performance on mathematical reasoning benchmarks while exhibits strong generalization to out-of-domain tasks. Code, data, and models will be fully open-sourced.
Authors: Genghan Zhang, Weixin Liang, Olivia Hsu, Kunle Olukotun
Abstract: ML libraries, often written in architecture-specific programming languages (ASPLs) that target domain-specific architectures, are key to efficient ML systems. However, writing these high-performance ML libraries is challenging because it requires expert knowledge of ML algorithms and the ASPL. Large language models (LLMs), on the other hand, have shown general coding capabilities. However, challenges remain when using LLMs for generating ML libraries using ASPLs because 1) this task is complicated even for experienced human programmers and 2) there are limited code examples because of the esoteric and evolving nature of ASPLs. Therefore, LLMs need complex reasoning with limited data in order to complete this task. To address these challenges, we introduce an adaptive self-improvement agentic system. In order to evaluate the effectiveness of our system, we construct a benchmark of a typical ML library and generate ASPL code with both open and closed-source LLMs on this benchmark. Our results show improvements of up to $3.9\times$ over a baseline single LLM.
Authors: Soheil Abbasloo
Abstract: Large Language Models (LLMs) have demonstrated impressive capabilities across numerous fields, presenting an opportunity to revolutionize optimization problem-solving, a crucial, ubiquitous, and complex domain. This paper explores the proficiency of LLMs in handling Sequential Optimization Problems (SOPs). We introduce WorldGen, a dynamic framework for generating unseen SOPs with controllable complexities, to evaluate LLM performance. Our initial observations reveal that while LLMs perform well on simple SOPs, their performance significantly degrades with increased complexity. Motivated by this, we revisit philosophical hypotheses on reasoning to enhance LLM performance. Inspired by the influential framework of Hegelian Dialectics, we propose ACE, demonstrating how the performance of LLMs in SOP contexts can be significantly improved without any retraining or further fine-tuning.
Authors: Alex Fl\"uckiger, Chantal Amrhein, Tim Graf, Philippe Schl\"apfer, Florian Schottmann, Samuel L\"aubli
Abstract: As strong machine translation (MT) systems are increasingly based on large language models (LLMs), reliable quality benchmarking requires methods that capture their ability to leverage extended context. This study compares two commercial MT systems -- DeepL and Supertext -- by assessing their performance on unsegmented texts. We evaluate translation quality across four language directions with professional translators assessing segments with full document-level context. While segment-level assessments indicate no strong preference between the systems in most cases, document-level analysis reveals a preference for Supertext in three out of four language directions, suggesting superior consistency across longer texts. We advocate for more context-sensitive evaluation methodologies to ensure that MT quality assessments reflect real-world usability. We release all evaluation data and scripts for further analysis and reproduction at https://github.com/supertext/evaluation_deepl_supertext.
URLs: https://github.com/supertext/evaluation_deepl_supertext.
Authors: Christian Ruiz, Fabio Martinez
Abstract: Sign Language Translation (SLT) systems support hearing-impaired people communication by finding equivalences between signed and spoken languages. This task is however challenging due to multiple sign variations, complexity in language and inherent richness of expressions. Computational approaches have evidenced capabilities to support SLT. Nonetheless, these approaches remain limited to cover gestures variability and support long sequence translations. This paper introduces a Transformer-based architecture that encodes spatio-temporal motion gestures, preserving both local and long-range spatial information through the use of multiple convolutional and attention mechanisms. The proposed approach was validated on the Colombian Sign Language Translation Dataset (CoL-SLTD) outperforming baseline approaches, and achieving a BLEU4 of 46.84%. Additionally, the proposed approach was validated on the RWTH-PHOENIX-Weather-2014T (PHOENIX14T), achieving a BLEU4 score of 30.77%, demonstrating its robustness and effectiveness in handling real-world variations
Authors: Xiaorui Ma, Haoran Xie, S. Joe Qin
Abstract: The integration of vision-language modalities has been a significant focus in multimodal learning, traditionally relying on Vision-Language Pretrained Models. However, with the advent of Large Language Models (LLMs), there has been a notable shift towards incorporating LLMs with vision modalities. Following this, the training paradigms for incorporating vision modalities into LLMs have evolved. Initially, the approach was to integrate the modalities through pretraining the modality integrator, named Single-stage Tuning. It has since branched out into methods focusing on performance enhancement, denoted as Two-stage Tuning, and those prioritizing parameter efficiency, referred to as Direct Adaptation. However, existing surveys primarily address the latest Vision Large Language Models (VLLMs) with Two-stage Tuning, leaving a gap in understanding the evolution of training paradigms and their unique parameter-efficient considerations. This paper categorizes and reviews 34 VLLMs from top conferences, journals, and highly cited Arxiv papers, focusing on parameter efficiency during adaptation from the training paradigm perspective. We first introduce the architecture of LLMs and parameter-efficient learning methods, followed by a discussion on vision encoders and a comprehensive taxonomy of modality integrators. We then review three training paradigms and their efficiency considerations, summarizing benchmarks in the VLLM field. To gain deeper insights into their effectiveness in parameter efficiency, we compare and discuss the experimental results of representative models, among which the experiment of the Direct Adaptation paradigm is replicated. Providing insights into recent developments and practical uses, this survey is a vital guide for researchers and practitioners navigating the efficient integration of vision modalities into LLMs.
Authors: Bo Pang, Tingrui Qiao, Caroline Walker, Chris Cunningham, Yun Sing Koh
Abstract: Large Language Models (LLMs) have significantly advanced natural language processing applications, yet their widespread use raises concerns regarding inherent biases that may reduce utility or harm for particular social groups. Despite the advancement in addressing LLM bias, existing research has two major limitations. First, existing LLM bias evaluation focuses on the U.S. cultural context, making it challenging to reveal stereotypical biases of LLMs toward other cultures, leading to unfair development and use of LLMs. Second, current bias evaluation often assumes models are familiar with the target social groups. When LLMs encounter words beyond their knowledge boundaries that are unfamiliar in their training data, they produce irrelevant results in the local context due to hallucinations and overconfidence, which are not necessarily indicative of inherent bias. This research addresses these limitations with a Local Integrated Bias Recognition and Assessment Framework (LIBRA) for measuring bias using datasets sourced from local corpora without crowdsourcing. Implementing this framework, we develop a dataset comprising over 360,000 test cases in the New Zealand context. Furthermore, we propose the Enhanced Idealized CAT Score (EiCAT), integrating the iCAT score with a beyond knowledge boundary score (bbs) and a distribution divergence-based bias measurement to tackle the challenge of LLMs encountering words beyond knowledge boundaries. Our results show that the BERT family, GPT-2, and Llama-3 models seldom understand local words in different contexts. While Llama-3 exhibits larger bias, it responds better to different cultural contexts. The code and dataset are available at: https://github.com/ipangbo/LIBRA.
Authors: Si-Ioi Ng, Pranav S. Ambadi, Kimberly D. Mueller, Julie Liss, Visar Berisha
Abstract: Existing methods for analyzing linguistic content from picture descriptions for assessment of cognitive-linguistic impairment often overlook the participant's visual narrative path, which typically requires eye tracking to assess. Spatio-semantic graphs are a useful tool for analyzing this narrative path from transcripts alone, however they are limited by the need for manual tagging of content information units (CIUs). In this paper, we propose an automated approach for estimation of spatio-semantic graphs (via automated extraction of CIUs) from the Cookie Theft picture commonly used in cognitive-linguistic analyses. The method enables the automatic characterization of the visual semantic path during picture description. Experiments demonstrate that the automatic spatio-semantic graphs effectively differentiate between cognitively impaired and unimpaired speakers. Statistical analyses reveal that the features derived by the automated method produce comparable results to the manual method, with even greater group differences between clinical groups of interest. These results highlight the potential of the automated approach for extracting spatio-semantic features in developing clinical speech models for cognitive impairment assessment.
Authors: Tianlin Zhang, En Yu, Yi Shao, Shuai Li, Sujuan Hou, Jiande Sun
Abstract: Multimodal fake news detection has garnered significant attention due to its profound implications for social security. While existing approaches have contributed to understanding cross-modal consistency, they often fail to leverage modal-specific representations and explicit discrepant features. To address these limitations, we propose a Multimodal Inverse Attention Network (MIAN), a novel framework that explores intrinsic discriminative features based on news content to advance fake news detection. Specifically, MIAN introduces a hierarchical learning module that captures diverse intra-modal relationships through local-to-global and local-to-local interactions, thereby generating enhanced unimodal representations to improve the identification of fake news at the intra-modal level. Additionally, a cross-modal interaction module employs a co-attention mechanism to establish and model dependencies between the refined unimodal representations, facilitating seamless semantic integration across modalities. To explicitly extract inconsistency features, we propose an inverse attention mechanism that effectively highlights the conflicting patterns and semantic deviations introduced by fake news in both intra- and inter-modality. Extensive experiments on benchmark datasets demonstrate that MIAN significantly outperforms state-of-the-art methods, underscoring its pivotal contribution to advancing social security through enhanced multimodal fake news detection.
Authors: Moses Ananta, Muhammad Farid Adilazuarda, Zayd Muhammad Kawakibi Zuhri, Ayu Purwarianti, Alham Fikri Aji
Abstract: Fine-tuning large language models (LLMs) is often constrained by the computational costs of processing massive datasets. We propose \textbf{QLESS} (Quantized Low-rank Gradient Similarity Search), which integrates gradient quantization with the LESS framework to enable memory-efficient data valuation and selection. QLESS employs a two-step compression process: first, it obtains low-dimensional gradient representations through LoRA-based random projection; then, it quantizes these gradients to low-bitwidth representations. Experiments on multiple LLM architectures (LLaMA, Mistral, Qwen) and benchmarks (MMLU, BBH, TyDiQA) show that QLESS achieves comparable data selection performance to LESS while reducing memory usage by up to 16x. Even 1-bit gradient quantization preserves data valuation quality. These findings underscore QLESS as a practical, scalable approach to identifying informative examples within strict memory constraints.
Authors: Huaye Zeng, Dongfu Jiang, Haozhe Wang, Ping Nie, Xiaotong Chen, Wenhu Chen
Abstract: Most progress in recent coder models has been driven by supervised fine-tuning (SFT), while the potential of reinforcement learning (RL) remains largely unexplored, primarily due to the lack of reliable reward data/model in the code domain. In this paper, we address this challenge by leveraging automated large-scale test-case synthesis to enhance code model training. Specifically, we design a pipeline that generates extensive (question, test-cases) pairs from existing code data. Using these test cases, we construct preference pairs based on pass rates over sampled programs to train reward models with Bradley-Terry loss. It shows an average of 10-point improvement for Llama-3.1-8B-Ins and 5-point improvement for Qwen2.5-Coder-7B-Ins through best-of-32 sampling, making the 7B model on par with 236B DeepSeek-V2.5. Furthermore, we conduct reinforcement learning with both reward models and test-case pass rewards, leading to consistent improvements across HumanEval, MBPP, BigCodeBench, and LiveCodeBench (V4). Notably, we follow the R1-style training to start from Qwen2.5-Coder-base directly and show that our RL training can improve model on HumanEval-plus by over 25\% and MBPP-plus by 6\% for merely 80 optimization steps. We believe our results highlight the huge potential of reinforcement learning in coder models.
Authors: Martijn Bartelds, Ananjan Nandi, Moussa Koulako Bala Doumbouya, Dan Jurafsky, Tatsunori Hashimoto, Karen Livescu
Abstract: Modern deep learning models often achieve high overall performance, but consistently fail on specific subgroups. Group distributionally robust optimization (group DRO) addresses this problem by minimizing the worst-group loss, but it fails when group losses misrepresent performance differences between groups. This is common in domains like speech, where the widely used connectionist temporal classification (CTC) loss scales with input length and varies with linguistic and acoustic properties, leading to spurious differences between group losses. We present CTC-DRO, which addresses the shortcomings of the group DRO objective by smoothing the group weight update to prevent overemphasis on consistently high-loss groups, while using input length-matched batching to mitigate CTC's scaling issues. We evaluate CTC-DRO on the task of multilingual automatic speech recognition (ASR) across five language sets from the ML-SUPERB 2.0 benchmark. CTC-DRO consistently outperforms group DRO and CTC-based baseline models, reducing the worst-language error by up to 65.9% and the average error by up to 47.7%. CTC-DRO can be applied to ASR with minimal computational costs, and offers the potential for reducing group disparities in other domains with similar challenges.
Authors: Pierre Ablin, Angelos Katharopoulos, Skyler Seto, David Grangier
Abstract: Machine learning models are routinely trained on a mixture of different data domains. Different domain weights yield very different downstream performances. We propose the Soup-of-Experts, a novel architecture that can instantiate a model at test time for any domain weights with minimal computational cost and without re-training the model. Our architecture consists of a bank of expert parameters, which are linearly combined to instantiate one model. We learn the linear combination coefficients as a function of the input domain weights. To train this architecture, we sample random domain weights, instantiate the corresponding model, and backprop through one batch of data sampled with these domain weights. We demonstrate how our approach obtains small specialized models on several language modeling tasks quickly. Soup-of-Experts are particularly appealing when one needs to ship many different specialist models quickly under a model size constraint.
Authors: Kemal Kurniawan, Meladel Mistica, Timothy Baldwin, Jey Han Lau
Abstract: Human label variation (HLV) challenges the standard assumption that an example has a single ground truth, instead embracing the natural variation in human labelling to train and evaluate models. While various training methods and metrics for HLV have been proposed, there has been no systematic meta-evaluation of HLV evaluation metrics, contributing to the lack of clarity in the best HLV training method. We propose new evaluation metrics and training methods and empirically meta-evaluate HLV evaluation metrics. We find that training on either disaggregated annotations or soft labels often performs best across metrics, and that our proposed soft metric correlates best with human preference.
Authors: Angelina Wang, Michelle Phan, Daniel E. Ho, Sanmi Koyejo
Abstract: Algorithmic fairness has conventionally adopted a perspective of racial color-blindness (i.e., difference unaware treatment). We contend that in a range of important settings, group difference awareness matters. For example, differentiating between groups may be necessary in legal contexts (e.g., the U.S. compulsory draft applies to men but not women) and harm assessments (e.g., calling a girl a terrorist may be less harmful than calling a Muslim person one). In our work we first introduce an important distinction between descriptive (fact-based), normative (value-based), and correlation (association-based) benchmarks. This distinction is significant because each category requires distinct interpretation and mitigation tailored to its specific characteristics. Then, we present a benchmark suite composed of eight different scenarios for a total of 16k questions that enables us to assess difference awareness. Finally, we show results across ten models that demonstrate difference awareness is a distinct dimension of fairness where existing bias mitigation strategies may backfire.
Authors: Oscar Skean, Md Rifat Arefin, Dan Zhao, Niket Patel, Jalal Naghiyev, Yann LeCun, Ravid Shwartz-Ziv
Abstract: From extracting features to generating text, the outputs of large language models (LLMs) typically rely on their final layers, following the conventional wisdom that earlier layers capture only low-level cues. However, our analysis shows that intermediate layers can encode even richer representations, often improving performance on a wide range of downstream tasks. To explain and quantify these hidden-layer properties, we propose a unified framework of representation quality metrics based on information theory, geometry, and invariance to input perturbations. Our framework highlights how each model layer balances information compression and signal preservation, revealing why mid-depth embeddings can exceed the last layer's performance. Through extensive experiments on 32 text-embedding tasks and comparisons across model architectures (transformers, state-space models) and domains (language, vision), we demonstrate that intermediate layers consistently provide stronger features. These findings challenge the standard focus on final-layer embeddings and open new directions for model analysis and optimization, including strategic use of mid-layer representations for more robust and accurate AI systems.
Authors: Georgios Margaritis, Periklis Petridis, Dimitris J. Bertsimas
Abstract: Recent advancements in machine learning (ML), natural language processing (NLP), and foundational models have shown promise for real-life applications in critical, albeit compute-constrainted fields like healthcare. In such areas, combining foundational models with supervised ML offers potential for automating tasks like diagnosis and treatment planning, but the limited availability of onsite computational resources pose significant challenges before applying these technologies effectively: Current approaches either yield subpar results when using pretrained models without task-specific adaptation, or require substantial computational resources for fine-tuning, which is often a barrier to entry in such environments. This renders them inaccessible in applications where performance and quality standards are high, but computational resources are scarce. To bridge the gap between best-in-class performance and accessibility, we propose a novel method for adapting foundational, multimodal embeddings to downstream tasks, without the need of expensive fine-tuning processes. Our method leverages frozen embeddings from Large Language Models (LLMs) and Vision Models, and uses contrastive learning to train a small, task-specific nonlinear projection that can be used in the downstream task, without having to fine-tune the original foundational models. We show that this efficient procedure leads to significant performance improvements across various downstream tasks, and perhaps more importantly with minimal computational overhead, offering a practical solution for the use of advanced, foundational ML models in resource-constrained settings.
Authors: Raghav Arora, Shivam Singh, Karthik Swaminathan, Ahana Datta, Snehasis Banerjee, Brojeshwar Bhowmick, Krishna Murthy Jatavallabhula, Mohan Sridharan, Madhava Krishna
Abstract: Assistive agents performing household tasks such as making the bed or cooking breakfast often compute and execute actions that accomplish one task at a time. However, efficiency can be improved by anticipating upcoming tasks and computing an action sequence that jointly achieves these tasks. State-of-the-art methods for task anticipation use data-driven deep networks and Large Language Models (LLMs), but they do so at the level of high-level tasks and/or require many training examples. Our framework leverages the generic knowledge of LLMs through a small number of prompts to perform high-level task anticipation, using the anticipated tasks as goals in a classical planning system to compute a sequence of finer-granularity actions that jointly achieve these goals. We ground and evaluate our framework's abilities in realistic scenarios in the VirtualHome environment and demonstrate a 31% reduction in execution time compared with a system that does not consider upcoming tasks.
Authors: Shivam Singh, Karthik Swaminathan, Nabanita Dash, Ramandeep Singh, Snehasis Banerjee, Mohan Sridharan, Madhava Krishna
Abstract: Embodied agents assisting humans are often asked to complete a new task in a new scenario. An agent preparing a particular dish in the kitchen based on a known recipe may be asked to prepare a new dish or to perform cleaning tasks in the storeroom. There may not be sufficient resources, e.g., time or labeled examples, to train the agent for these new situations. Large Language Models (LLMs) trained on considerable knowledge across many domains are able to predict a sequence of abstract actions for such new tasks and scenarios, although it may not be possible for the agent to execute this action sequence due to task-, agent-, or domain-specific constraints. Our framework addresses these challenges by leveraging the generic predictions provided by LLM and the prior domain-specific knowledge encoded in a Knowledge Graph (KG), enabling an agent to quickly adapt to new tasks and scenarios. The robot also solicits and uses human input as needed to refine its existing knowledge. Based on experimental evaluation over cooking and cleaning tasks in simulation domains, we demonstrate that the interplay between LLM, KG, and human input leads to substantial performance gains compared with just using the LLM output.
Authors: Ruisi Zhang, Neusha Javidnia, Nojan Sheybani, Farinaz Koushanfar
Abstract: This paper introduces RoSe, the first-of-its-kind ML/Crypto codesign watermarking framework that regulates LLM-generated code to avoid intellectual property rights violations and inappropriate misuse in software development. High-quality watermarks adhering to the detectability-fidelity-robustness tri-objective are limited due to codes' low-entropy nature. Watermark verification, however, often needs to reveal the signature and requires re-encoding new ones for code reuse, which potentially compromising the system's usability. To overcome these challenges, RoSe obtains high-quality watermarks by training the watermark insertion and extraction modules end-to-end to ensure (i) unaltered watermarked code functionality and (ii) enhanced detectability and robustness leveraging pre-trained CodeT5 as the insertion backbone to enlarge the code syntactic and variable rename transformation search space. In the deployment, RoSe uses zero-knowledge proofs for secure verification without revealing the underlying signatures. Extensive evaluations demonstrated RoSe achieves high detection accuracy while preserving the code functionality. RoSe is also robust against attacks and provides efficient secure watermark verification.
Authors: Yuan Gao, Mattia Piccinini, Johannes Betz
Abstract: Large Language Models (LLMs) can capture nuanced contextual relationships, reasoning, and complex problem-solving. By leveraging their ability to process and interpret large-scale information, LLMs have shown potential to address domain-specific challenges, including those in autonomous driving systems. This paper proposes a novel framework that leverages LLMs for risk-aware analysis of generated driving scenarios. We hypothesize that LLMs can effectively evaluate whether driving scenarios generated by autonomous driving testing simulators are safety-critical. To validate this hypothesis, we conducted an empirical evaluation to assess the effectiveness of LLMs in performing this task. This framework will also provide feedback to generate the new safety-critical scenario by using adversarial method to modify existing non-critical scenarios and test their effectiveness in validating motion planning algorithms. Code and scenarios are available at: https://github.com/yuangao-tum/Riskaware-Scenario-analyse
URLs: https://github.com/yuangao-tum/Riskaware-Scenario-analyse
Authors: Thien Q. Tran, Akifumi Wachi, Rei Sato, Takumi Tanabe, Youhei Akimoto
Abstract: Safety alignment is an essential research topic for real-world AI applications. Despite the multifaceted nature of safety and trustworthiness in AI, current safety alignment methods often focus on a comprehensive notion of safety. By carefully assessing models from the existing safety-alignment methods, we found that, while they generally improved overall safety performance, they failed to ensure safety in specific categories. Our study first identified the difficulty of eliminating such vulnerabilities without sacrificing the model's helpfulness. We observed that, while smaller KL penalty parameters, increased training iterations, and dataset cleansing can enhance safety, they do not necessarily improve the trade-off between safety and helpfulness. We discovered that safety alignment could even induce undesired effects and result in a model that prefers generating negative tokens leading to rejective responses, regardless of the input context. To address this, we introduced a learning-free method, Token-level Safety-Debiased Inference (TSDI), to estimate and correct this bias during the generation process using randomly constructed prompts. Our experiments demonstrated that our method could enhance the model's helpfulness while maintaining safety, thus improving the trade-off Pareto-front.
Authors: Xiangzhi Eric Wang, Zackary P. T. Sin, Ye Jia, Daniel Archer, Wynonna H. Y. Fong, Qing Li, Chen Li
Abstract: In our daily lives, we can naturally convey instructions for the spatial manipulation of objects using words and gestures. Transposing this form of interaction into virtual reality (VR) object manipulation can be beneficial. We propose VR Mover, an LLM-empowered solution that can understand and interpret the user's vocal instruction to support object manipulation. By simply pointing and speaking, the LLM can manipulate objects without structured input. Our user study demonstrates that VR Mover enhances user usability, overall experience and performance on multi-object manipulation, while also reducing workload and arm fatigue. Users prefer the proposed natural interface for broad movements and may complementarily switch to gizmos or virtual hands for finer adjustments. These findings are believed to contribute to design implications for future LLM-based object manipulation interfaces, highlighting the potential for more intuitive and efficient user interactions in VR environments.
Authors: Wangtao Sun, Haotian Xu, Huanxuan Liao, Xuanqing Yu, Zhongtao Jiang, Shizhu He, Jun Zhao, Kang Liu
Abstract: How to interact with LLMs through \emph{instructions} has been widely studied by researchers. However, previous studies have treated the emergence of instructions and the training of LLMs on task data as separate processes, overlooking the inherent unity between the two. This paper proposes a neural network framework, VaiBot, that integrates VAE and VIB, designed to uniformly model, learn, and infer both deduction and induction tasks under LLMs. Through experiments, we demonstrate that VaiBot performs on par with existing baseline methods in terms of deductive capabilities while significantly surpassing them in inductive capabilities. We also find that VaiBot can scale up using general instruction-following data and exhibits excellent one-shot induction abilities. We finally synergistically integrate the deductive and inductive processes of VaiBot. Through T-SNE dimensionality reduction, we observe that its inductive-deductive process significantly improves the distribution of training parameters, enabling it to outperform baseline methods in inductive reasoning tasks. The code and data for this paper can be found at https://anonymous.4open.science/r/VaiBot-021F.
Authors: Yuan Tian, Chuhan Zhang, Xiaotong Wang, Sitong Pan, Weiwei Cui, Haidong Zhang, Dazhen Deng, Yingcai Wu
Abstract: Creating data reports is time-consuming, as it requires iterative exploration and understanding of data, followed by summarizing the insights. While large language models (LLMs) are powerful tools for data processing and text generation, they often struggle to produce complete data reports that fully meet user expectations. One significant challenge is effectively communicating the entire analysis logic to LLMs. Moreover, determining a comprehensive analysis logic can be mentally taxing for users. To address these challenges, we propose ReSpark, an LLM-based method that leverages existing data reports as references for creating new ones. Given a data table, ReSpark searches for similar-topic reports, parses them into interdependent segments corresponding to analytical objectives, and executes them with new data. It identifies inconsistencies and customizes the objectives, data transformations, and textual descriptions. ReSpark allows users to review real-time outputs, insert new objectives, and modify report content. Its effectiveness was evaluated through comparative and user studies.
Authors: Sidak Pal Singh, Hossein Mobahi, Atish Agarwala, Yann Dauphin
Abstract: Curvature regularization techniques like Sharpness Aware Minimization (SAM) have shown great promise in improving generalization on vision tasks. However, we find that SAM performs poorly in domains like natural language processing (NLP), often degrading performance -- even with twice the compute budget. We investigate the discrepancy across domains and find that in the NLP setting, SAM is dominated by regularization of the logit statistics -- instead of improving the geometry of the function itself. We use this observation to develop an alternative algorithm we call Functional-SAM, which regularizes curvature only through modification of the statistics of the overall function implemented by the neural network, and avoids spurious minimization through logit manipulation. Furthermore, we argue that preconditioning the SAM perturbation also prevents spurious minimization, and when combined with Functional-SAM, it gives further improvements. Our proposed algorithms show improved performance over AdamW and SAM baselines when trained for an equal number of steps, in both fixed-length and Chinchilla-style training settings, at various model scales (including billion-parameter scale). On the whole, our work highlights the importance of more precise characterizations of sharpness in broadening the applicability of curvature regularization to large language models (LLMs).
Authors: Abdelrahman Abdallah, Jamshid Mozafari, Bhawna Piryani, Mohammed Ali, Adam Jatowt
Abstract: Retrieval, re-ranking, and retrieval-augmented generation (RAG) are critical components of modern natural language processing (NLP) applications in information retrieval, question answering, and knowledge-based text generation. However, existing solutions are often fragmented, lacking a unified framework that easily integrates these essential processes. The absence of a standardized implementation, coupled with the complexity of retrieval and re-ranking workflows, makes it challenging for researchers to compare and evaluate different approaches in a consistent environment. While existing toolkits such as Rerankers and RankLLM provide general-purpose reranking pipelines, they often lack the flexibility required for fine-grained experimentation and benchmarking. In response to these challenges, we introduce \textbf{Rankify}, a powerful and modular open-source toolkit designed to unify retrieval, re-ranking, and RAG within a cohesive framework. Rankify supports a wide range of retrieval techniques, including dense and sparse retrievers, while incorporating state-of-the-art re-ranking models to enhance retrieval quality. Additionally, Rankify includes a collection of pre-retrieved datasets to facilitate benchmarking, available at Huggingface (https://huggingface.co/datasets/abdoelsayed/reranking-datasets). To encourage adoption and ease of integration, we provide comprehensive documentation (http://rankify.readthedocs.io/), an open-source implementation on GitHub(https://github.com/DataScienceUIBK/rankify), and a PyPI package for effortless installation(https://pypi.org/project/rankify/). By providing a unified and lightweight framework, Rankify allows researchers and practitioners to advance retrieval and re-ranking methodologies while ensuring consistency, scalability, and ease of use.
URLs: https://huggingface.co/datasets/abdoelsayed/reranking-datasets)., http://rankify.readthedocs.io/),, https://github.com/DataScienceUIBK/rankify),, https://pypi.org/project/rankify/).
Authors: Dylan Sam, Ayan Chakrabarti, Afshin Rostamizadeh, Srikumar Ramalingam, Gui Citovsky, Sanjiv Kumar
Abstract: Similarity between training examples is used to curate pretraining datasets for language models by many methods -- for diversification and to select examples similar to high-quality data. However, similarity is typically measured with off-the-shelf embedding models that are generic or trained for tasks such as retrieval. This paper introduces a framework to analyze the suitability of embedding models specifically for data curation in the language model pretraining setting. We quantify the correlation between similarity in the embedding space to similarity in pretraining loss between different training examples, and how diversifying in the embedding space affects pretraining quality. We analyze a variety of embedding models in our framework, with experiments using the Pile dataset for pretraining a 1.7B parameter decoder-only language model. We find that the embedding models we consider are all useful for pretraining data curation. Moreover, a simple approach of averaging per-token embeddings proves to be surprisingly competitive with more sophisticated embedding models -- likely because the latter are not designed specifically for pretraining data curation. Indeed, we believe our analysis and evaluation framework can serve as a foundation for the design of embedding models that specifically reason about similarity in pretraining datasets.
Authors: Han Zhou, Xingchen Wan, Ruoxi Sun, Hamid Palangi, Shariq Iqbal, Ivan Vuli\'c, Anna Korhonen, Sercan \"O. Ar{\i}k
Abstract: Large language models, employed as multiple agents that interact and collaborate with each other, have excelled at solving complex tasks. The agents are programmed with prompts that declare their functionality, along with the topologies that orchestrate interactions across agents. Designing prompts and topologies for multi-agent systems (MAS) is inherently complex. To automate the entire design process, we first conduct an in-depth analysis of the design space aiming to understand the factors behind building effective MAS. We reveal that prompts together with topologies play critical roles in enabling more effective MAS design. Based on the insights, we propose Multi-Agent System Search (MASS), a MAS optimization framework that efficiently exploits the complex MAS design space by interleaving its optimization stages, from local to global, from prompts to topologies, over three stages: 1) block-level (local) prompt optimization; 2) workflow topology optimization; 3) workflow-level (global) prompt optimization, where each stage is conditioned on the iteratively optimized prompts/topologies from former stages. We show that MASS-optimized multi-agent systems outperform a spectrum of existing alternatives by a substantial margin. Based on the MASS-found systems, we finally propose design principles behind building effective multi-agent systems.
Authors: Ga\v{s}per Begu\v{s}, Maksymilian D\k{a}bkowski, Ryan Rhodes
Abstract: The performance of large language models (LLMs) has recently improved to the point where the models can perform well on many language tasks. We show here that -- for the first time -- the models can also generate valid metalinguistic analyses of language data. We outline a research program where the behavioral interpretability of LLMs on these tasks is tested via prompting. LLMs are trained primarily on text -- as such, evaluating their metalinguistic abilities improves our understanding of their general capabilities and sheds new light on theoretical models in linguistics. We show that OpenAI's o1 vastly outperforms other models on tasks involving drawing syntactic trees and phonological generalization. We speculate that OpenAI o1's unique advantage over other models may result from the model's chain-of-thought mechanism, which mimics the structure of human reasoning used in complex cognitive tasks, such as linguistic analysis.
Authors: Zach Nussbaum, John X. Morris, Brandon Duderstadt, Andriy Mulyar
Abstract: This technical report describes the training of nomic-embed-text-v1, the first fully reproducible, open-source, open-weights, open-data, 8192 context length English text embedding model that outperforms both OpenAI Ada-002 and OpenAI text-embedding-3-small on the short-context MTEB benchmark and the long context LoCo benchmark. We release the training code and model weights under an Apache 2.0 license. In contrast with other open-source models, we release the full curated training data and code that allows for full replication of nomic-embed-text-v1. You can find code and data to replicate the model at https://github.com/nomic-ai/contrastors.
Authors: Rena Gao, Carsten Roever, Jey Han Lau
Abstract: We present an evaluation framework for interactive dialogue assessment in the context of English as a Second Language (ESL) speakers. Our framework collects dialogue-level interactivity labels (e.g., topic management; 4 labels in total) and micro-level span features (e.g., backchannels; 17 features in total). Given our annotated data, we study how the micro-level features influence the (higher level) interactivity quality of ESL dialogues by constructing various machine learning-based models. Our results demonstrate that certain micro-level features strongly correlate with interactivity quality, like reference word (e.g., she, her, he), revealing new insights about the interaction between higher-level dialogue quality and lower-level linguistic signals. Our framework also provides a means to assess ESL communication, which is useful for language assessment.
Authors: Nitay Calderon, Roi Reichart
Abstract: Recent advancements in NLP systems, particularly with the introduction of LLMs, have led to widespread adoption of these systems by a broad spectrum of users across various domains, impacting decision-making, the job market, society, and scientific research. This surge in usage has led to an explosion in NLP model interpretability and analysis research, accompanied by numerous technical surveys. Yet, these surveys often overlook the needs and perspectives of explanation stakeholders. In this paper, we address three fundamental questions: Why do we need interpretability, what are we interpreting, and how? By exploring these questions, we examine existing interpretability paradigms, their properties, and their relevance to different stakeholders. We further explore the practical implications of these paradigms by analyzing trends from the past decade across multiple research fields. To this end, we retrieved thousands of papers and employed an LLM to characterize them. Our analysis reveals significant disparities between NLP developers and non-developer users, as well as between research fields, underscoring the diverse needs of stakeholders. For example, explanations of internal model components are rarely used outside the NLP field. We hope this paper informs the future design, development, and application of methods that align with the objectives and requirements of various stakeholders.
Authors: Guangya Wan, Yuqi Wu, Jie Chen, Sheng Li
Abstract: Self-Consistency mitigates hallucinations in Large Language Models (LLMs) by sampling multiple reasoning paths,but it lacks a systematic approach to determine the optimal number of samples or select the most faithful rationale. To address this limitation, we introduce Reasoning-Aware Self-Consistency (RASC), a novel framework that enhances sampling efficiency and reasoning faithfulness by dynamically evaluating both outputs and rationales. RASC assesses the quality of reasoning and the consistency of answers for each generated sample, using these assessments to guide early stopping decisions and rationale selection. The framework employs criteria-based stopping and weighted majority voting, enabling more informed choices on when to halt sampling and which rationale to select. Our comprehensive experiments across diverse question-answering datasets demonstrate that RASC outperforms existing methods, reducing sample usage by approximately 70% while maintaining accuracy. Moreover, RASC facilitates the selection of high-fidelity rationales, thereby improving the faithfulness of LLM outputs. Our approach effectively addresses the efficiency-accuracy trade-off in LLM reasoning tasks, offering a new perspective for more nuanced, faithful, and effective utilization of LLMs in resource-constrained environments.
Authors: Blessed Guda, Gabrial Zencha Ashungafac, Lawrence Francis, Carlee Joe-Wong
Abstract: Large Language models (LLMs) have brought about substantial advancements in the field of Question Answering (QA) systems. These models do remarkably well in addressing intricate inquiries in a variety of disciplines. However, because of domain-specific vocabulary, complex technological concepts, and the requirement for exact responses applying LLMs to specialized sectors like telecommunications presents additional obstacles. GPT-3.5 has been used in recent work, to obtain noteworthy accuracy for telecom-related questions in a Retrieval Augmented Generation (RAG) framework. Notwithstanding these developments, the practical use of models such as GPT-3.5 is restricted by their proprietary nature and high computing demands. This paper introduces QMOS, an innovative approach which uses a Question-Masked loss and Option Shuffling trick to enhance the performance of LLMs in answering Multiple-Choice Questions in the telecommunications domain. Our focus was on using opensource, smaller language models (Phi-2 and Falcon-7B) within an enhanced RAG framework. Our multi-faceted approach involves several enhancements to the whole LLM-RAG pipeline of finetuning, retrieval, prompt engineering and inference. Our approaches significantly outperform existing results, achieving accuracy improvements from baselines of 24.70% to 49.30% with Falcon-7B and from 42.07% to 84.65% with Phi-2.
Authors: Ali Mohammadjafari, Anthony S. Maida, Raju Gottumukkala
Abstract: LLMs when used with Retrieval Augmented Generation (RAG), are greatly improving the SOTA of translating natural language queries to structured and correct SQL. Unlike previous reviews, this survey provides a comprehensive study of the evolution of LLM-based text-to-SQL systems, from early rule-based models to advanced LLM approaches that use (RAG) systems. We discuss benchmarks, evaluation methods, and evaluation metrics. Also, we uniquely study the use of Graph RAGs for better contextual accuracy and schema linking in these systems. Finally, we highlight key challenges such as computational efficiency, model robustness, and data privacy toward improvements of LLM-based text-to-SQL systems.
Authors: Fillipe dos Santos Silva, Gabriel Kenzo Kakimoto, Julio Cesar dos Reis, Marcelo S. Reis
Abstract: Cluster analysis plays a crucial role in various domains and applications, such as customer segmentation in marketing. These contexts often involve multimodal data, including both tabular and textual datasets, making it challenging to represent hidden patterns for obtaining meaningful clusters. This study introduces ERASMO, a framework designed to fine-tune a pretrained language model on textually encoded tabular data and generate embeddings from the fine-tuned model. ERASMO employs a textual converter to transform tabular data into a textual format, enabling the language model to process and understand the data more effectively. Additionally, ERASMO produces contextually rich and structurally representative embeddings through techniques such as random feature sequence shuffling and number verbalization. Extensive experimental evaluations were conducted using multiple datasets and baseline approaches. Our results demonstrate that ERASMO fully leverages the specific context of each tabular dataset, leading to more precise and nuanced embeddings for accurate clustering. This approach enhances clustering performance by capturing complex relationship patterns within diverse tabular data.
Authors: David Reber, Sean Richardson, Todd Nief, Cristina Garbacea, Victor Veitch
Abstract: Reward models are widely used as proxies for human preferences when aligning or evaluating LLMs. However, reward models are black boxes, and it is often unclear what, exactly, they are actually rewarding. In this paper we develop Rewrite-based Attribute Treatment Estimator (RATE) as an effective method for measuring the sensitivity of a reward model to high-level attributes of responses, such as sentiment, helpfulness, or complexity. Importantly, RATE measures the causal effect of an attribute on the reward. RATE uses LLMs to rewrite responses to produce imperfect counterfactual examples that can be used to measure causal effects. A key challenge is that these rewrites are imperfect in a manner that can induce substantial bias in the estimated sensitivity of the reward model to the attribute. The core idea of RATE is to adjust for this imperfect-rewrite effect by rewriting twice. We establish the validity of the RATE procedure and show empirically that it is an effective estimator.
Authors: Zihan Chen, Bike Xie, Jundong Li, Cong Shen
Abstract: Large Language Models (LLMs) have demonstrated remarkable success across a wide range of language tasks, but their deployment on edge devices remains challenging due to the substantial memory requirements imposed by their large parameter sizes. Weight-only quantization presents a promising solution to reduce the memory footprint of LLMs. However, existing approaches primarily focus on integer-bit quantization, limiting their adaptability to fractional-bit quantization tasks and preventing the full utilization of available storage space on devices. In this paper, we introduce Channel-Wise Mixed-Precision Quantization (CMPQ), a novel mixed-precision quantization method that allocates quantization precision in a channel-wise pattern based on activation distributions. By assigning different precision levels to different weight channels, CMPQ can adapt to any bit-width constraint. CMPQ employs a non-uniform quantization strategy and incorporates two outlier extraction techniques that collaboratively preserve the critical information, thereby minimizing the quantization loss. Experiments on different sizes of LLMs demonstrate that CMPQ not only enhances performance in integer-bit quantization tasks but also achieves significant performance gains with a modest increase in memory usage. CMPQ thus represents an adaptive and effective approach to LLM quantization, offering substantial benefits across diverse device capabilities.
Authors: Xuan Zhang, Fengzhuo Zhang, Cunxiao Du, Chao Du, Tianyu Pang, Wei Gao, Min Lin
Abstract: Scaling language models to handle longer contexts introduces substantial memory challenges due to the growing cost of key-value (KV) caches. Motivated by the efficiency gains of hybrid models and the broad availability of pretrained large transformer backbones, we explore transitioning transformer models into hybrid architectures for a more efficient generation. In this work, we propose LightTransfer, a lightweight method that transforms models such as LLaMA into hybrid variants. Our approach identifies lazy layers -- those focusing on recent or initial tokens -- and replaces their full attention with streaming attention. This transformation can be performed without any training for long-context understanding tasks or with minimal fine-tuning for o1-like long reasoning generation tasks that require stronger reasoning capabilities. Experiments across diverse benchmarks and models (e.g., LLaMA, Mistral, QwQ-STILL) demonstrate that, even when half of the layers are identified as lazy, LightTransfer achieves up to 2.17$\times$ throughput improvement with minimal performance loss ($<1.5\%$ on LongBench) and achieves 53.3\% on math benchmark AIME24 of advanced o1-like long reasoning model QwQ-STILL.
Authors: SeongYeub Chu, JongWoo Kim, Bryan Wong, MunYong Yi
Abstract: Existing automated essay scoring (AES) has solely relied on essay text without using explanatory rationales for the scores, thereby forgoing an opportunity to capture the specific aspects evaluated by rubric indicators in a fine-grained manner. This paper introduces Rationale-based Multiple Trait Scoring (RMTS), a novel approach for multi-trait essay scoring that integrates prompt-engineering-based large language models (LLMs) with a fine-tuning-based essay scoring model using a smaller large language model (S-LLM). RMTS uses an LLM-based trait-wise rationale generation system where a separate LLM agent generates trait-specific rationales based on rubric guidelines, which the scoring model uses to accurately predict multi-trait scores. Extensive experiments on benchmark datasets, including ASAP, ASAP++, and Feedback Prize, show that RMTS significantly outperforms state-of-the-art models and vanilla S-LLMs in trait-specific scoring. By assisting quantitative assessment with fine-grained qualitative rationales, RMTS enhances the trait-wise reliability, providing partial explanations about essays. The code is available at https://github.com/BBeeChu/RMTS.git.
Authors: Prakhar Verma, Sukruta Prakash Midigeshi, Gaurav Sinha, Arno Solin, Nagarajan Natarajan, Amit Sharma
Abstract: We introduce Plan*RAG, a novel framework that enables structured multi-hop reasoning in retrieval-augmented generation (RAG) through test-time reasoning plan generation. While existing approaches such as ReAct maintain reasoning chains within the language model's context window, we observe that this often leads to plan fragmentation and execution failures. Our key insight is that by isolating the reasoning plan as a directed acyclic graph (DAG) outside the LM's working memory, we can enable (1) systematic exploration of reasoning paths, (2) atomic subqueries enabling precise retrievals and grounding, and (3) efficiency through parallel execution and bounded context window utilization. Moreover, Plan*RAG's modular design allows it to be integrated with existing RAG methods, thus providing a practical solution to improve current RAG systems. On standard multi-hop reasoning benchmarks, Plan*RAG consistently achieves improvements over recently proposed methods such as RQ-RAG and Self-RAG, while maintaining comparable computational costs.
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 41 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 74 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 are publicly available to foster open research.
Authors: Ekaterina Artemova, Jason Lucas, Saranya Venkatraman, Jooyoung Lee, Sergei Tilga, Adaku Uchendu, Vladislav Mikhailov
Abstract: The rapid proliferation of large language models (LLMs) has increased the volume of machine-generated texts (MGTs) and blurred text authorship in various domains. However, most existing MGT benchmarks include single-author texts (human-written and machine-generated). This conventional design fails to capture more practical multi-author scenarios, where the user refines the LLM response for natural flow, coherence, and factual correctness. Our paper introduces the Benchmark of Expert-edited Machine-generated Outputs (Beemo), which includes 6.5k texts written by humans, generated by ten instruction-finetuned LLMs, and edited by experts for various use cases, ranging from creative writing to summarization. Beemo additionally comprises 13.1k machine-generated and LLM-edited texts, allowing for diverse MGT detection evaluation across various edit types. We document Beemo's creation protocol and present the results of benchmarking 33 configurations of MGT detectors in different experimental setups. We find that expert-based editing evades MGT detection, while LLM-edited texts are unlikely to be recognized as human-written. Beemo and all materials are publicly available.
Authors: Muhammad Khalifa, Yi-Chern Tan, Arash Ahmadian, Tom Hosking, Honglak Lee, Lu Wang, Ahmet \"Ust\"un, Tom Sherborne, Matthias Gall\'e
Abstract: Model merging has shown great promise at combining expert models, but the benefit of merging is unclear when merging "generalist" models trained on many tasks. We explore merging in the context of large (~100B) models, by recycling checkpoints that exhibit tradeoffs among different tasks. Such checkpoints are often created in the process of developing a frontier model, and the suboptimal ones are usually discarded. Given a pool of model checkpoints obtained from different training runs (e.g., different stages, objectives, hyperparameters, and data mixtures), which naturally show tradeoffs across different language capabilities (e.g., instruction following vs. code generation), we investigate whether merging can recycle such suboptimal models into a Pareto-optimal one. Our optimization algorithm tunes the weight of each checkpoint in a linear combination, resulting in such an optimal model that outperforms both individual models and merge-based baselines. Further analysis shows that good merges tend to include almost all checkpoints with non-zero weights, indicating that even seemingly bad initial checkpoints can contribute to good final merges.
Authors: Shengmin Piao, Sanghyun Park
Abstract: Large Language Models exhibit impressive reasoning capabilities across diverse tasks, motivating efforts to distill these capabilities into smaller models through generated reasoning data. However, direct training on such synthesized reasoning data may lead to superficial imitation of reasoning process, rather than fostering a genuine integration of reasoning capabilities with underlying knowledge. To address this, we propose TinyThinker, a framework introducing two novel approaches. First, we introduce a three-stage process that incrementally guides the student model through the reasoning process, progressively refining knowledge from coarse to fine granularity. Second, we develop a two-phase training framework comprising an initial reasoning acquisition phase followed by a self-reflection phase utilizing self-generated data. Experiments on commonsense reasoning benchmarks demonstrate that TinyThinker achieves superior performance compared to baselines. Ablation studies further validate the effectiveness of each component in our framework. We expect that TinyThinker can be extended to other knowledge-intensive reasoning tasks, offering an alternative strategy for developing effective reasoning capabilities in smaller language models. Codes are available at https://github.com/shengminp/TinyThinker
Authors: Zhenni Bi, Kai Han, Chuanjian Liu, Yehui Tang, Yunhe Wang
Abstract: Large Language Models (LLMs) have demonstrated remarkable abilities across various language tasks, but solving complex reasoning problems remains a significant challenge. While existing methods, such as Chain-of-Thought (CoT) and Tree-of-Thought (ToT), enhance reasoning by decomposing problems or structuring prompts, they typically perform a single pass of reasoning and may fail to revisit flawed paths, compromising accuracy. To address this limitation, we propose a novel reasoning framework called Forest-of-Thought (FoT), which integrates multiple reasoning trees to leverage collective decision-making for solving complex logical problems. FoT employs sparse activation strategies to select the most relevant reasoning paths, improving both efficiency and accuracy. Additionally, we introduce a dynamic self-correction strategy that enables real-time error correction, along with consensus-guided decision-making strategies to optimize both correctness and computational resources. Experimental results demonstrate that the FoT framework, combined with these strategies, significantly enhances the reasoning capabilities of LLMs, enabling them to solve complex tasks with greater precision and efficiency.
Authors: Trishita Tiwari, G. Edward Suh
Abstract: This work performs an analysis of sequence-level probabilities for quantifying the of risk training data extraction from Large Language Models (LLMs). Per-sequence extraction probabilities provide finer-grained information than has been studied in prior work. We re-analyze the effects of decoding schemes, model sizes, prefix lengths, partial sequence leakages, and token positions to uncover new insights that were not possible in previous works due to their choice of metrics. We perform this study on two pre-trained models, Llama and OPT, trained on the Common Crawl and The Pile respectively. We discover that 1) Extraction Rate, the predominant metric used in prior quantification work, underestimates the threat of leakage of training data in randomized LLMs by as much as 2.14X. 2) Although on average, larger models and longer prefixes can extract more data, this is not true for a substantial portion of individual sequences. 30.4-41.5% of our sequences are easier to extract with either shorter prefixes or smaller models. 3) Contrary to previous beliefs, partial leakage in commonly used decoding schemes like top-k and top-p is not easier than leaking verbatim training data. 4) Extracting later tokens in a sequence is as much as 10.12X easier than extracting earlier tokens. The insights gained from our analysis shed light on the nature of memorization of training data on a per-sequence basis.
Authors: Hyeonchu Park, Byungjun Kim, Bugeun Kim
Abstract: As large language models (LLMs) generate more human-like texts, concerns about the side effects of AI-generated texts (AIGT) have grown. So, researchers have developed methods for detecting AIGT. However, two challenges remain. First, the performance of detecting black-box LLMs is low because existing models focus on probabilistic features. Second, most AIGT detectors have been tested on a single-candidate setting, which assumes that we know the origin of an AIGT and which may deviate from the real-world scenario. To resolve these challenges, we propose DART, which consists of four steps: rephrasing, semantic parsing, scoring, and multiclass classification. We conducted three experiments to test the performance of DART. The experimental result shows that DART can discriminate multiple black-box LLMs without probabilistic features and the origin of AIGT.
Authors: Guoxuan Chen, Han Shi, Jiawei Li, Yihang Gao, Xiaozhe Ren, Yimeng Chen, Xin Jiang, Zhenguo Li, Weiyang Liu, Chao Huang
Abstract: Large Language Models (LLMs) have exhibited exceptional performance across a spectrum of natural language processing tasks. However, their substantial sizes pose considerable challenges, particularly in computational demands and inference speed, due to their quadratic complexity. In this work, we have identified a key pattern: certain seemingly meaningless special tokens (i.e., separators) contribute disproportionately to attention scores compared to semantically meaningful tokens. This observation suggests that information of the segments between these separator tokens can be effectively condensed into the separator tokens themselves without significant information loss. Guided by this insight, we introduce SepLLM, a plug-and-play framework that accelerates inference by compressing these segments and eliminating redundant tokens. Additionally, we implement efficient kernels for training acceleration. Experimental results across training-free, training-from-scratch, and post-training settings demonstrate SepLLM's effectiveness. Notably, using the Llama-3-8B backbone, SepLLM achieves over 50% reduction in KV cache on the GSM8K-CoT benchmark while maintaining comparable performance. Furthermore, in streaming settings, SepLLM effectively processes sequences of up to 4 million tokens or more while maintaining consistent language modeling capabilities.
Authors: Adam Ishay, Joohyung Lee
Abstract: Large Language Models (LLMs) have made significant strides in various intelligent tasks but still struggle with complex action reasoning tasks that require systematic search. To address this limitation, we propose a method that bridges the natural language understanding capabilities of LLMs with the symbolic reasoning strengths of action languages. Our approach, termed "LLM+AL," leverages the LLM's strengths in semantic parsing and commonsense knowledge generation alongside the action language's proficiency in automated reasoning based on encoded knowledge. We compare LLM+AL against state-of-the-art LLMs, including ChatGPT-4, Claude 3 Opus, Gemini Ultra 1.0, and o1-preview, using benchmarks for complex reasoning about actions. Our findings indicate that, although all methods exhibit errors, LLM+AL, with relatively minimal human corrections, consistently leads to correct answers, whereas standalone LLMs fail to improve even with human feedback. LLM+AL also contributes to automated generation of action languages.
Authors: Sara Abdali, Can Goksen, Saeed Amizadeh, Kazuhito Koishida
Abstract: Investigating NLP through a philosophical lens has recently caught researcher's eyes as it connects computational methods with classical schools of philosophy. This paper introduces a philosophical approach inspired by the Hegelian Dialectic for LLMs' self-reflection, utilizing a self-dialectical approach to emulate internal critiques and then synthesize new ideas by resolving the contradicting points. Moreover, this paper investigates the effect of LLMs' temperature for generation by establishing a dynamic annealing approach, which promotes the creativity in the early stages and gradually refines it by focusing on the nuances, as well as a fixed temperature strategy for generation. Our proposed approach is examined to determine its ability to generate novel ideas from an initial proposition. Additionally, a Multi Agent Majority Voting (MAMV) strategy is leveraged to assess the validity and novelty of the generated ideas, which proves beneficial in the absence of domain experts. Our experiments show promise in generating new ideas and provide a stepping stone for future research.
Authors: Eunjung Yeo, Julie Liss, Visar Berisha, David Mortensen
Abstract: Purpose: This commentary introduces how artificial intelligence (AI) can be leveraged to advance cross-language intelligibility assessment of dysarthric speech. Method: We propose a conceptual framework consisting of a universal model that captures language-universal speech impairments and a language-specific intelligibility model that incorporates linguistic nuances. Additionally, we identify key barriers to cross-language intelligibility assessment, including data scarcity, annotation complexity, and limited linguistic insights, and present AI-driven solutions to overcome these challenges. Conclusion: Advances in AI offer transformative opportunities to enhance cross-language intelligibility assessment for dysarthric speech by balancing scalability across languages and adaptability by languages.
Authors: David Salinas, Omar Swelam, Frank Hutter
Abstract: Evaluating Large Language Models (LLMs) often requires costly human annotations. To address this, LLM-based judges have been proposed, which compare the outputs of two LLMs enabling the ranking of models without human intervention. While several approaches have been proposed, many confounding factors are present between different papers. For instance the model, the prompt and other hyperparameters are typically changed at the same time making apple-to-apple comparisons challenging. In this paper, we propose to systematically analyze and tune hyperparameter of LLM judges. To alleviate the high cost of evaluating a judge, we propose to leverage multi-objective multi-fidelity which allows to find judges that trades accuracy for cost and also reduce significantly the cost of the search. Our method identifies judges that not only outperform existing benchmarks in accuracy and cost-efficiency but also utilize open-weight models, ensuring greater accessibility and reproducibility.
Authors: Adeyola Opaluwah
Abstract: This paper examines the output of cultural items generated by Chat Generative PreTrained Transformer Pro in response to three structured prompts to translate three anthologies of African poetry. The first prompt was broad, the second focused on poetic structure, and the third prompt emphasized cultural specificity. To support this analysis, four comparative tables were created. The first table presents the results of the cultural items produced after the three prompts, the second categorizes these outputs based on Aixela framework of Proper nouns and Common expressions, the third table summarizes the cultural items generated by human translators, a custom translation engine, and a Large Language Model. The final table outlines the strategies employed by Chat Generative PreTrained Transformer Pro following the culture specific prompt. Compared to the outputs of cultural items from reference human translation and the custom translation engine in prior studies the findings indicate that the culture oriented prompts used with Chat Generative PreTrained Transformer Pro did not yield significant enhancements of cultural items during the translation of African poetry from English to French. Among the fifty four cultural items, the human translation produced thirty three cultural items in repetition, the custom translation engine generated Thirty eight cultural items in repetition while Chat Generative PreTrained Transformer Pro produced forty one cultural items in repetition. The untranslated cultural items revealed inconsistencies in Large language models approach to translating cultural items in African poetry from English to French.
Authors: Manish Sanwal
Abstract: Large Language Models (LLMs) leverage chain-of-thought (CoT) prompting to provide step-by-step rationales, improving performance on complex tasks. Despite its benefits, vanilla CoT often fails to fully verify intermediate inferences and can produce misleading explanations. In this work, we propose Layered Chain-of-Thought (Layered-CoT) Prompting, a novel framework that systematically segments the reasoning process into multiple layers, each subjected to external checks and optional user feedback. We expand on the key concepts, present three scenarios -- medical triage, financial risk assessment, and agile engineering -- and demonstrate how Layered-CoT surpasses vanilla CoT in terms of transparency, correctness, and user engagement. By integrating references from recent arXiv papers on interactive explainability, multi-agent frameworks, and agent-based collaboration, we illustrate how Layered-CoT paves the way for more reliable and grounded explanations in high-stakes domains.
Authors: Jingjing Xu, Eugen Beck, Zijian Yang, Ralf Schl\"uter
Abstract: ASR systems are deployed across diverse environments, each with specific hardware constraints. We use supernet training to jointly train multiple encoders of varying sizes, enabling dynamic model size adjustment to fit hardware constraints without redundant training. Moreover, we introduce a novel method called OrthoSoftmax, which applies multiple orthogonal softmax functions to efficiently identify optimal subnets within the supernet, avoiding resource-intensive search. This approach also enables more flexible and precise subnet selection by allowing selection based on various criteria and levels of granularity. Our results with CTC on Librispeech and TED-LIUM-v2 show that FLOPs-aware component-wise selection achieves the best overall performance. With the same number of training updates from one single job, WERs for all model sizes are comparable to or slightly better than those of individually trained models. Furthermore, we analyze patterns in the selected components and reveal interesting insights.
Authors: Ling Xing, Alex Jinpeng Wang, Rui Yan, Jinhui Tang
Abstract: Large Language Models (LLMs) have revolutionized natural language processing, excelling in handling longer sequences. However, the inefficiency and redundancy in processing extended in-context tokens remain a challenge. Many attempts to address this rely on compressing tokens with smaller text encoders, yet we question whether text encoders are truly indispensable. Our journey leads to an unexpected discovery-a much smaller vision encoder, applied directly to sequences of text tokens, can rival text encoders on text tasks. When pre-trained on large amounts of data and transferred to multiple mid-sized or small text understanding benchmarks, VIST leads to comparable results with 16% fewer FLOPs and 50% less memory usage. We further uncover significant token redundancy and devise a frequency-based masking strategy to guide the focus of the visual encoder toward the most critical tokens. Interestingly, we observe the trained visual encoder performs like a summarizer, selectively ignoring less important words such as prepositions and conjunctions. This approach delivers remarkable results, outperforming traditional text encoder-based methods by 5.7% on average over benchmarks like TriviaQA, NQ, PopQA, TREF, SST2, and SST5, setting a new standard for token efficiency in LLMs.
Authors: Taewoo Kang, Kjerstin Thorson, Tai-Quan Peng, Dan Hiaeshutter-Rice, Sanguk Lee, Stuart Soroka
Abstract: This study attempts to advancing content analysis methodology from consensus-oriented to coordination-oriented practices, thereby embracing diverse coding outputs and exploring the dynamics among differential perspectives. As an exploratory investigation of this approach, we evaluate six GPT-4o configurations to analyze sentiment in Fox News and MSNBC transcripts on Biden and Trump during the 2020 U.S. presidential campaign, examining patterns across these models. By assessing each model's alignment with ideological perspectives, we explore how partisan selective processing could be identified in LLM-Assisted Content Analysis (LACA). Findings reveal that partisan persona LLMs exhibit stronger ideological biases when processing politically congruent content. Additionally, intercoder reliability is higher among same-partisan personas compared to cross-partisan pairs. This approach enhances the nuanced understanding of LLM outputs and advances the integrity of AI-driven social science research, enabling simulations of real-world implications.
Authors: Yuhang Zhou, Giannis Karamanolakis, Victor Soto, Anna Rumshisky, Mayank Kulkarni, Furong Huang, Wei Ai, Jianhua Lu
Abstract: The recent success of specialized Large Language Models (LLMs) in domains such as mathematical reasoning and coding has led to growing interest in methods for merging these expert LLMs into a unified Mixture-of-Experts (MoE) model, with the goal of enhancing performance in each domain while retaining effectiveness on general tasks. However, the effective merging of expert models remains an open challenge, especially for models with highly divergent weight parameters or different architectures. State-of-the-art MoE merging methods only work with homogeneous model architectures and rely on simple unweighted averaging to merge expert layers, which does not address parameter interference and requires extensive fine-tuning of the merged MoE to restore performance. To address these limitations, this paper introduces new MoE merging techniques, including strategies to mitigate parameter interference, routing heuristics to reduce the need for MoE fine-tuning, and a novel method for merging experts with different architectures. Extensive experiments across multiple domains demonstrate the effectiveness of our proposed methods, reducing fine-tuning costs, improving performance over state-of-the-art methods, and expanding the applicability of MoE merging.
Authors: Benjamin A. Spiegel, Lucas Gelfond, George Konidaris
Abstract: Abstract symbolic writing systems are semiotic codes that are ubiquitous in modern society but are otherwise absent in the animal kingdom. Anthropological evidence suggests that the earliest forms of some writing systems originally consisted of iconic pictographs, which signify their referent via visual resemblance. While previous studies have examined the emergence and, separately, the evolution of pictographic writing systems through a computational lens, most employ non-naturalistic methodologies that make it difficult to draw clear analogies to human and animal cognition. We develop a multi-agent reinforcement learning testbed for emergent communication called a Signification Game, and formulate a model of inferential communication that enables agents to leverage visual theory of mind to communicate actions using pictographs. Our model, which is situated within a broader formalism for animal communication, sheds light on the cognitive and cultural processes that led to the development of early writing systems.
Authors: Yu Zhang, Rongjie Huang, Ruiqi Li, JinZheng He, Yan Xia, Feiyang Chen, Xinyu Duan, Baoxing Huai, Zhou Zhao
Abstract: Style transfer for out-of-domain (OOD) singing voice synthesis (SVS) focuses on generating high-quality singing voices with unseen styles (such as timbre, emotion, pronunciation, and articulation skills) derived from reference singing voice samples. However, the endeavor to model the intricate nuances of singing voice styles is an arduous task, as singing voices possess a remarkable degree of expressiveness. Moreover, existing SVS methods encounter a decline in the quality of synthesized singing voices in OOD scenarios, as they rest upon the assumption that the target vocal attributes are discernible during the training phase. To overcome these challenges, we propose StyleSinger, the first singing voice synthesis model for zero-shot style transfer of out-of-domain reference singing voice samples. StyleSinger incorporates two critical approaches for enhanced effectiveness: 1) the Residual Style Adaptor (RSA) which employs a residual quantization module to capture diverse style characteristics in singing voices, and 2) the Uncertainty Modeling Layer Normalization (UMLN) to perturb the style attributes within the content representation during the training phase and thus improve the model generalization. Our extensive evaluations in zero-shot style transfer undeniably establish that StyleSinger outperforms baseline models in both audio quality and similarity to the reference singing voice samples. Access to singing voice samples can be found at https://aaronz345.github.io/StyleSingerDemo/.
Authors: Yimu Wang, Shuai Yuan, Bo Xue, Xiangru Jian, Wei Pang, Mushi Wang, Ning Yu
Abstract: Recent progress in video-text retrieval has been driven largely by advancements in model architectures and training strategies. However, the representation learning capabilities of videotext retrieval models remain constrained by lowquality and limited training data annotations. To address this issue, we present a novel ViDeoText Retrieval Paradigm with RElevance-based AugMentation, namely DREAM, which enhances video and text data using large foundation models to learn more generalized features. Specifically, we first adopt a simple augmentation method, which generates self-similar data by randomly duplicating or dropping subwords and frames. In addition, inspired by the recent advancement in visual and language generative models, we propose a more robust augmentation method through textual paraphrasing and video stylization using large language models (LLMs) and visual generative models (VGMs). To further enrich video and text information, we propose a relevance-based augmentation method, where LLMs and VGMs generate and integrate new relevant information into the original data. Leveraging this enriched data, extensive experiments on several video-text retrieval benchmarks demonstrate the superiority of DREAM over existing methods.
Authors: Adam S. Shai, Sarah E. Marzen, Lucas Teixeira, Alexander Gietelink Oldenziel, Paul M. Riechers
Abstract: What computational structure are we building into large language models when we train them on next-token prediction? Here, we present evidence that this structure is given by the meta-dynamics of belief updating over hidden states of the data-generating process. Leveraging the theory of optimal prediction, we anticipate and then find that belief states are linearly represented in the residual stream of transformers, even in cases where the predicted belief state geometry has highly nontrivial fractal structure. We investigate cases where the belief state geometry is represented in the final residual stream or distributed across the residual streams of multiple layers, providing a framework to explain these observations. Furthermore we demonstrate that the inferred belief states contain information about the entire future, beyond the local next-token prediction that the transformers are explicitly trained on. Our work provides a general framework connecting the structure of training data to the geometric structure of activations inside transformers.
Authors: Delong Ran, Jinyuan Liu, Yichen Gong, Jingyi Zheng, Xinlei He, Tianshuo Cong, Anyu Wang
Abstract: Jailbreak attacks induce Large Language Models (LLMs) to generate harmful responses, posing severe misuse threats. Though research on jailbreak attacks and defenses is emerging, there is no consensus on evaluating jailbreaks, i.e., the methods to assess the harmfulness of an LLM's response are varied. Each approach has its own set of strengths and weaknesses, impacting their alignment with human values, as well as the time and financial cost. This diversity challenges researchers in choosing suitable evaluation methods and comparing different attacks and defenses. In this paper, we conduct a comprehensive analysis of jailbreak evaluation methodologies, drawing from nearly 90 jailbreak research published between May 2023 and April 2024. Our study introduces a systematic taxonomy of jailbreak evaluators, offering indepth insights into their strengths and weaknesses, along with the current status of their adaptation. To aid further research, we propose JailbreakEval, a toolkit for evaluating jailbreak attempts. JailbreakEval includes various evaluators out-of-the-box, enabling users to obtain results with a single command or customized evaluation workflows. In summary, we regard JailbreakEval to be a catalyst that simplifies the evaluation process in jailbreak research and fosters an inclusive standard for jailbreak evaluation within the community.
Authors: Chufan Gao, Jathurshan Pradeepkumar, Trisha Das, Shivashankar Thati, Jimeng Sun
Abstract: Background: The global cost of drug discovery and development exceeds $200 billion annually, with clinical trial outcomes playing a critical role in the regulatory approval of new drugs and impacting patient outcomes. Despite their significance, large-scale, high-quality clinical trial outcome data are not readily available to the public, limiting advances in trial outcome predictive modeling. Methods: We introduce the Clinical Trial Outcome (CTO) knowledge base, a fully reproducible, large-scale (around 125K drug and biologics trials), open-source of clinical trial information including large language model (LLM) interpretations of publications, matched trials over phases, sentiment analysis from news, stock prices of trial sponsors, and other trial-related metrics. From this knowledge base, we additionally performed manual annotation of a set of recent clinical trials from 2020-2024. Results: We evaluated the quality of our knowledge base by generating high-quality trial outcome labels that demonstrate strong agreement with previously published expert annotations, achieving an F1 score of 94 for Phase 3 trials and 91 across all phases. Additionally, we benchmarked a suite of standard machine learning models on our manually annotated set, highlighting the distribution shift of recent trials and the need for continuously updated labeling methods. Conclusions: By analyzing CTO's performance on recent trials, we showed a need for recent, high-quality trial outcome labels. We release our knowledge base and labels to the public at https://chufangao.github.io/CTOD, which will also be regularly updated to support ongoing research in clinical trial outcomes, offering insights that could optimize the drug development process.
Authors: Pankayaraj Pathmanathan, Souradip Chakraborty, Xiangyu Liu, Yongyuan Liang, Furong Huang
Abstract: Recent advancements in Reinforcement Learning with Human Feedback (RLHF) have significantly impacted the alignment of Large Language Models (LLMs). The sensitivity of reinforcement learning algorithms such as Proximal Policy Optimization (PPO) has led to new line work on Direct Policy Optimization (DPO), which treats RLHF in a supervised learning framework. The increased practical use of these RLHF methods warrants an analysis of their vulnerabilities. In this work, we investigate the vulnerabilities of DPO to poisoning attacks under different scenarios and compare the effectiveness of preference poisoning, a first of its kind. We comprehensively analyze DPO's vulnerabilities under different types of attacks, i.e., backdoor and non-backdoor attacks, and different poisoning methods across a wide array of language models, i.e., LLama 7B, Mistral 7B, and Gemma 7B. We find that unlike PPO-based methods, which, when it comes to backdoor attacks, require at least 4\% of the data to be poisoned to elicit harmful behavior, we exploit the true vulnerabilities of DPO more simply so we can poison the model with only as much as 0.5\% of the data. We further investigate the potential reasons behind the vulnerability and how well this vulnerability translates into backdoor vs non-backdoor attacks.
Authors: Xisen Jin, Xiang Ren
Abstract: Large Language models (LLMs) suffer from forgetting of upstream data when fine-tuned. Despite efforts on mitigating forgetting, few have investigated whether, and how forgotten upstream examples are dependent on newly learned tasks. Insights on such dependencies enable efficient and targeted mitigation of forgetting. In this paper, we empirically analyze forgetting that occurs in $N$ upstream examples of language modeling or instruction-tuning after fine-tuning LLMs on one of $M$ new tasks, visualized in $M\times N$ matrices. We show that the matrices are often well-approximated with low-rank matrices, indicating the dominance of simple associations between the learned tasks and forgotten upstream examples. Leveraging the analysis, we predict forgetting of upstream examples when fine-tuning on unseen tasks with matrix completion over the empirical associations. This enables fast identification of most forgotten examples without expensive inference on the entire upstream data. The approach, despite simplicity, outperforms prior approaches that learn semantic relationships of learned tasks and upstream examples with LMs for predicting forgetting. We demonstrate the practical utility of our analysis by showing statistically significantly reduced forgetting as we upweight predicted examples for replay at fine-tuning. Project page: https://inklab.usc.edu/lm-forgetting-prediction/
Authors: Zhengzhuo Xu, Bowen Qu, Yiyan Qi, Sinan Du, Chengjin Xu, Chun Yuan, Jian Guo
Abstract: Automatic chart understanding is crucial for content comprehension and document parsing. Multimodal large language models (MLLMs) have demonstrated remarkable capabilities in chart understanding through domain-specific alignment and fine-tuning. However, the application of alignment training within the chart domain is still underexplored. To address this, we propose ChartMoE, which employs the mixture of expert (MoE) architecture to replace the traditional linear projector to bridge the modality gap. Specifically, we train multiple linear connectors through distinct alignment tasks, which are utilized as the foundational initialization parameters for different experts. Additionally, we introduce ChartMoE-Align, a dataset with over 900K chart-table-JSON-code quadruples to conduct three alignment tasks (chart-table/JSON/code). Combined with the vanilla connector, we initialize different experts in four distinct ways and adopt high-quality knowledge learning to further refine the MoE connector and LLM parameters. Extensive experiments demonstrate the effectiveness of the MoE connector and our initialization strategy, e.g., ChartMoE improves the accuracy of the previous state-of-the-art from 80.48% to 84.64% on the ChartQA benchmark.
Authors: Yu Zhang, Changhao Pan, Wenxiang Guo, Ruiqi Li, Zhiyuan Zhu, Jialei Wang, Wenhao Xu, Jingyu Lu, Zhiqing Hong, Chuxin Wang, LiChao Zhang, Jinzheng He, Ziyue Jiang, Yuxin Chen, Chen Yang, Jiecheng Zhou, Xinyu Cheng, Zhou Zhao
Abstract: The scarcity of high-quality and multi-task singing datasets significantly hinders the development of diverse controllable and personalized singing tasks, as existing singing datasets suffer from low quality, limited diversity of languages and singers, absence of multi-technique information and realistic music scores, and poor task suitability. To tackle these problems, we present GTSinger, a large global, multi-technique, free-to-use, high-quality singing corpus with realistic music scores, designed for all singing tasks, along with its benchmarks. Particularly, (1) we collect 80.59 hours of high-quality singing voices, forming the largest recorded singing dataset; (2) 20 professional singers across nine widely spoken languages offer diverse timbres and styles; (3) we provide controlled comparison and phoneme-level annotations of six commonly used singing techniques, helping technique modeling and control; (4) GTSinger offers realistic music scores, assisting real-world musical composition; (5) singing voices are accompanied by manual phoneme-to-audio alignments, global style labels, and 16.16 hours of paired speech for various singing tasks. Moreover, to facilitate the use of GTSinger, we conduct four benchmark experiments: technique-controllable singing voice synthesis, technique recognition, style transfer, and speech-to-singing conversion. The corpus and demos can be found at http://aaronz345.github.io/GTSingerDemo/. We provide the dataset and the code for processing data and conducting benchmarks at https://huggingface.co/datasets/GTSinger/GTSinger and https://github.com/AaronZ345/GTSinger.
URLs: http://aaronz345.github.io/GTSingerDemo/., https://huggingface.co/datasets/GTSinger/GTSinger, https://github.com/AaronZ345/GTSinger.
Authors: Yu Zhang, Ziyue Jiang, Ruiqi Li, Changhao Pan, Jinzheng He, Rongjie Huang, Chuxin Wang, Zhou Zhao
Abstract: Zero-shot singing voice synthesis (SVS) with style transfer and style control aims to generate high-quality singing voices with unseen timbres and styles (including singing method, emotion, rhythm, technique, and pronunciation) from audio and text prompts. However, the multifaceted nature of singing styles poses a significant challenge for effective modeling, transfer, and control. Furthermore, current SVS models often fail to generate singing voices rich in stylistic nuances for unseen singers. To address these challenges, we introduce TCSinger, the first zero-shot SVS model for style transfer across cross-lingual speech and singing styles, along with multi-level style control. Specifically, TCSinger proposes three primary modules: 1) the clustering style encoder employs a clustering vector quantization model to stably condense style information into a compact latent space; 2) the Style and Duration Language Model (S\&D-LM) concurrently predicts style information and phoneme duration, which benefits both; 3) the style adaptive decoder uses a novel mel-style adaptive normalization method to generate singing voices with enhanced details. Experimental results show that TCSinger outperforms all baseline models in synthesis quality, singer similarity, and style controllability across various tasks, including zero-shot style transfer, multi-level style control, cross-lingual style transfer, and speech-to-singing style transfer. Singing voice samples can be accessed at https://aaronz345.github.io/TCSingerDemo/.
Authors: Catherine Yeh, Donghao Ren, Yannick Assogba, Dominik Moritz, Fred Hohman
Abstract: Data augmentation is crucial to make machine learning models more robust and safe. However, augmenting data can be challenging as it requires generating diverse data points to rigorously evaluate model behavior on edge cases and mitigate potential harms. Creating high-quality augmentations that cover these "unknown unknowns" is a time- and creativity-intensive task. In this work, we introduce Amplio, an interactive tool to help practitioners navigate "unknown unknowns" in unstructured text datasets and improve data diversity by systematically identifying empty data spaces to explore. Amplio includes three human-in-the-loop data augmentation techniques: Augment With Concepts, Augment by Interpolation, and Augment with Large Language Model. In a user study with 18 professional red teamers, we demonstrate the utility of our augmentation methods in helping generate high-quality, diverse, and relevant model safety prompts. We find that Amplio enabled red teamers to augment data quickly and creatively, highlighting the transformative potential of interactive augmentation workflows.
Authors: Ethan Ewer, Daewon Chae, Thomas Zeng, Jinkyu Kim, Kangwook Lee
Abstract: Next-token prediction is conventionally done using decoder-only Transformers with causal attention, as this approach allows for efficient reuse of keys and values. What if we were not compute-limited, should we still use decoder-only Transformers? In this work, we introduce Encoder-only Next Token Prediction (ENTP). We explore the differences between ENTP and decoder-only Transformers in expressive power and complexity, highlighting potential advantages of ENTP in settings with unbounded compute. We introduce the $\operatorname{Count3}$ task and show, both theoretically and experimentally, that while ENTP can perform this task easily, a decoder-only Transformer cannot. Finally, we empirically demonstrate the superior performance of ENTP across representative tasks where next-token prediction based Transformers can be evaluated, including addition, in-context learning, and language modeling.
Authors: Jaeyeon Kim, Sehyun Kwon, Joo Young Choi, Jongho Park, Jaewoong Cho, Jason D. Lee, Ernest K. Ryu
Abstract: In-context learning (ICL) describes a language model's ability to generate outputs based on a set of input demonstrations and a subsequent query. To understand this remarkable capability, researchers have studied simplified, stylized models. These studies have consistently observed long loss plateaus, during which models exhibit minimal improvement, followed by a sudden, rapid surge of learning. In this work, we reveal that training on multiple diverse ICL tasks simultaneously shortens the loss plateaus, making each task easier to learn. This finding is surprising as it contradicts the natural intuition that the combined complexity of multiple ICL tasks would lengthen the learning process, not shorten it. Our result suggests that the recent success in large-scale training of language models may be attributed not only to the richness of the data at scale but also to the easier optimization (training) induced by the diversity of natural language training data.
Authors: Bohan Lyu, Yadi Cao, Duncan Watson-Parris, Leon Bergen, Taylor Berg-Kirkpatrick, Rose Yu
Abstract: Large Language Models (LLMs) demonstrate promising capabilities in solving simple scientific problems but, even with domain-specific fine-tuning, often produce hallucinations for complex ones. While integrating LLMs with tools can mitigate this reliability issue, models finetuned on tool usage only often over-rely on them, incurring unnecessary costs from resource-intensive scientific tools even for simpler problems. Inspired by how human experts assess the complexity of the problem before choosing the solutions, we propose a novel two-component fine-tuning method, Adapting While Learning (AWL). In the first component, World Knowledge Learning (WKL), LLMs internalize scientific knowledge by learning from tools-generated solutions. In the second component, Tool Usage Adaptation (TUA), we classify questions as easy or hard based on the WKL-trained model's accuracy, and train it to maintain direct reasoning for simple problems while switching to tools for challenging ones. We validate our method on 6 scientific benchmark datasets in climate science, epidemiology, and mathematics. Compared to the base 8B model, our trained models achieve 28.27% higher answer accuracy and 13.76% better tool usage accuracy, even surpassing state-of-the-art models including GPT-4 and Claude-3.5 on 4 custom-created datasets.
Authors: Vinay Kumar Sankarapu, Chintan Chitroda, Yashwardhan Rathore, Neeraj Kumar Singh, Pratinav Seth
Abstract: The rapid growth of AI has led to more complex deep learning models, often operating as opaque "black boxes" with limited transparency in their decision-making. This lack of interpretability poses challenges, especially in high-stakes applications where understanding model output is crucial. This work highlights the importance of interpretability in fostering trust, accountability, and responsible deployment. To address these challenges, we introduce DLBacktrace, a novel, model-agnostic technique designed to provide clear insights into deep learning model decisions across a wide range of domains and architectures, including MLPs, CNNs, and Transformer-based LLM models. We present a comprehensive overview of DLBacktrace and benchmark its performance against established interpretability methods such as SHAP, LIME, and GradCAM. Our results demonstrate that DLBacktrace effectively enhances understanding of model behavior across diverse tasks. DLBacktrace is compatible with models developed in both PyTorch and TensorFlow, supporting architectures such as BERT, ResNet, U-Net, and custom DNNs for tabular data. The library is open-sourced and available at https://github.com/AryaXAI/DLBacktrace .
Authors: Utkarsh Saxena, Sayeh Sharify, Kaushik Roy, Xin Wang
Abstract: Post-training quantization (PTQ) of large language models (LLMs) holds the promise in reducing the prohibitive computational cost at inference time. Quantization of all weight, activation and key-value (KV) cache tensors to 4-bit without significantly degrading generalizability is challenging, due to the high quantization error caused by extreme outliers in activations. To tackle this problem, we propose ResQ, a PTQ method that pushes further the state-of-the-art. By means of principal component analysis (PCA), it identifies a low-rank subspace (in practice 1/8 of the hidden dimension) in which activation variances are highest, and keep the coefficients within this subspace in high precision, e.g. 8-bit, while quantizing the rest to 4-bit. Within each subspace, invariant random rotation is applied to further suppress outliers. We show that this is a provably optimal mixed precision quantization scheme that minimizes error. With the Llama and Qwen2.5 families of models, we demonstrate that ResQ outperforms recent uniform and mixed precision PTQ methods on a variety of benchmarks, achieving up to 33\% lower perplexity on Wikitext than the next best method SpinQuant, and upto 3\times speedup over 16-bit baseline. Code is available at https://github.com/utkarsh-dmx/project-resq.
Authors: Aditi Singh, Abul Ehtesham, Saket Kumar, Tala Talaei Khoei
Abstract: Large Language Models (LLMs) have revolutionized artificial intelligence (AI) by enabling human like text generation and natural language understanding. However, their reliance on static training data limits their ability to respond to dynamic, real time queries, resulting in outdated or inaccurate outputs. Retrieval Augmented Generation (RAG) has emerged as a solution, enhancing LLMs by integrating real time data retrieval to provide contextually relevant and up-to-date responses. Despite its promise, traditional RAG systems are constrained by static workflows and lack the adaptability required for multistep reasoning and complex task management. Agentic Retrieval-Augmented Generation (Agentic RAG) transcends these limitations by embedding autonomous AI agents into the RAG pipeline. These agents leverage agentic design patterns reflection, planning, tool use, and multiagent collaboration to dynamically manage retrieval strategies, iteratively refine contextual understanding, and adapt workflows to meet complex task requirements. This integration enables Agentic RAG systems to deliver unparalleled flexibility, scalability, and context awareness across diverse applications. This survey provides a comprehensive exploration of Agentic RAG, beginning with its foundational principles and the evolution of RAG paradigms. It presents a detailed taxonomy of Agentic RAG architectures, highlights key applications in industries such as healthcare, finance, and education, and examines practical implementation strategies. Additionally, it addresses challenges in scaling these systems, ensuring ethical decision making, and optimizing performance for real-world applications, while providing detailed insights into frameworks and tools for implementing Agentic RAG.
Authors: Zihan Qiu, Zeyu Huang, Bo Zheng, Kaiyue Wen, Zekun Wang, Rui Men, Ivan Titov, Dayiheng Liu, Jingren Zhou, Junyang Lin
Abstract: This paper revisits the implementation of $\textbf{L}$oad-$\textbf{b}$alancing $\textbf{L}$oss (LBL) when training Mixture-of-Experts (MoEs) models. Specifically, LBL for MoEs is defined as $N_E \sum_{i=1}^{N_E} f_i p_i$, where $N_E$ is the total number of experts, $f_i$ represents the frequency of expert $i$ being selected, and $p_i$ denotes the average gating score of the expert $i$. Existing MoE training frameworks usually employ the parallel training strategy so that $f_i$ and the LBL are calculated within a $\textbf{micro-batch}$ and then averaged across parallel groups. In essence, a micro-batch for training billion-scale LLMs normally contains very few sequences. So, the micro-batch LBL is almost at the sequence level, and the router is pushed to distribute the token evenly within each sequence. Under this strict constraint, even tokens from a domain-specific sequence ($\textit{e.g.}$, code) are uniformly routed to all experts, thereby inhibiting expert specialization. In this work, we propose calculating LBL using a $\textbf{global-batch}$ to loose this constraint. Because a global-batch contains much more diverse sequences than a micro-batch, which will encourage load balance at the corpus level. Specifically, we introduce an extra communication step to synchronize $f_i$ across micro-batches and then use it to calculate the LBL. Through experiments on training MoEs-based LLMs (up to $\textbf{42.8B}$ total parameters and $\textbf{400B}$ tokens), we surprisingly find that the global-batch LBL strategy yields excellent performance gains in both pre-training perplexity and downstream tasks. Our analysis reveals that the global-batch LBL also greatly improves the domain specialization of MoE experts.
Authors: Sergey Berezin, Reza Farahbakhsh, Noel Crespi
Abstract: We present a novel class of jailbreak adversarial attacks on LLMs, termed Task-in-Prompt (TIP) attacks. Our approach embeds sequence-to-sequence tasks (e.g., cipher decoding, riddles, code execution) into the model's prompt to indirectly generate prohibited inputs. To systematically assess the effectiveness of these attacks, we introduce the PHRYGE benchmark. We demonstrate that our techniques successfully circumvent safeguards in six state-of-the-art language models, including GPT-4o and LLaMA 3.2. Our findings highlight critical weaknesses in current LLM safety alignments and underscore the urgent need for more sophisticated defence strategies. Warning: this paper contains examples of unethical inquiries used solely for research purposes.
Authors: Kamer Ali Yuksel, Hassan Sawaf
Abstract: Financial metrics like the Sharpe ratio are pivotal in evaluating investment performance by balancing risk and return. However, traditional metrics often struggle with robustness and generalization, particularly in dynamic and volatile market conditions. This paper introduces AlphaSharpe, a novel framework leveraging large language models (LLMs) to iteratively evolve and optimize financial metrics to discover enhanced risk-return metrics that outperform traditional approaches in robustness and correlation with future performance metrics by employing iterative crossover, mutation, and evaluation. Key contributions of this work include: (1) a novel use of LLMs to generate and refine financial metrics with implicit domain-specific knowledge, (2) a scoring mechanism to ensure that evolved metrics generalize effectively to unseen data, and (3) an empirical demonstration of 3x predictive power for future risk-returns, and 2x portfolio performance. Experimental results in a real-world dataset highlight the superiority of discovered metrics, making them highly relevant to portfolio managers and financial decision-makers. This framework not only addresses the limitations of existing metrics but also showcases the potential of LLMs in advancing financial analytics, paving the way for informed and robust investment strategies.
Authors: Clovis Varangot-Reille, Christophe Bouvard, Antoine Gourru, Mathieu Ciancone, Marion Schaeffer, Fran\c{c}ois Jacquenet
Abstract: Large Language Models (LLM)-based systems, i.e. interconnected elements that include an LLM as a central component (e.g., conversational agents), are typically monolithic static architectures that rely on a single LLM for all user queries. However, they often require different preprocessing strategies, levels of reasoning, or knowledge. Generalist LLMs (e.g. GPT-4) trained on very large multi-topic corpora can perform well in a variety of tasks. They require significant financial, energy, and hardware resources that may not be justified for basic tasks. This implies potentially investing in unnecessary costs for a given query. To overcome this problem, a routing mechanism routes user queries to the most suitable components, such as smaller LLMs or experts in specific topics. This approach may improve response quality while minimising costs. Routing can be expanded to other components of the conversational agent architecture, such as the selection of optimal embedding strategies. This paper explores key considerations for integrating routing into LLM-based systems, focusing on resource management, cost definition, and strategy selection. Our main contributions include a formalisation of the problem, a novel taxonomy of existing approaches emphasising relevance and resource efficiency, and a comparative analysis of these strategies in relation to industry practices. Finally, we identify critical challenges and directions for future research.
Authors: Teng Xiao, Yige Yuan, Zhengyu Chen, Mingxiao Li, Shangsong Liang, Zhaochun Ren, Vasant G Honavar
Abstract: Existing preference optimization objectives for language model alignment require additional hyperparameters that must be extensively tuned to achieve optimal performance, increasing both the complexity and time required for fine-tuning large language models. In this paper, we propose a simple yet effective hyperparameter-free preference optimization algorithm for alignment. We observe that promising performance can be achieved simply by optimizing inverse perplexity, which is calculated as the inverse of the exponentiated average log-likelihood of the chosen and rejected responses in the preference dataset. The resulting simple learning objective, SimPER, is easy to implement and eliminates the need for expensive hyperparameter tuning and a reference model, making it both computationally and memory efficient. Extensive experiments on widely used real-world benchmarks, including MT-Bench, AlpacaEval 2, and 10 key benchmarks of the Open LLM Leaderboard with 5 base models, demonstrate that SimPER consistently and significantly outperforms existing approaches-even without any hyperparameters or a reference model . For example, despite its simplicity, SimPER outperforms state-of-the-art methods by up to 5.7 points on AlpacaEval 2 and achieves the highest average ranking across 10 benchmarks on the Open LLM Leaderboard. The source code for SimPER is publicly available at: https://github.com/tengxiao1/SimPER.
Authors: Zhiteng Li, Mingyuan Xia, Jingyuan Zhang, Zheng Hui, Linghe Kong, Yulun Zhang, Xiaokang Yang
Abstract: Large language models (LLMs) have achieved remarkable success in natural language processing (NLP) tasks, yet their substantial memory requirements present significant challenges for deployment on resource-constrained devices. Singular Value Decomposition (SVD) has emerged as a promising compression technique for LLMs, offering considerable reductions in memory overhead. However, existing SVD-based methods often struggle to effectively mitigate the errors introduced by SVD truncation, leading to a noticeable performance gap when compared to the original models. Furthermore, applying a uniform compression ratio across all transformer layers fails to account for the varying importance of different layers. To address these challenges, we propose AdaSVD, an adaptive SVD-based LLM compression approach. Specifically, AdaSVD introduces adaComp, which adaptively compensates for SVD truncation errors by alternately updating the singular matrices U and V^T. Additionally, AdaSVD introduces adaCR, which adaptively assigns layer-specific compression ratios based on the relative importance of each layer. Extensive experiments across multiple LLM families and evaluation metrics demonstrate that AdaSVD consistently outperforms state-of-the-art (SOTA) SVD-based methods, achieving superior performance with significantly reduced memory requirements. The code and models will be available at https://github.com/ZHITENGLI/AdaSVD.