While large generative artificial intelligence (GenAI) models have achieved significant success, they also raise growing concerns about online information security due to their potential misuse for generating deceptive content. Out-of-context (OOC) multimodal misinformation detection, which often retrieves Web evidence to identify the repurposing of images in false contexts, faces the issue of reasoning over GenAI-polluted evidence to derive accurate predictions. Existing works simulate GenAI-powered pollution at the claim level with stylistic rewriting to conceal linguistic cues, and ignore evidence-level pollution for such information-seeking applications. In this work, we investigate how polluted evidence affects the performance of existing OOC detectors, revealing a performance degradation of more than 9 percentage points. We propose two strategies, cross-modal evidence reranking and cross-modal claim-evidence reasoning, to address the challenges posed by polluted evidence. Extensive experiments on two benchmark datasets show that these strategies can effectively enhance the robustness of existing out-of-context detectors amidst polluted evidence.

This research aims to unravel how large language models (LLMs) iteratively refine token predictions (or, in a general sense, vector predictions). We utilized a logit lens technique to analyze the model's token predictions derived from intermediate representations. Specifically, we focused on how LLMs access and use information from input contexts, and how positioning of relevant information affects the model's token prediction refinement process. Our findings for multi-document question answering task, by varying input context lengths (the number of documents), using GPT-2, revealed that the number of layers between the first layer that the model predicted next tokens correctly and the later layers that the model finalized its correct predictions, as a function of the position of relevant information (i.e., placing the relevant one at the beginning, middle, or end of the input context), has a nearly inverted U shape. We found that the gap between these two layers, on average, diminishes when relevant information is positioned at the beginning or end of the input context, suggesting that the model requires more refinements when processing longer contexts with relevant information situated in the middle, and highlighting which layers are essential for determining the correct output. Our analysis provides insights about how token predictions are distributed across different conditions, and establishes important connections to existing hypotheses and previous findings in AI safety research and development.

Recent advances in natural language processing have leveraged instruction tuning to enhance Large Language Models (LLMs) for table-related tasks. However, previous works train different base models with different training data, lacking an apples-to-apples comparison across the result table LLMs. To address this, we fine-tune base models from the Mistral, OLMo, and Phi families on existing public training datasets. Our replication achieves performance on par with or surpassing existing table LLMs, establishing new state-of-the-art performance on Hitab, a table question-answering dataset. More importantly, through systematic out-of-domain evaluation, we decouple the contributions of training data and the base model, providing insight into their individual impacts. In addition, we assess the effects of table-specific instruction tuning on general-purpose benchmarks, revealing trade-offs between specialization and generalization.

Advances in Large Language Models (LLMs) have sparked interest in their ability to solve Olympiad-level math problems. However, the training and evaluation of these models are constrained by the limited size and quality of available datasets, as creating large-scale data for such advanced problems requires extensive effort from human experts. In addition, current benchmarks are prone to contamination, leading to unreliable evaluations. In this paper, we present an automated pipeline that leverages the rich resources of the Art of Problem Solving (AoPS) forum, which predominantly features Olympiad-level problems and community-driven solutions. Using open-source LLMs, we develop a method to extract question-answer pairs from the forum, resulting in AoPS-Instruct, a dataset of more than 600,000 high-quality QA pairs. Our experiments demonstrate that fine-tuning LLMs on AoPS-Instruct improves their reasoning abilities across various benchmarks. Moreover, we build an automatic pipeline that introduces LiveAoPSBench, an evolving evaluation set with timestamps, derived from the latest forum data, providing a contamination-resistant benchmark for assessing LLM performance. Notably, we observe a significant decline in LLM performance over time, suggesting their success on older examples may stem from pre-training exposure rather than true reasoning ability. Our work presents a scalable approach to creating and maintaining large-scale, high-quality datasets for advanced math reasoning, offering valuable insights into the capabilities and limitations of LLMs in this domain. Our benchmark and code is available at https://github.com/DSL-Lab/aops. Large language models (LLMs) have shown tremendous success in solving various tasks such as code generation (Li et al., 2022), math reasoning (Shao et al., 2024), and commonsense reasoning (Zellers et al., 2019; Achiam et al., 2023), suggesting that current models may show signs of artificial general intelligence (AGI) (Bubeck et al., 2023). Math reasoning is perhaps one of the most challenging tasks for the LLMs, since mathematics is inherently structured, requiring not just recall of facts but also rigorous logical inference, abstraction, and understanding of formal symbolic systems. As such, there have been grand challenges (Selsam et al., 2019) and million-dollar prizes AIMO (2023) established for a model capable of solving Olympiad-level math problems. On the training side, despite significant progress in certain areas, such as geometry, particularly with the assistance of symbolic methods (Trinh et al., 2024), the performance of LLMs remains limited on Olympiad-level problems (He et al., 2024).

To achieve generalized and robust natural-to-formal language conversion (N2F), large language models (LLMs) need to have strong capabilities of decomposition and composition in N2F when faced with an unfamiliar formal language and be able to cope with compositional gaps and counter-intuitive symbolic names. To investigate whether LLMs have this set of basic capabilities in N2F, we propose the DEDC framework. This framework semi-automatically performs sample and task construction, allowing decoupled evaluation of the set of decomposition and composition capabilities of LLMs in N2F. Based on this framework, we evaluate and analyze the most advanced LLMs, and the main findings include that: (1) the LLMs are deficient in both decomposition and composition; (2) the LLMs show a wide coverage of error types that can be attributed to deficiencies in natural language understanding and the learning and use of symbolic systems; (3) compositional gaps and counter-intuitive symbolic names both affect the decomposition and composition of the LLMs. Our work provides a new perspective for investigating the basic capabilities of decomposition and composition of LLMs in N2F. The detailed analysis of deficiencies and attributions can help subsequent improvements of LLMs.

This paper presents a novel methodology of fine-tuning for large language models-dynamic LoRA. Building from the standard Low-Rank Adaptation framework, this methodology further adds dynamic adaptation mechanisms to improve efficiency and performance. The key contribution of dynamic LoRA lies within its adaptive weight allocation mechanism coupled with an input feature-based adaptive strategy. These enhancements allow for a more precise fine-tuning process that is more tailored to specific tasks. Traditional LoRA methods use static adapter settings, not considering the different importance of model layers. In contrast, dynamic LoRA introduces a mechanism that dynamically evaluates the layer's importance during fine-tuning. This evaluation enables the reallocation of adapter parameters to fit the unique demands of each individual task, which leads to better optimization results. Another gain in flexibility arises from the consideration of the input feature distribution, which helps the model generalize better when faced with complicated and diverse datasets. The joint approach boosts not only the performance over each single task but also the generalization ability of the model. The efficiency of the dynamic LoRA was validated in experiments on benchmark datasets, such as GLUE, with surprising results. More specifically, this method achieved 88.1% accuracy with an F1-score of 87.3%. Noticeably, these improvements were made at a slight increase in computational costs: only 0.1% more resources than standard LoRA. This balance between performance and efficiency positions dynamic LoRA as a practical, scalable solution for fine-tuning LLMs, especially in resource-constrained scenarios. To take it a step further, its adaptability makes it a promising foundation for much more advanced applications, including multimodal tasks.

Logical reasoning is a critical component of Large Language Models (LLMs), and substantial research efforts in recent years have aimed to enhance their deductive reasoning capabilities. However, existing deductive reasoning benchmarks, which are crucial for evaluating and advancing LLMs, are inadequate due to their lack of task complexity, presence of prior knowledge as a confounder, and superficial error analysis. To address these deficiencies, we introduce JustLogic, a synthetically generated deductive reasoning benchmark designed for rigorous evaluation of LLMs. JustLogic is (i) highly complex, capable of generating a diverse range of linguistic patterns, vocabulary, and argument structures; (ii) prior knowledge independent, eliminating the advantage of models possessing prior knowledge and ensuring that only deductive reasoning is used to answer questions; and (iii) capable of in-depth error analysis on the heterogeneous effects of reasoning depth and argument form on model accuracy. Our experimental results on JustLogic reveal that most state-of-the-art (SOTA) LLMs perform significantly worse than the human average, demonstrating substantial room for model improvement. All code and data are available at https://github.com/michaelchen-lab/JustLogic

This paper introduces the open-source dataset WanJuanSiLu, designed to provide high-quality training corpora for low-resource languages, thereby advancing the research and development of multilingual models. To achieve this, we have developed a systematic data processing framework tailored for low-resource languages. This framework encompasses key stages such as data extraction, corpus cleaning, content deduplication, security filtering, quality evaluation, and theme classification. Through the implementation of this framework, we have significantly improved both the quality and security of the dataset, while maintaining its linguistic diversity. As of now, data for all five languages have been fully open-sourced.

Real-world tasks such as garment manipulation and table rearrangement demand robots to perform generalizable, highly precise, and long-horizon actions. Although imitation learning has proven to be an effective approach for teaching robots new skills, large amounts of expert demonstration data are still indispensible for these complex tasks, resulting in high sample complexity and costly data collection. To address this, we propose Semantic Keypoint Imitation Learning (SKIL), a framework which automatically obtain semantic keypoints with help of vision foundation models, and forms the descriptor of semantic keypoints that enables effecient imitation learning of complex robotic tasks with significantly lower sample complexity. In real world experiments, SKIL doubles the performance of baseline methods in tasks such as picking a cup or mouse, while demonstrating exceptional robustness to variations in objects, environmental changes, and distractors. For long-horizon tasks like hanging a towel on a rack where previous methods fail completely, SKIL achieves a mean success rate of 70\% with as few as 30 demonstrations. Furthermore, SKIL naturally supports cross-embodiment learning due to its semantic keypoints abstraction, our experiments demonstrate that even human videos bring considerable improvement to the learning performance. All these results demonstrate the great success of SKIL in achieving data-efficint generalizable robotic learning. Visualizations and code are available at: https://skil-robotics.github.io/SKIL-robotics/.

Maintaining consistent model performance across domains is a fundamental challenge in machine learning. While recent work has explored using LLM-generated data for fine-tuning, its impact on cross-domain generalization remains poorly understood. In this paper, we present a systematic analysis revealing that fine-tuning with LLM-generated data not only improves target task performance but also reduces out-of-domain (OOD) degradation compared to fine-tuning with ground truth data. Through analyzing the data sequence in tasks of various domains, we demonstrate that this enhanced OOD robustness stems from a reduced prevalence of high perplexity tokens in LLM-generated sequences. Following this hypothesis we showed that masking high perplexity tokens in ground truth training data also achieves similar OOD preservation comparable to using LLM-generated data. Extensive experiments across diverse model architectures and scales, including Gemma2-2B, Mistral-7B and Llama3-8B, corroborate the consistency of our findings. To the best of our knowledge, this work provides the first mechanistic explanation for the superior OOD robustness conferred by LLM-generated training data, offering valuable insights for developing more robust fine-tuning strategies.