We study how to subvert language models from following the rules. We model rule-following as inference in propositional Horn logic, a mathematical system in which rules have the form "if $P$ and $Q$, then $R$" for some propositions $P$, $Q$, and $R$. We prove that although transformers can faithfully abide by such rules, maliciously crafted prompts can nevertheless mislead even theoretically constructed models. Empirically, we find that attacks on our theoretical models mirror popular attacks on large language models. Our work suggests that studying smaller theoretical models can help understand the behavior of large language models in rule-based settings like logical reasoning and jailbreak attacks.

Despite the significant progress of large language models (LLMs) in various tasks, they often produce factual errors due to their limited internal knowledge. Retrieval-Augmented Generation (RAG), which enhances LLMs with external knowledge sources, offers a promising solution. However, these methods can be misled by irrelevant paragraphs in retrieved documents. Due to the inherent uncertainty in LLM generation, inputting the entire document may introduce off-topic information, causing the model to deviate from the central topic and affecting the relevance of the generated content. To address these issues, we propose the Retrieve-Plan-Generation (RPG) framework. RPG generates plan tokens to guide subsequent generation in the plan stage. In the answer stage, the model selects relevant fine-grained paragraphs based on the plan and uses them for further answer generation. This plan-answer process is repeated iteratively until completion, enhancing generation relevance by focusing on specific topics. To implement this framework efficiently, we utilize a simple but effective multi-task prompt-tuning method, enabling the existing LLMs to handle both planning and answering. We comprehensively compare RPG with baselines across 5 knowledge-intensive generation tasks, demonstrating the effectiveness of our approach.

The recent success of Large Language Models (LLMs) has sparked concerns about their potential to spread misinformation. As a result, there is a pressing need for tools to identify ``fake arguments'' generated by such models. To create these tools, examples of texts generated by LLMs are needed. This paper introduces a methodology to obtain good, bad and ugly arguments from argumentative essays produced by ChatGPT, OpenAI's LLM. We then describe a novel dataset containing a set of diverse arguments, ArGPT. We assess the effectiveness of our dataset and establish baselines for several argumentation-related tasks. Finally, we show that the artificially generated data relates well to human argumentation and thus is useful as a tool to train and test systems for the defined tasks.

Evaluation of multilingual Large Language Models (LLMs) is challenging due to a variety of factors -- the lack of benchmarks with sufficient linguistic diversity, contamination of popular benchmarks into LLM pre-training data and the lack of local, cultural nuances in translated benchmarks. In this work, we study human and LLM-based evaluation in a multilingual, multi-cultural setting. We evaluate 30 models across 10 Indic languages by conducting 90K human evaluations and 30K LLM-based evaluations and find that models such as GPT-4o and Llama-3 70B consistently perform best for most Indic languages. We build leaderboards for two evaluation settings - pairwise comparison and direct assessment and analyse the agreement between humans and LLMs. We find that humans and LLMs agree fairly well in the pairwise setting but the agreement drops for direct assessment evaluation especially for languages such as Bengali and Odia. We also check for various biases in human and LLM-based evaluation and find evidence of self-bias in the GPT-based evaluator. Our work presents a significant step towards scaling up multilingual evaluation of LLMs.

Large language models (LLMs) are being applied to time series tasks, particularly time series forecasting. However, are language models actually useful for time series? After a series of ablation studies on three recent and popular LLM-based time series forecasting methods, we find that removing the LLM component or replacing it with a basic attention layer does not degrade the forecasting results -- in most cases the results even improved. We also find that despite their significant computational cost, pretrained LLMs do no better than models trained from scratch, do not represent the sequential dependencies in time series, and do not assist in few-shot settings. Additionally, we explore time series encoders and reveal that patching and attention structures perform similarly to state-of-the-art LLM-based forecasters.

The proliferation of misinformation and harmful narratives in online discourse has underscored the critical need for effective Counter Narrative (CN) generation techniques. However, existing automatic evaluation methods often lack interpretability and fail to capture the nuanced relationship between generated CNs and human perception. Aiming to achieve a higher correlation with human judgments, this paper proposes a novel approach to asses generated CNs that consists on the use of a Large Language Model (LLM) as a evaluator. By comparing generated CNs pairwise in a tournament-style format, we establish a model ranking pipeline that achieves a correlation of $0.88$ with human preference. As an additional contribution, we leverage LLMs as zero-shot (ZS) CN generators and conduct a comparative analysis of chat, instruct, and base models, exploring their respective strengths and limitations. Through meticulous evaluation, including fine-tuning experiments, we elucidate the differences in performance and responsiveness to domain-specific data. We conclude that chat-aligned models in ZS are the best option for carrying out the task, provided they do not refuse to generate an answer due to security concerns.

Warning: Contents of this paper may be upsetting. Public attitudes towards key societal issues, expressed on online media, are of immense value in policy and reform efforts, yet challenging to understand at scale. We study one such social issue: homelessness in the U.S., by leveraging the remarkable capabilities of large language models to assist social work experts in analyzing millions of posts from Twitter. We introduce a framing typology: Online Attitudes Towards Homelessness (OATH) Frames: nine hierarchical frames capturing critiques, responses and perceptions. We release annotations with varying degrees of assistance from language models, with immense benefits in scaling: 6.5x speedup in annotation time while only incurring a 3 point F1 reduction in performance with respect to the domain experts. Our experiments demonstrate the value of modeling OATH-Frames over existing sentiment and toxicity classifiers. Our large-scale analysis with predicted OATH-Frames on 2.4M posts on homelessness reveal key trends in attitudes across states, time periods and vulnerable populations, enabling new insights on the issue. Our work provides a general framework to understand nuanced public attitudes at scale, on issues beyond homelessness.

Digital storytelling, essential in entertainment, education, and marketing, faces challenges in production scalability and flexibility. The StoryAgent framework, introduced in this paper, utilizes Large Language Models and generative tools to automate and refine digital storytelling. Employing a top-down story drafting and bottom-up asset generation approach, StoryAgent tackles key issues such as manual intervention, interactive scene orchestration, and narrative consistency. This framework enables efficient production of interactive and consistent narratives across multiple modalities, democratizing content creation and enhancing engagement. Our results demonstrate the framework's capability to produce coherent digital stories without reference videos, marking a significant advancement in automated digital storytelling.

Large language models (LLMs) have shown remarkable performance on code generation tasks. A recent application of LLMs for code generation is iterative code repair, where a model fixes an incorrect program by rationalizing about errors and generating a new program. However, code repair is primarily studied on high-resource languages like Python, and the framework's efficacy is under-explored on low-resource languages. To apply code repair for low-resource languages, we propose Distilling Low-Resource Repairs (DistiLRR), an approach that transfers the reasoning and code generation ability from a teacher model to a student model. Our results show that DistiLRR consistently outperforms baselines on low-resource languages, but has similar performance on high-resource languages. To investigate this behavior, we perform a further analysis and find that the correlation between rationale quality and code correctness is weaker than previously perceived. We hypothesize this weakness is magnified in low-resource settings where base models lack deep knowledge of a programming language, leading to wavering benefits of code repair between high-resource and low-resource languages.

While large language models (LLMs) have showcased impressive capabilities, they struggle with addressing legal queries due to the intricate complexities and specialized expertise required in the legal field. In this paper, we introduce InternLM-Law, a specialized LLM tailored for addressing diverse legal queries related to Chinese laws, spanning from responding to standard legal questions (e.g., legal exercises in textbooks) to analyzing complex real-world legal situations. We meticulously construct a dataset in the Chinese legal domain, encompassing over 1 million queries, and implement a data filtering and processing pipeline to ensure its diversity and quality. Our training approach involves a novel two-stage process: initially fine-tuning LLMs on both legal-specific and general-purpose content to equip the models with broad knowledge, followed by exclusive fine-tuning on high-quality legal data to enhance structured output generation. InternLM-Law achieves the highest average performance on LawBench, outperforming state-of-the-art models, including GPT-4, on 13 out of 20 subtasks. We make InternLM-Law and our dataset publicly available to facilitate future research in applying LLMs within the legal domain.