Extending LLMs to New Languages: A Case Study of Llama and Persian Adaptation

Large language models (LLMs) have made great progress in classification and text generation tasks. However, they are mainly trained on English data and often struggle with low-resource languages. In this study, we explore adding a new language, i.e., Persian, to Llama (a model with a limited understanding of Persian) using parameter-efficient fine-tuning. We employ a multi-stage approach involving pretraining on monolingual Persian data, aligning representations through bilingual pretraining and instruction datasets, and instruction-tuning with task-specific datasets. We evaluate the model's performance at each stage on generation and classification tasks. Our findings suggest that incorporating the Persian language, through bilingual data alignment, can enhance classification accuracy for Persian tasks, with no adverse impact and sometimes even improvements on English tasks. Additionally, the results highlight the model's initial strength as a critical factor when working with limited training data, with cross-lingual alignment offering minimal benefits for the low-resource language. Knowledge transfer from English to Persian has a marginal effect, primarily benefiting simple classification tasks.

An Empirical Analysis on Spatial Reasoning Capabilities of Large Multimodal Models

Large Multimodal Models (LMMs) have achieved strong performance across a range of vision and language tasks. However, their spatial reasoning capabilities are under-investigated. In this paper, we construct a novel VQA dataset, Spatial-MM, to comprehensively study LMMs' spatial understanding and reasoning capabilities. Our analyses on object-relationship and multi-hop reasoning reveal several important findings. Firstly, bounding boxes and scene graphs, even synthetic ones, can significantly enhance LMMs' spatial reasoning. Secondly, LMMs struggle more with questions posed from the human perspective than the camera perspective about the image. Thirdly, chain of thought (CoT) prompting does not improve model performance on complex multi-hop questions involving spatial relations. % Moreover, spatial reasoning steps are much less accurate than non-spatial ones across MLLMs. Lastly, our perturbation analysis on GQA-spatial reveals that LMMs are much stronger at basic object detection than complex spatial reasoning. We believe our benchmark dataset and in-depth analyses can spark further research on LMMs spatial reasoning. Spatial-MM benchmark is available at: https://github.com/FatemehShiri/Spatial-MM

Audio Is the Achilles' Heel: Red Teaming Audio Large Multimodal Models

Large Multimodal Models (LMMs) have demonstrated the ability to interact with humans under real-world conditions by combining Large Language Models (LLMs) and modality encoders to align multimodal information (visual and auditory) with text. However, such models raise new safety challenges of whether models that are safety-aligned on text also exhibit consistent safeguards for multimodal inputs. Despite recent safety-alignment research on vision LMMs, the safety of audio LMMs remains under-explored. In this work, we comprehensively red team the safety of five advanced audio LMMs under three settings: (i) harmful questions in both audio and text formats, (ii) harmful questions in text format accompanied by distracting non-speech audio, and (iii) speech-specific jailbreaks. Our results under these settings demonstrate that open-source audio LMMs suffer an average attack success rate of 69.14% on harmful audio questions, and exhibit safety vulnerabilities when distracted with non-speech audio noise. Our speech-specific jailbreaks on Gemini-1.5-Pro achieve an attack success rate of 70.67% on the harmful query benchmark. We provide insights on what could cause these reported safety-misalignments. Warning: this paper contains offensive examples.

The Best of Both Worlds: Bridging Quality and Diversity in Data Selection with Bipartite Graph

The performance of large language models (LLMs) in natural language processing (NLP) tasks is significantly influenced by the quality and diversity of data used for supervised fine-tuning (SFT). Current data selection methods often focus solely on quality or diversity, leading to underperforming models due to suboptimal training data. In this paper, we introduce GraphFilter, a novel method that represents the dataset as a bipartite graph, linking sentences to their constituent n-grams. This representation effectively captures the relationships between sentences and linguistic patterns, facilitating the selection of sentences that enhance n-gram diversity. To balance quality and diversity during selection, we propose a priority function that combines the quality metric with the diversity metric in a multiplicative manner. GraphFilter iteratively selects high-priority sentences, updates the bipartite graph by removing covered n-grams, and re-calculates priorities to reflect the evolving data landscape. We conduct extensive experiments using three model backbones across six widely used benchmarks. The results demonstrate that GraphFilter outperforms all nine baseline approaches, achieving superior model performance and computational efficiency. Our analyses validate the effectiveness of our design choices, examine the subsets selected by GraphFilter and other methods, highlight the importance of instruction diversity, and explore the role of quality and diversity in relation to subset sizes. GraphFilter establishes a new foundation for effective data selection strategies, encouraging further research in data selection for LLMs.

Graph-constrained Reasoning: Faithful Reasoning on Knowledge Graphs with Large Language Models

Large language models (LLMs) have demonstrated impressive reasoning abilities, but they still struggle with faithful reasoning due to knowledge gaps and hallucinations. To address these issues, knowledge graphs (KGs) have been utilized to enhance LLM reasoning through their structured knowledge. However, existing KG-enhanced methods, either retrieval-based or agent-based, encounter difficulties in accurately retrieving knowledge and efficiently traversing KGs at scale. In this work, we introduce graph-constrained reasoning (GCR), a novel framework that bridges structured knowledge in KGs with unstructured reasoning in LLMs. To eliminate hallucinations, GCR ensures faithful KG-grounded reasoning by integrating KG structure into the LLM decoding process through KG-Trie, a trie-based index that encodes KG reasoning paths. KG-Trie constrains the decoding process, allowing LLMs to directly reason on graphs and generate faithful reasoning paths grounded in KGs. Additionally, GCR leverages a lightweight KG-specialized LLM for graph-constrained reasoning alongside a powerful general LLM for inductive reasoning over multiple reasoning paths, resulting in accurate reasoning with zero reasoning hallucination. Extensive experiments on several KGQA benchmarks demonstrate that GCR achieves state-of-the-art performance and exhibits strong zero-shot generalizability to unseen KGs without additional training.

Jigsaw Puzzles: Splitting Harmful Questions to Jailbreak Large Language Models

Large language models (LLMs) have exhibited outstanding performance in engaging with humans and addressing complex questions by leveraging their vast implicit knowledge and robust reasoning capabilities. However, such models are vulnerable to jailbreak attacks, leading to the generation of harmful responses. Despite recent research on single-turn jailbreak strategies to facilitate the development of defence mechanisms, the challenge of revealing vulnerabilities under multi-turn setting remains relatively under-explored. In this work, we propose Jigsaw Puzzles (JSP), a straightforward yet effective multi-turn jailbreak strategy against the advanced LLMs. JSP splits questions into harmless fractions as the input of each turn, and requests LLMs to reconstruct and respond to questions under multi-turn interaction. Our experimental results demonstrate that the proposed JSP jailbreak bypasses original safeguards against explicitly harmful content, achieving an average attack success rate of 93.76% on 189 harmful queries across 5 advanced LLMs (Gemini-1.5-Pro, Llama-3.1-70B, GPT-4, GPT-4o, GPT-4o-mini). Moreover, JSP achieves a state-of-the-art attack success rate of 92% on GPT-4 on the harmful query benchmark, and exhibits strong resistant to defence strategies. Warning: this paper contains offensive examples.

Scalable Frame-based Construction of Sociocultural NormBases for Socially-Aware Dialogues

Sociocultural norms serve as guiding principles for personal conduct in social interactions, emphasizing respect, cooperation, and appropriate behavior, which is able to benefit tasks including conversational information retrieval, contextual information retrieval and retrieval-enhanced machine learning. We propose a scalable approach for constructing a Sociocultural Norm (SCN) Base using Large Language Models (LLMs) for socially aware dialogues. We construct a comprehensive and publicly accessible Chinese Sociocultural NormBase. Our approach utilizes socially aware dialogues, enriched with contextual frames, as the primary data source to constrain the generating process and reduce the hallucinations. This enables extracting of high-quality and nuanced natural-language norm statements, leveraging the pragmatic implications of utterances with respect to the situation. As real dialogue annotated with gold frames are not readily available, we propose using synthetic data. Our empirical results show: (i) the quality of the SCNs derived from synthetic data is comparable to that from real dialogues annotated with gold frames, and (ii) the quality of the SCNs extracted from real data, annotated with either silver (predicted) or gold frames, surpasses that without the frame annotations. We further show the effectiveness of the extracted SCNs in a RAG-based (Retrieval-Augmented Generation) model to reason about multiple downstream dialogue tasks.

From Pre-training Corpora to Large Language Models: What Factors Influence LLM Performance in Causal Discovery Tasks?

Recent advances in artificial intelligence have seen Large Language Models (LLMs) demonstrate notable proficiency in causal discovery tasks. This study explores the factors influencing the performance of LLMs in causal discovery tasks. Utilizing open-source LLMs, we examine how the frequency of causal relations within their pre-training corpora affects their ability to accurately respond to causal discovery queries. Our findings reveal that a higher frequency of causal mentions correlates with better model performance, suggesting that extensive exposure to causal information during training enhances the models' causal discovery capabilities. Additionally, we investigate the impact of context on the validity of causal relations. Our results indicate that LLMs might exhibit divergent predictions for identical causal relations when presented in different contexts. This paper provides the first comprehensive analysis of how different factors contribute to LLM performance in causal discovery tasks.

Towards Probing Speech-Specific Risks in Large Multimodal Models: A Taxonomy, Benchmark, and Insights

Large Multimodal Models (LMMs) have achieved great success recently, demonstrating a strong capability to understand multimodal information and to interact with human users. Despite the progress made, the challenge of detecting high-risk interactions in multimodal settings, and in particular in speech modality, remains largely unexplored. Conventional research on risk for speech modality primarily emphasises the content (e.g., what is captured as transcription). However, in speech-based interactions, paralinguistic cues in audio can significantly alter the intended meaning behind utterances. In this work, we propose a speech-specific risk taxonomy, covering 8 risk categories under hostility (malicious sarcasm and threats), malicious imitation (age, gender, ethnicity), and stereotypical biases (age, gender, ethnicity). Based on the taxonomy, we create a small-scale dataset for evaluating current LMMs capability in detecting these categories of risk. We observe even the latest models remain ineffective to detect various paralinguistic-specific risks in speech (e.g., Gemini 1.5 Pro is performing only slightly above random baseline). Warning: this paper contains biased and offensive examples.

CausalScore: An Automatic Reference-Free Metric for Assessing Response Relevance in Open-Domain Dialogue Systems

Automatically evaluating the quality of responses in open-domain dialogue systems is a challenging but crucial task. Current evaluation metrics often fail to align with human judgments, especially when assessing responses that are grammatically correct. To address this issue, we propose a novel metric, called CausalScore, which assesses the relevance of responses by measuring the causal strength between dialogue histories and responses. The causal strength is estimated by utilizing both unconditional dependence and conditional dependencies from the dialogue history to responses. We compare our metric with the existing competitive metrics in terms of their alignment with human judgements. Our experimental results demonstrate that CausalScore significantly surpasses existing state-of-the-art metrics by aligning better with human judgements. Additionally, we collect a new dialogue dataset CGDIALOG+ with human-annotated causal relations and a set of pairwise human judgements to facilitate the development of future automatic metrics.