Trojan backdoors can be injected into large language models at various stages, including pretraining, fine-tuning, and in-context learning, posing a significant threat to the model's alignment. Due to the nature of causal language modeling, detecting these triggers is challenging given the vast search space. In this study, we propose a multistage framework for detecting Trojan triggers in large language models consisting of token filtration, trigger identification, and trigger verification. We discuss existing trigger identification methods and propose two variants of a black-box trigger inversion method that rely on output logits, utilizing beam search and greedy decoding respectively. We show that the verification stage is critical in the process and propose semantic-preserving prompts and special perturbations to differentiate between actual Trojan triggers and other adversarial strings that display similar characteristics. The evaluation of our approach on the TrojAI and RLHF poisoned model datasets demonstrates promising results.

Scaling data and model size has been proven effective for boosting the performance of large language models. In addition to training-time scaling, recent studies have revealed that increasing test-time computational resources can further improve performance. In this work, we introduce Aggregation Fine-Tuning (AFT), a supervised finetuning paradigm where the model learns to synthesize multiple draft responses, referred to as proposals, into a single, refined answer, termed aggregation. At inference time, a propose-and-aggregate strategy further boosts performance by iteratively generating proposals and aggregating them. Empirical evaluations on benchmark datasets show that AFT-trained models substantially outperform standard SFT. Notably, an AFT model, fine-tuned from Llama3.1-8B-Base with only 64k data, achieves a 41.3% LC win rate on AlpacaEval 2, surpassing significantly larger LLMs such as Llama3.1-405B-Instruct and GPT4. By combining sequential refinement and parallel sampling, the propose-and-aggregate framework scales inference-time computation in a flexible manner. Overall, These findings position AFT as a promising approach to unlocking additional capabilities of LLMs without resorting to increasing data volume or model size.

Vision-Language Models (VLMs) have demonstrated remarkable capabilities in understanding multimodal inputs and have been widely integrated into Retrieval-Augmented Generation (RAG) based conversational systems. While current VLM-powered chatbots can provide textual source references in their responses, they exhibit significant limitations in referencing contextually relevant images during conversations. In this paper, we introduce Contextual Image Reference -- the ability to appropriately reference relevant images from retrieval documents based on conversation context -- and systematically investigate VLMs' capability in this aspect. We conduct the first evaluation for contextual image referencing, comprising a dedicated testing dataset and evaluation metrics. Furthermore, we propose ImageRef-VL, a method that significantly enhances open-source VLMs' image referencing capabilities through instruction fine-tuning on a large-scale, manually curated multimodal conversation dataset. Experimental results demonstrate that ImageRef-VL not only outperforms proprietary models but also achieves an 88% performance improvement over state-of-the-art open-source VLMs in contextual image referencing tasks. Our code is available at https://github.com/bytedance/ImageRef-VL.

In real-world scientific discovery, human beings always make use of the accumulated prior knowledge with imagination pick select one or a few most promising hypotheses from large and noisy data analysis results. In this study, we introduce a new type of graph structure, the text-numeric graph (TNG), which is defined as graph entities and associations have both text-attributed information and numeric information. The TNG is an ideal data structure model for novel scientific discovery via graph reasoning because it integrates human-understandable textual annotations or prior knowledge, with numeric values that represent the observed or activation levels of graph entities or associations in different samples. Together both the textual information and numeric values determine the importance of graph entities and associations in graph reasoning for novel scientific knowledge discovery. We further propose integrating large language models (LLMs) and graph neural networks (GNNs) to analyze the TNGs for graph understanding and reasoning. To demonstrate the utility, we generated the text-omic(numeric) signaling graphs (TOSG), as one type of TNGs, in which all graphs have the same entities, associations and annotations, but have sample-specific entity numeric (omic) values using single cell RNAseq (scRNAseq) datasets of different diseases. We proposed joint LLM-GNN models for key entity mining and signaling pathway mining on the TOSGs. The evaluation results showed the LLM-GNN and TNGs models significantly improve classification accuracy and network inference. In conclusion, the TNGs and joint LLM-GNN models are important approaches for scientific discovery.

Societal and personal values are transmitted to younger generations through interaction and exposure. Traditionally, children and adolescents learned values from parents, educators, or peers. Nowadays, social platforms serve as a significant channel through which youth (and adults) consume information, as the main medium of entertainment, and possibly the medium through which they learn different values. In this paper we extract implicit values from TikTok movies uploaded by online influencers targeting children and adolescents. We curated a dataset of hundreds of TikTok movies and annotated them according to the Schwartz Theory of Personal Values. We then experimented with an array of Masked and Large language model, exploring how values can be detected. Specifically, we considered two pipelines -- direct extraction of values from video and a 2-step approach in which videos are first converted to elaborated scripts and then values are extracted. Achieving state-of-the-art results, we find that the 2-step approach performs significantly better than the direct approach and that using a trainable Masked Language Model as a second step significantly outperforms a few-shot application of a number of Large Language Models. We further discuss the impact of fine-tuning and compare the performance of the different models on identification of values present or contradicted in the TikTok. Finally, we share the first values-annotated dataset of TikTok videos. Our results pave the way to further research on influence and value transmission in video-based social platforms.

Restless Multi-Armed Bandits (RMABs) have been successfully applied to resource allocation problems in a variety of settings, including public health. With the rapid development of powerful large language models (LLMs), they are increasingly used to design reward functions to better match human preferences. Recent work has shown that LLMs can be used to tailor automated allocation decisions to community needs using language prompts. However, this has been studied primarily for English prompts and with a focus on task performance only. This can be an issue since grassroots workers, especially in developing countries like India, prefer to work in local languages, some of which are low-resource. Further, given the nature of the problem, biases along population groups unintended by the user are also undesirable. In this work, we study the effects on both task performance and fairness when the DLM algorithm, a recent work on using LLMs to design reward functions for RMABs, is prompted with non-English language commands. Specifically, we run the model on a synthetic environment for various prompts translated into multiple languages. The prompts themselves vary in complexity. Our results show that the LLM-proposed reward functions are significantly better when prompted in English compared to other languages. We also find that the exact phrasing of the prompt impacts task performance. Further, as prompt complexity increases, performance worsens for all languages; however, it is more robust with English prompts than with lower-resource languages. On the fairness side, we find that low-resource languages and more complex prompts are both highly likely to create unfairness along unintended dimensions.

The task of converting natural language questions (NLQs) into executable SQL queries, known as text-to-SQL, has gained significant interest in recent years, as it enables non-technical users to interact with relational databases. Many benchmarks, such as SPIDER and WikiSQL, have contributed to the development of new models and the evaluation of their performance. In addition, other datasets, like SEDE and BIRD, have introduced more challenges and complexities to better map real-world scenarios. However, these datasets primarily focus on high-resource languages such as English and Chinese. In this work, we introduce Dialect2SQL, the first large-scale, cross-domain text-to-SQL dataset in an Arabic dialect. It consists of 9,428 NLQ-SQL pairs across 69 databases in various domains. Along with SQL-related challenges such as long schemas, dirty values, and complex queries, our dataset also incorporates the complexities of the Moroccan dialect, which is known for its diverse source languages, numerous borrowed words, and unique expressions. This demonstrates that our dataset will be a valuable contribution to both the text-to-SQL community and the development of resources for low-resource languages.

Recent years have witnessed extensive efforts to enhance Large Language Models (LLMs) across various domains, alongside growing attention to their ethical implications. However, a critical challenge remains largely overlooked: LLMs must balance between rejecting harmful requests for safety and accommodating legitimate ones for utility. This paper presents a Direct Preference Optimization (DPO) based alignment framework that achieves better overall performance by addressing this ethical-utility trade-off, using chemical domain applications as a proof-of-concept. Our alignment pipeline starts with a GPT-assisted three-phase data generation scheme, in which we create LibraChemQA, a chemical question-answering dataset comprising 31.6k triplet instances. By incorporating an innovative balanced seed in the data generation process, our framework systematically considers both legitimate and illegitimate requests. The framework also introduces a rephrasing mechanism for efficient data augmentation that enhances the model's chemical comprehension. We further develop a novel hybrid evaluation scheme with LLM judges for precise assessment of both safety and utility. Experimental results demonstrate our model's substantial improvements in overall performance where both safety and utility are considered - our resulting model, LibraChem, outperforms leading LLMs including Claude-3, GPT-4o, and LLaMA-3 by margins of 13.44%, 7.16%, and 7.10% respectively on our released benchmark.

The task of policy compliance detection (PCD) is to determine if a scenario is in compliance with respect to a set of written policies. In a conversational setting, the results of PCD can indicate if clarifying questions must be asked to determine compliance status. Existing approaches usually claim to have reasoning capabilities that are latent or require a large amount of annotated data. In this work, we propose logical decomposition for policy compliance (LDPC): a neuro-symbolic framework to detect policy compliance using large language models (LLMs) in a few-shot setting. By selecting only a few exemplars alongside recently developed prompting techniques, we demonstrate that our approach soundly reasons about policy compliance conversations by extracting sub-questions to be answered, assigning truth values from contextual information, and explicitly producing a set of logic statements from the given policies. The formulation of explicit logic graphs can in turn help answer PCDrelated questions with increased transparency and explainability. We apply this approach to the popular PCD and conversational machine reading benchmark, ShARC, and show competitive performance with no task-specific finetuning. We also leverage the inherently interpretable architecture of LDPC to understand where errors occur, revealing ambiguities in the ShARC dataset and highlighting the challenges involved with reasoning for conversational question answering.

Introduction. Objective Structured Clinical Examinations (OSCEs) are widely used to assess medical students' communication skills, but scoring interview-based assessments is time-consuming and potentially subject to human bias. This study explored the potential of large language models (LLMs) to automate OSCE evaluations using the Master Interview Rating Scale (MIRS). Methods. We compared the performance of four state-of-the-art LLMs (GPT-4o, Claude 3.5, Llama 3.1, and Gemini 1.5 Pro) in evaluating OSCE transcripts across all 28 items of the MIRS under the conditions of zero-shot, chain-of-thought (CoT), few-shot, and multi-step prompting. The models were benchmarked against a dataset of 10 OSCE cases with 174 expert consensus scores available. Model performance was measured using three accuracy metrics (exact, off-by-one, thresholded). Results. Averaging across all MIRS items and OSCE cases, LLMs performed with low exact accuracy (0.27 to 0.44), and moderate to high off-by-one accuracy (0.67 to 0.87) and thresholded accuracy (0.75 to 0.88). A zero temperature parameter ensured high intra-rater reliability ($\alpha = 0.98$ for GPT-4o). CoT, few-shot, and multi-step techniques proved valuable when tailored to specific assessment items. The performance was consistent across MIRS items independent of encounter phases and communication domains. Conclusion. We demonstrated the feasibility of AI-assisted OSCE evaluation and provided benchmarking of multiple LLMs across multiple prompt techniques. Our work provides a baseline performance assessment for LLMs that lays a foundation for future research in automated assessment of clinical communication skills.