In this study, we address the challenge of consistently following emotional support strategies in long conversations by large language models (LLMs). We introduce the Strategy-Relevant Attention (SRA) metric, a model-agnostic measure designed to evaluate the effectiveness of LLMs in adhering to strategic prompts in emotional support contexts. By analyzing conversations within the Emotional Support Conversations dataset (ESConv) using LLaMA models, we demonstrate that SRA is significantly correlated with a model's ability to sustain the outlined strategy throughout the interactions. Our findings reveal that the application of SRA-informed prompts leads to enhanced strategic adherence, resulting in conversations that more reliably exhibit the desired emotional support strategies over longer conversations. Furthermore, we contribute a comprehensive, multi-branch synthetic conversation dataset for ESConv, featuring a variety of strategy continuations informed by our optimized prompting method. The code and data are publicly available on our Github.

In light of recent advances in large language models (LLMs), the expectations for the next generation of virtual assistants include enhanced naturalness and adaptability across diverse usage scenarios. However, the creation of high-quality annotated data for Task-Oriented Dialog (TOD) is recognized to be slow and costly. To address these challenges, we introduce Task-Oriented Automatic Dialogs (TOAD), a novel and scalable TOD dataset along with its automatic generation pipeline. The TOAD dataset simulates realistic app context interaction and provide a variety of system response style options. Two aspects of system response styles are considered, verbosity level and users' expression mirroring. We benchmark TOAD on two response generation tasks and the results show that modelling more verbose or responses without user expression mirroring is more challenging.

We introduce API Pack, a multilingual dataset featuring over one million instruction-API call pairs aimed at advancing large language models' API call generation capabilities. Through experiments, we demonstrate API Pack's efficacy in enhancing models for this specialized task while maintaining their overall proficiency at general coding. Fine-tuning CodeLlama-13B on just 20,000 Python instances yields over 10% and 5% higher accuracy than GPT-3.5 and GPT-4 respectively in generating unseen API calls. Scaling to 100k examples improves generalization to new APIs not seen during training. In addition, cross-lingual API call generation is achieved without needing extensive data per language. The dataset, fine-tuned models, and overall code base are publicly available at https://github.com/zguo0525/API-Pack.

A drug molecule is a substance that changes the organism's mental or physical state. Every approved drug has an indication, which refers to the therapeutic use of that drug for treating a particular medical condition. While the Large Language Model (LLM), a generative Artificial Intelligence (AI) technique, has recently demonstrated effectiveness in translating between molecules and their textual descriptions, there remains a gap in research regarding their application in facilitating the translation between drug molecules and indications, or vice versa, which could greatly benefit the drug discovery process. The capability of generating a drug from a given indication would allow for the discovery of drugs targeting specific diseases or targets and ultimately provide patients with better treatments. In this paper, we first propose a new task, which is the translation between drug molecules and corresponding indications, and then test existing LLMs on this new task. Specifically, we consider nine variations of the T5 LLM and evaluate them on two public datasets obtained from ChEMBL and DrugBank. Our experiments show the early results of using LLMs for this task and provide a perspective on the state-of-the-art. We also emphasize the current limitations and discuss future work that has the potential to improve the performance on this task. The creation of molecules from indications, or vice versa, will allow for more efficient targeting of diseases and significantly reduce the cost of drug discovery, with the potential to revolutionize the field of drug discovery in the era of generative AI.

As LLMs become more pervasive across various users and scenarios, identifying potential issues when using these models becomes essential. Examples include bias, inconsistencies, and hallucination. Although auditing the LLM for these problems is desirable, it is far from being easy or solved. An effective method is to probe the LLM using different versions of the same question. This could expose inconsistencies in its knowledge or operation, indicating potential for bias or hallucination. However, to operationalize this auditing method at scale, we need an approach to create those probes reliably and automatically. In this paper we propose an automatic and scalable solution, where one uses a different LLM along with human-in-the-loop. This approach offers verifiability and transparency, while avoiding circular reliance on the same LLMs, and increasing scientific rigor and generalizability. Specifically, we present a novel methodology with two phases of verification using humans: standardized evaluation criteria to verify responses, and a structured prompt template to generate desired probes. Experiments on a set of questions from TruthfulQA dataset show that we can generate a reliable set of probes from one LLM that can be used to audit inconsistencies in a different LLM. The criteria for generating and applying auditing probes is generalizable to various LLMs regardless of the underlying structure or training mechanism.

With the development of LLMs, the security threats of LLMs are getting more and more attention. Numerous jailbreak attacks have been proposed to assess the security defense of LLMs. Current jailbreak attacks primarily utilize scenario camouflage techniques. However their explicitly mention of malicious intent will be easily recognized and defended by LLMs. In this paper, we propose an indirect jailbreak attack approach, Puzzler, which can bypass the LLM's defense strategy and obtain malicious response by implicitly providing LLMs with some clues about the original malicious query. In addition, inspired by the wisdom of "When unable to attack, defend" from Sun Tzu's Art of War, we adopt a defensive stance to gather clues about the original malicious query through LLMs. Extensive experimental results show that Puzzler achieves a query success rate of 96.6% on closed-source LLMs, which is 57.9%-82.7% higher than baselines. Furthermore, when tested against the state-of-the-art jailbreak detection approaches, Puzzler proves to be more effective at evading detection compared to baselines.

The past year has seen rapid acceleration in the development of large language models (LLMs). However, without proper steering and safeguards, LLMs will readily follow malicious instructions, provide unsafe advice, and generate toxic content. We introduce SimpleSafetyTests (SST) as a new test suite for rapidly and systematically identifying such critical safety risks. The test suite comprises 100 test prompts across five harm areas that LLMs, for the vast majority of applications, should refuse to comply with. We test 11 open-access and open-source LLMs and four closed-source LLMs, and find critical safety weaknesses. While some of the models do not give a single unsafe response, most give unsafe responses to more than 20% of the prompts, with over 50% unsafe responses in the extreme. Prepending a safety-emphasising system prompt substantially reduces the occurrence of unsafe responses, but does not completely stop them from happening. Trained annotators labelled every model response to SST (n = 3,000). We use these annotations to evaluate five AI safety filters (which assess whether a models' response is unsafe given a prompt) as a way of automatically evaluating models' performance on SST. The filters' performance varies considerably. There are also differences across the five harm areas, and on the unsafe versus safe responses. The widely-used Perspective API has 72% accuracy and a newly-created zero-shot prompt to OpenAI's GPT-4 performs best with 89% accuracy. Content Warning: This paper contains prompts and responses that relate to child abuse, suicide, self-harm and eating disorders, scams and fraud, illegal items, and physical harm.

Recent advancements in Large Language Models (LLMs) have showcased their remarkable capabilities in text understanding and generation. However, even stronger LLMs are susceptible to acquiring erroneous or obsolete information from the training corpus. Direct secondary fine-tuning with data containing new knowledge may be ineffective in updating knowledge due to the conflict between old and new knowledge. In this paper, we propose a new paradigm for fine-tuning called F-Learning (Forgetting before Learning), which employs parametric arithmetic to facilitate the forgetting of old knowledge and learning of new knowledge. Experimental results on two publicly available datasets demonstrate that our proposed F-Learning can obviously improve the knowledge updating performance of both full fine-tuning and LoRA fine-tuning, simultaneously outperforming the existing baselines in most cases. Moreover, we have also discovered that forgetting old knowledge by subtracting the parameters of LoRA can yield a similar effect to subtracting the parameters of full fine-tuning, and occasionally even surpass it significantly.

Extensive efforts in automated approaches for content moderation have been focused on developing models to identify toxic, offensive, and hateful content with the aim of lightening the load for moderators. Yet, it remains uncertain whether improvements on those tasks have truly addressed moderators' needs in accomplishing their work. In this paper, we surface gaps between past research efforts that have aimed to provide automation for aspects of content moderation and the needs of volunteer content moderators, regarding identifying violations of various moderation rules. To do so, we conduct a model review on Hugging Face to reveal the availability of models to cover various moderation rules and guidelines from three exemplar forums. We further put state-of-the-art LLMs to the test, evaluating how well these models perform in flagging violations of platform rules from one particular forum. Finally, we conduct a user survey study with volunteer moderators to gain insight into their perspectives on useful moderation models. Overall, we observe a non-trivial gap, as missing developed models and LLMs exhibit moderate to low performance on a significant portion of the rules. Moderators' reports provide guides for future work on developing moderation assistant models.

In the Bling Zoo, a tiger wears a giant gold medallion, a monkey sports a bejeweled crown, and a turtle munches on a bowl of diamonds. Unfortunately, this fantastical destination does not exist. 'Bling Zoo' was just one of a series of videos Sora created Thursday when CEO Sam Altman asked his followers on X (formerly Twitter) to submit commands that were generated into movies. The results were so ultra-realistic, they led one observer to comment: 'This one convinced me the future is here and it's going to be OK.' One user requested that Sora create, 'An instructional cooking session for homemade gnocchi hosted by a grandmother social media influencer set in a rustic Tuscan country kitchen with cinematic lighting' This prompt led to the most realistic video containing a human that Altman posted on Thursday.