We present TheraGen, an advanced AI-powered mental health chatbot utilizing the LLaMA 2 7B model. This approach builds upon recent advancements in language models and transformer architectures. TheraGen provides all-day personalized, compassionate mental health care by leveraging a large dataset of 1 million conversational entries, combining anonymized therapy transcripts, online mental health discussions, and psychological literature, including APA resources. Our implementation employs transfer learning, fine-tuning, and advanced training techniques to optimize performance. TheraGen offers a user-friendly interface for seamless interaction, providing empathetic responses and evidence-based coping strategies. Evaluation results demonstrate high user satisfaction rates, with 94% of users reporting improved mental well-being. The system achieved a BLEU score of 0.67 and a ROUGE score of 0.62, indicating strong response accuracy. With an average response time of 1395 milliseconds, TheraGen ensures real-time, efficient support. While not a replacement for professional therapy, TheraGen serves as a valuable complementary tool, significantly improving user well-being and addressing the accessibility gap in mental health treatments. This paper details TheraGen's architecture, training methodology, ethical considerations, and future directions, contributing to the growing field of AI-assisted mental healthcare and offering a scalable solution to the pressing need for mental health support.

The retrieval augmented generation (RAG) system such as Retro has been shown to improve language modeling capabilities and reduce toxicity and hallucinations by retrieving from a database of non-parametric memory containing trillions of entries. We introduce Retro-li that shows retrieval can also help using a small-scale database, but it demands more accurate and better neighbors when searching in a smaller hence sparser non-parametric memory. This can be met by using a proper semantic similarity search. We further propose adding a regularization to the non-parametric memory for the first time: it significantly reduces perplexity when the neighbor search operations are noisy during inference, and it improves generalization when a domain shift occurs. We also show that Retro-li's non-parametric memory can potentially be implemented on analog in-memory computing hardware, exhibiting O(1) search time while causing noise in retrieving neighbors, with minimal (<1%) performance loss. Our code is available at: https://github.com/IBM/Retrieval-Enhanced-Transformer-Little.

A new "empathic voice interface" launched today by Hume AI, a New York–based startup, makes it possible to add a range of emotionally expressive voices, plus an emotionally attuned ear, to large language models from Anthropic, Google, Meta, Mistral, and OpenAI--portending an era when AI helpers may more routinely get all gushy on us. "We specialize in building empathic personalities that speak in ways people would speak, rather than stereotypes of AI assistants," says Hume AI cofounder Alan Cowen, a psychologist who has coauthored a number of research papers on AI and emotion, and who previously worked on emotional technologies at Google and Facebook. WIRED tested Hume's latest voice technology, called EVI 2 and found its output to be similar to that developed by OpenAI for ChatGPT. Later, a real movie star, Scarlett Johansson, claimed OpenAI had ripped off her voice.) Like ChatGPT, Hume is far more emotionally expressive than most conventional voice interfaces. If you tell it that your pet has died, for example, it will adopt a suitable somber and sympathetic tone.

Generative AI models, such as the GPT and Llama series, have significant potential to assist laypeople in answering legal questions. However, little prior work focuses on the data sourcing, inference, and evaluation of these models in the context of laypersons. To this end, we propose a human-centric legal NLP pipeline, covering data sourcing, inference, and evaluation. We introduce and release a dataset, LegalQA, with real and specific legal questions spanning from employment law to criminal law, corresponding answers written by legal experts, and citations for each answer. We develop an automatic evaluation protocol for this dataset, then show that retrieval-augmented generation from only 850 citations in the train set can match or outperform internet-wide retrieval, despite containing 9 orders of magnitude less data. Finally, we propose future directions for open-sourced efforts, which fall behind closed-sourced models.

Humans perceive the world through multisensory integration, blending the information of different modalities to adapt their behavior. Contrastive learning offers an appealing solution for multimodal self-supervised learning. Indeed, by considering each modality as a different view of the same entity, it learns to align features of different modalities in a shared representation space. However, this approach is intrinsically limited as it only learns shared or redundant information between modalities, while multimodal interactions can arise in other ways. In this work, we introduce CoMM, a Contrastive MultiModal learning strategy that enables the communication between modalities in a single multimodal space. Instead of imposing cross- or intra- modality constraints, we propose to align multimodal representations by maximizing the mutual information between augmented versions of these multimodal features. Our theoretical analysis shows that shared, synergistic and unique terms of information naturally emerge from this formulation, allowing us to estimate multimodal interactions beyond redundancy. We test CoMM both in a controlled and in a series of real-world settings: in the former, we demonstrate that CoMM effectively captures redundant, unique and synergistic information between modalities. In the latter, CoMM learns complex multimodal interactions and achieves state-of-the-art results on the six multimodal benchmarks.

Large language models (LLMs) have shown remarkable abilities in different fields, including standard Natural Language Processing (NLP) tasks. To elicit knowledge from LLMs, prompts play a key role, consisting of natural language instructions. Most open and closed source LLMs are trained on available labeled and unlabeled resources--digital content such as text, images, audio, and videos. Hence, these models have better knowledge for high-resourced languages but struggle with low-resourced languages. Since prompts play a crucial role in understanding their capabilities, the language used for prompts remains an important research question. Although there has been significant research in this area, it is still limited, and less has been explored for medium to low-resourced languages. In this study, we investigate different prompting strategies (native vs. non-native) on 11 different NLP tasks associated with 12 different Arabic datasets (9.7K data points). In total, we conducted 197 experiments involving 3 LLMs, 12 datasets, and 3 prompting strategies. Our findings suggest that, on average, the non-native prompt performs the best, followed by mixed and native prompts.

The rapid development of Large Language Models (LLMs) for healthcare applications has spurred calls for holistic evaluation beyond frequently-cited benchmarks like USMLE, to better reflect real-world performance. While real-world assessments are valuable indicators of utility, they often lag behind the pace of LLM evolution, likely rendering findings obsolete upon deployment. This temporal disconnect necessitates a comprehensive upfront evaluation that can guide model selection for specific clinical applications. We introduce MEDIC, a framework assessing LLMs across five critical dimensions of clinical competence: medical reasoning, ethics and bias, data and language understanding, in-context learning, and clinical safety. MEDIC features a novel cross-examination framework quantifying LLM performance across areas like coverage and hallucination detection, without requiring reference outputs. We apply MEDIC to evaluate LLMs on medical question-answering, safety, summarization, note generation, and other tasks. Our results show performance disparities across model sizes, baseline vs medically finetuned models, and have implications on model selection for applications requiring specific model strengths, such as low hallucination or lower cost of inference. MEDIC's multifaceted evaluation reveals these performance trade-offs, bridging the gap between theoretical capabilities and practical implementation in healthcare settings, ensuring that the most promising models are identified and adapted for diverse healthcare applications.

Second language (L2) learners can improve their pronunciation by imitating golden speech, especially when the speech that aligns with their respective speech characteristics. This study explores the hypothesis that learner-specific golden speech generated with zero-shot text-to-speech (ZS-TTS) techniques can be harnessed as an effective metric for measuring the pronunciation proficiency of L2 learners. Building on this exploration, the contributions of this study are at least two-fold: 1) design and development of a systematic framework for assessing the ability of a synthesis model to generate golden speech, and 2) in-depth investigations of the effectiveness of using golden speech in automatic pronunciation assessment (APA). Comprehensive experiments conducted on the L2-ARCTIC and Speechocean762 benchmark datasets suggest that our proposed modeling can yield significant performance improvements with respect to various assessment metrics in relation to some prior arts. To our knowledge, this study is the first to explore the role of golden speech in both ZS-TTS and APA, offering a promising regime for computer-assisted pronunciation training (CAPT).

If GPT-4V can accurately digitize North Carolina, underwent urban renewal, a national hand-written documents through carefully crafted program aimed at modernizing "blighted" areas prompts, it could become a valuable tool for nonexperts (Lee et al. 2017). This process, mostly impacting in transcribing historical documents on a African-American neighborhoods, displaced families, large scale. Alternatively, if it falls short, it is still businesses, and organizations for economic crucial to understand and discuss the implications and infrastructure development.

In this paper we explore evaluation of LLM capabilities. We present measurements of GPT-4 performance on several deterministic tasks; each task involves a basic calculation and takes as input parameter some element drawn from a large well-defined population (e.g., count elements in a list, multiply two k-digit numbers, etc). We examine several conditions per-task and perform enough trials so that statistically significant differences can be detected. This allows us to investigate the sensitivity of task-accuracy both to query phrasing and input parameter population. We find that seemingly trivial modifications in the task-prompt or input population can yield differences far larger than can be explained by sampling effects. For example, performance on a simple list-counting task varies with query-phrasing and list-length, but also with list composition (i.e., the thing-to-be-counted) and object frequency (e.g., success when an element accounts for $\approx$ 50\% of a list is different from when it accounts for $\approx$ 70\% etc). We conclude that efforts to quantify LLM capabilities easily succumb to the language-as-fixed-effect fallacy, where experimental observations are improperly generalized beyond what the data supports. A consequence appears to be that intuitions that have been formed based on interactions with humans form a very unreliable guide as to which input modifications should ``make no difference'' to LLM performance.