Uncategorized

Large Language Models (LLMs) are increasingly employed for various domains, with use cases including creative writing, chatbots, and semantic search. Many of these applications are inherently subjective and require generations catering to different demographics, cultural and societal norms, or individual preferences. Through their large-scale training, current language models are exposed to diverse data that allows…

One effective method to improve the reasoning skills of LLMs is to employ supervised fine-tuning (SFT) with chain-of-thought (CoT) annotations. However, this approach has limitations in terms of generalization because it heavily depends on the provided CoT data. In scenarios like math problem-solving, each question in the training data typically has only one annotated reasoning…

The inherent capabilities of pretrained large language models are notable, yet achieving desired behaviors often requires additional adaptation. When dealing with models whose weights are kept private, the challenge intensifies, rendering tuning either excessively costly or outright impossible. As a result, striking the right balance between customization and resource efficiency remains a persistent concern in…

Large Language Models (LLMs) have exhibited remarkable prowess across various natural language processing tasks. However, applying them to Information Retrieval (IR) tasks remains a challenge due to the scarcity of IR-specific concepts in natural language. Addressing this, the idea of instruction tuning has emerged as a pivotal method to elevate LLMs’ capabilities and control. While…

Stable AI has recently released a new state-of-the-art model, Stable-Code-3B, designed for code completion in various programming languages with multiple additional capabilities. The model is a follow-up on the Stable Code Alpha 3B. It is trained on 1.3 trillion tokens including both natural language data and code data in 18 programming languages and codes.  Compared…

Large Language Models (LLMs) have emerged as a transformative force in artificial intelligence, offering remarkable capabilities in processing and generating language-based responses. LLMs are being used in many applications, from automated customer service to generating creative content. However, one critical challenge surfacing with using LLMs is their ability to utilize external tools to accomplish intricate…

The Natural Language Generation (NLG) field stands at the intersection of linguistics and artificial intelligence. It focuses on the creation of human-like text by machines. Recent advancements in Large Language Models (LLMs) have revolutionized NLG, significantly enhancing the ability of systems to generate coherent and contextually relevant text. This evolving field necessitates robust evaluation methodologies…

Mixture-of-Experts (MoE) is an architecture based on the “divide and conquer” principle to solve complex tasks. Multiple individual machine learning (ML) models (called experts) work individually based on their specializations to provide the most optimal results. To better understand their use cases, Mistral AI recently released Mixtral, an open-source high-quality MoE model that outperformed or…

The emergence of large language models (LLMs) like GPT, Claude, Gemini, LLaMA, Mistral, etc., has greatly accelerated recent advances in natural language processing (NLP). Instruction tweaking is a well-known approach to training LLMs. This method allows LLMs to improve their pre-trained representations to follow human instructions using large-scale, well-formatted instruction data. However, these tasks are…

In the constantly evolving field of machine learning, particularly in semantic segmentation, the accurate estimation and validation of uncertainty have become increasingly vital. Despite numerous studies claiming advances in uncertainty methods, there remains a disconnection between theoretical development and practical application. Fundamental questions linger, such as whether it is feasible to separate data-related (aleatoric) and…