Latest Innovations in Recommendation systems with LLMs
How LLMs are changing the recommendation industry as we speak
Technology is rapidly evolving with the emergence of large language models (LLMs), notably transforming industries like recommendation systems. These systems influence a variety of user experiences—from the next song Spotify plays to the posts in your TikTok feed. By anticipating user preferences and integrating smoothly into daily technology, recommendation systems enhance user satisfaction and increase engagement by keeping content relevant and engaging.
However, integrating LLMs directly into production environments poses challenges, primarily due to high latency issues. Consequently, rather than being used directly, LLMs are often employed to enhance other components of the system.
In this article, we'll explore the latest studies on how LLMs are changing recommendation systems in every stage.
A Brief Overview of Modern Recommendation Systems
Modern recommendation systems typically operate in a four-stage loop:
Input Processing and Feature Generation: This stage involves processing data derived from user interactions—both explicit signals such as reviews and ratings, and implicit signals like clicks—into relevant features that can be used for making recommendations.
Candidates Generation: When a user request is received, the system retrieves an initial list of potential recommendations (candidates), such as videos and music, from a large corpus. It employs retrieval techniques like keyword matching to identify the candidates most relevant to the user.
Ranking of Candidates: Each candidate is scored based on a relevance function. Predictive models, such as twin tower neural networks, are used to transform both user and candidate data into embeddings. The relevance is then determined by computing the similarity between these embeddings.
User Interaction and Data collection: The ranked list of recommendations is presented to the user, who interacts with it by clicking on or ignoring certain items. These interactions can than be collected for future processing to refine and improve the recommendation algorithms.
The application of large language models (LLMs) in recommendation systems can be divided into two main categories: discriminative and generative.
Discriminative Cases: Traditional LLMs such as BERT are utilized for classifying or predicting specific outcomes, focusing on categorizing user inputs or predicting user behaviors based on predefined categories.
Generative Cases: More recent developments concentrate on models like GPT, which are capable of generating new content or suggestions in human-readable text.
We will explore the generative applications where advanced models, such as GPT-4 and LLaMA 3.1, are being employed to enhance recommendation systems at every stage.
Where to adapt LLMs in the recommendation systems
As recommendation systems evolve, the strategic integration of large language models (LLMs) is becoming increasingly crucial. This section explores some of the latest research on how LLMs are being adapted at various stages of the recommendation pipeline.
1) Feature Generation
Feature Augmentation (Li et al., 2023)
LLMs can be employed to generate specific features for users or items. For instance, consider the approach used in a product categorization study (refer to the schema described in this paper). The model utilizes structured instruction, which includes components like task description, prompts, input text, candidate labels, output constraints and the desired output. This structured approach helps in generating precise features tailored to specific recommendation tasks. Few-shot learning and fine-tuning strategies can be used to adapt large language models to specific tasks further.
Few-shot Learning: This approach “trains” a LLM on a new task by providing it with only a few examples within the prompt, leveraging its pre-trained knowledge to quickly adapt.
Fine-Tuning: This technique involves additional training of a pre-trained model on a larger, task-specific dataset to enhance its performance by adjusting for the new task
Li et al., 2023: An overview of relevant feature extractions from EcomGPT for diverse E-commerce tasks
Synthetic label generation (Mehrdad et al., 2024)
Synthetic data generation involves creating realistic, artificial data entries for training purposes, as demonstrated in this study The process involves two main steps:
Context Setting: This initial step establishes a domain-specific context to guide the data generation. For example, using a prompt such as "Imagine you are a movie reviewer" helps set the scene for generating data relevant to movie reviews.
Data Generation Prompt: After setting the context, the LLM is given specific instructions about the desired output. This includes details on the style of text (e.g., movie review), the sentiment (positive or negative), and any constraints like word count or specific terminology. This ensures the synthetic data align closely with the requirements of the recommendation system, making them useful for enhancing its accuracy and effectiveness.
2) Candidates Generation
The retrieval is designed to select set of potential candidates for the target user before they can be ranked. There are essentially two types of retrieval mechanisms:
Bag of Words Retrieval: Converts text into vectors of word frequencies and retrieves documents by measuring similarity (e.g., cosine similarity) between these frequency vectors and the user.
Embedding-Based Retrieval: Transforms text into dense semantic embeddings and retrieves documents by comparing the semantic similarity (cosine similarity) between the user’s and documents' embeddings.
LLM augmented retrieval (Wu et al., 2024)
LLM enhances both bag of words and embedding based retrieval by adding synthetic text (essentially augmenting user request) to better represent both the user side and the document side embeddings. This was presented in this study where each user request query is augmented with similar queries to improve retrieval.
3) Ranking
Candidate scoring task
The candidate scoring task involves utilizing a large language model (LLM) as a pointwise function which assigns a score for every target user u, and for every candidate c in candidate set C. The final ranked list of items is generated by sorting these utility scores for each item in the candidate set.
where k1, k2.. kn is list of ranked candidates.
For candidate scoring tasks, there are three primary approaches being used:
Approach 1: Knowledge Distillation (Cui et al., 2024)
This method employs a dual-model strategy where a student model (typically smaller language model) learns from a dataset generated by a more complex teacher model, such as an LLaMA 3.1, often enhanced by integrating user attributes. Again techniques such as few-shot learning and fine-tuning are used for adapting the LLM to specific domain cases, allowing the student model to effectively mimic the teacher’s performance while being more efficient. This approach not only improves model scalability but also retains high performance due to only smaller model used for inference, as detailed in the study available at this link.
Approach 2: Score Generation (Zhiyuli et al., 2023)
Score generation techniques, as discussed in this research, involve the LLM generating scores directly from the given prompts, which can then be used to rank items according to these scores by exposing it through caching mechanisms.
Approach 3: Score prediction from LLM (Wu et al., 2023)
In this method, the candidate scoring task is transformed into a binary question-answering challenge. The process starts with a detailed textual description of the user profile, behaviors, and target item. The model then answers a question aimed at determining user preferences, generating an estimated score or probability. To refine this approach, modifications include replacing the decoder layer of the LLM with a Multi-Layer Perceptron (MLP) for predictions, or using a bidimensional softmax function on the logits of binary answers like "Yes" or "No" to calculate scores (See this study). This simplifies the complex scoring task into a straightforward binary classification problem.
Candidate generation task (Luo et al., 2024)
In item generation tasks, a large language model functions as a generative mechanism, delivering a final ranked list of items with just a single forward pass of the function.
This method primarily depends on the inherent reasoning skills of the LLM to assess user preferences and generate an appropriately ranked list of suggestions. This is what was proposed in this study.
This ranked list can then be used with caching the most frequently occurring users / requests and updating it with batch processing.
4) User Interaction and Data collection (Dong et al., 2024)
LLMs are evolving the user experience in recommendation systems in several significant and noticeable ways. Perhaps the most transformative application is embedding the recommendation process directly into conversational interfaces like in case of conversational search engines like perplexity. As discussed in this study, chatbots powered by LLMs that are fine-tuned on attribute-based conversations—where the bot asks users about their preferences before making suggestions—represent a major shift in how recommendations are delivered. This method leverages LLMs directly, moving beyond merely enhancing existing models.
It introduces a new paradigm where technologies like Retrieval Augmented Generation (RAG) play a crucial role. In the context of recommendation systems, particularly those integrated into conversational agents, RAG can help by pulling relevant and timely information to make more informed suggestions based on the user's current context or query. This can lead to more personalized and accurate recommendations.
Conclusion
Throughout latest studies, we've seen how LLMs are reshaping recommendation systems despite challenges like high latency that prevent them from being the primary inference model. While LLMs currently empower backend improvements in feature generation and predictive accuracy, ongoing advancements like quantization in computational efficiency will eventually mitigate latency issues. This will enable the direct integration of LLMs, dramatically enhancing their responsiveness and deliver seamless, intuitive experiences that improve user experience.
The possibilities are endless !! As always, Happy Learning!
🎉 Good Reads for the weekends
ML:
Career and Leadership:
Awesome overview of the recommendation landscape, with lots of good links and rabbit holes to go down into. Keep up the great work!
Thanks for summarizing so many awesome papers. My best one was the knowledge distillation with student- teacher model. this is where SLM (Small Language Models) will have its strong use cases. They learn fast and run fast :)