Yonsei University
AI Insights
- Replay buffer: A memory buffer in SPRING that stores a subset of the user's interactions, used to update the parametric adapter. [3]
- Drift score: A measure of how much a user's interaction history has changed over time, used to select samples for the replay buffer. [3]
- ROUGE scores: Metrics used to evaluate the quality of generated text, including ROUGE-1 and ROUGE-L. [3]
- The paper presents SPRING, a framework for adapting large language models (LLMs) to individual users based on their interaction history. [2]
- SPRING: A framework for adapting large language models to individual users based on their interaction history. [1]
Abstract
Personalizing Large Language Models typically relies on static retrieval or one-time adaptation, assuming user preferences remain invariant over time. However, real-world interactions are dynamic, where user interests continuously evolve, posing a challenge for models to adapt to preference drift without catastrophic forgetting. Standard continual learning approaches often struggle in this context, as they indiscriminately update on noisy interaction streams, failing to distinguish genuine preference shifts from transient contexts. To address this, we introduce SPRInG, a novel semi-parametric framework designed for effective continual personalization. During training, SPRInG employs drift-driven selective adaptation, which utilizes a likelihood-based scoring function to identify high-novelty interactions. This allows the model to selectively update the user-specific adapter on drift signals while preserving hard-to-learn residuals in a replay buffer. During inference, we apply strict relevance gating and fuse parametric knowledge with retrieved history via logit interpolation. Experiments on the long-form personalized generation benchmark demonstrate that SPRInG outperforms existing baselines, validating its robustness for real-world continual personalization.
Personalizing Large Language Models typically relies on static retrieval or one-time adaptation, assuming user preferences remain invariant over time. However, real-world interactions are dynamic, where user interests continuously evolve, posing a challenge for models to adapt to preference drift without catastrophic forgetting. Standard continual learning approaches often struggle in this context, as they indiscriminately update on noisy interaction streams, failing to distinguish genuine preference shifts from transient contexts. To address this, we introduce SPRInG, a novel semi-parametric framework designed for effective continual personalization. During training, SPRInG employs drift-driven selective adaptation, which utilizes a likelihood-based scoring function to identify high-novelty interactions. This allows the model to selectively update the user-specific adapter on drift signals while preserving hard-to-learn residuals in a replay buffer. During inference, we apply strict relevance gating and fuse parametric knowledge with retrieved history via logit interpolation. Experiments on the long-form personalized generation benchmark demonstrate that SPRInG outperforms existing baselines, validating its robustness for real-world continual personalization.
Why we are recommending this paper?
Due to your Interest in Personalization
This paper directly addresses personalization, a core interest, by proposing a method for continual adaptation of LLMs to evolving user preferences. The focus on retrieval-interpolated generation aligns perfectly with the user’s interest in personalization platforms and data-driven CRM strategies.
Due to your Interest in Personalization
This paper directly addresses personalization, a core interest, by proposing a method for continual adaptation of LLMs to evolving user preferences. The focus on retrieval-interpolated generation aligns perfectly with the user’s interest in personalization platforms and data-driven CRM strategies.
Fudan University
AI Insights
- On objective tasks, personalized information may not always improve model performance and can even lead to factual errors or logical biases. [3]
- On subjective personalized tasks, personalized information is crucial for improving model performance. [3]
- The type of personalized information used can significantly impact model performance. [3]
- Aligned personas tend to perform better than unaligned personas on both objective and subjective tasks. [3]
- To maximize the benefits of personalized information, it is essential to carefully design and implement the personaDual framework, taking into account the specific requirements and constraints of each application domain. [3]
- Personalized information can have a double-edged effect on the performance of large language models (LLMs). [2]
Abstract
As users increasingly expect LLMs to align with their preferences, personalized information becomes valuable. However, personalized information can be a double-edged sword: it can improve interaction but may compromise objectivity and factual correctness, especially when it is misaligned with the question. To alleviate this problem, we propose PersonaDual, a framework that supports both general-purpose objective reasoning and personalized reasoning in a single model, and adaptively switches modes based on context. PersonaDual is first trained with SFT to learn two reasoning patterns, and then further optimized via reinforcement learning with our proposed DualGRPO to improve mode selection. Experiments on objective and personalized benchmarks show that PersonaDual preserves the benefits of personalization while reducing interference, achieving near interference-free performance and better leveraging helpful personalized signals to improve objective problem-solving.
As users increasingly expect LLMs to align with their preferences, personalized information becomes valuable. However, personalized information can be a double-edged sword: it can improve interaction but may compromise objectivity and factual correctness, especially when it is misaligned with the question. To alleviate this problem, we propose PersonaDual, a framework that supports both general-purpose objective reasoning and personalized reasoning in a single model, and adaptively switches modes based on context. PersonaDual is first trained with SFT to learn two reasoning patterns, and then further optimized via reinforcement learning with our proposed DualGRPO to improve mode selection. Experiments on objective and personalized benchmarks show that PersonaDual preserves the benefits of personalization while reducing interference, achieving near interference-free performance and better leveraging helpful personalized signals to improve objective problem-solving.
Why we are recommending this paper?
Due to your Interest in Personalization
Given the user’s interest in personalization, this paper’s exploration of balancing personalization with objectivity is highly relevant. The approach of adaptive reasoning offers a valuable perspective on managing the potential downsides of personalization within a CRM context.
Due to your Interest in Personalization
Given the user’s interest in personalization, this paper’s exploration of balancing personalization with objectivity is highly relevant. The approach of adaptive reasoning offers a valuable perspective on managing the potential downsides of personalization within a CRM context.
ToulouseINP
AI Insights
- The convergence rate of Stochastic Gradient Descent (SGD) depends on the learning rate and the variance of the gradient estimate. [3]
- A constant step size leads to a linear convergence rate for strongly convex functions, but not for convex or nonconvex functions. [3]
- The condition number of the Hessian matrix affects the convergence rate and stability of SGD. [3]
- Large eigenvalues in some directions can lead to slower progress and increased noise amplification. [3]
- First-order methods like SGD struggle with ill-conditioned problems due to their sensitivity to small eigenvalues. [3]
- Strongly convex function: A function that has a positive definite Hessian matrix everywhere. [3]
- Condition number: The ratio of the largest to smallest eigenvalue of a matrix, indicating how close it is to being singular. [3]
- For convex functions, SGD converges sublinearly with a fixed step size, while for nonconvex functions, it achieves stationarity bounds. [2]
Abstract
Optimization is central to both modern machine learning (ML) and scientific machine learning (SciML), yet the structure of the underlying optimization problems differs substantially across these domains. Classical ML typically relies on stochastic, sample-separable objectives that favor first-order and adaptive gradient methods. In contrast, SciML often involves physics-informed or operator-constrained formulations in which differential operators induce global coupling, stiffness, and strong anisotropy in the loss landscape. As a result, optimization behavior in SciML is governed by the spectral properties of the underlying physical models rather than by data statistics, frequently limiting the effectiveness of standard stochastic methods and motivating deterministic or curvature-aware approaches. This document provides a unified introduction to optimization methods in ML and SciML, emphasizing how problem structure shapes algorithmic choices. We review first- and second-order optimization techniques in both deterministic and stochastic settings, discuss their adaptation to physics-constrained and data-driven SciML models, and illustrate practical strategies through tutorial examples, while highlighting open research directions at the interface of scientific computing and scientific machine learning.
Optimization is central to both modern machine learning (ML) and scientific machine learning (SciML), yet the structure of the underlying optimization problems differs substantially across these domains. Classical ML typically relies on stochastic, sample-separable objectives that favor first-order and adaptive gradient methods. In contrast, SciML often involves physics-informed or operator-constrained formulations in which differential operators induce global coupling, stiffness, and strong anisotropy in the loss landscape. As a result, optimization behavior in SciML is governed by the spectral properties of the underlying physical models rather than by data statistics, frequently limiting the effectiveness of standard stochastic methods and motivating deterministic or curvature-aware approaches. This document provides a unified introduction to optimization methods in ML and SciML, emphasizing how problem structure shapes algorithmic choices. We review first- and second-order optimization techniques in both deterministic and stochastic settings, discuss their adaptation to physics-constrained and data-driven SciML models, and illustrate practical strategies through tutorial examples, while highlighting open research directions at the interface of scientific computing and scientific machine learning.
Why we are recommending this paper?
Due to your Interest in CRM Optimization
This paper’s focus on optimization methods for training SciML models aligns with the user’s interest in data-driven CRM and MLOps. The core concept of optimization is fundamental to building effective personalization models.
Due to your Interest in CRM Optimization
This paper’s focus on optimization methods for training SciML models aligns with the user’s interest in data-driven CRM and MLOps. The core concept of optimization is fundamental to building effective personalization models.
National University of Singapore
AI Insights
- The paper proposes a novel framework called LEAN-LLM-OPT for automatic optimization model formulation using Large Language Models (LLMs). [2]
- The framework involves two workflows: one for generating the optimization model and another for translating it into program codes that can be input to optimization solvers. [1]
Abstract
Large-scale optimization is a key backbone of modern business decision-making. However, building these models is often labor-intensive and time-consuming. We address this by proposing LEAN-LLM-OPT, a LightwEight AgeNtic workflow construction framework for LLM-assisted large-scale OPTimization auto-formulation. LEAN-LLM-OPT takes as input a problem description together with associated datasets and orchestrates a team of LLM agents to produce an optimization formulation. Specifically, upon receiving a query, two upstream LLM agents dynamically construct a workflow that specifies, step-by-step, how optimization models for similar problems can be formulated. A downstream LLM agent then follows this workflow to generate the final output. Leveraging LLMs' text-processing capabilities and common modeling practices, the workflow decomposes the modeling task into a sequence of structured sub-tasks and offloads mechanical data-handling operations to auxiliary tools. This design alleviates the downstream agent's burden related to planning and data handling, allowing it to focus on the most challenging components that cannot be readily standardized. Extensive simulations show that LEAN-LLM-OPT, instantiated with GPT-4.1 and the open source gpt-oss-20B, achieves strong performance on large-scale optimization modeling tasks and is competitive with state-of-the-art approaches. In addition, in a Singapore Airlines choice-based revenue management use case, LEAN-LLM-OPT demonstrates practical value by achieving leading performance across a range of scenarios. Along the way, we introduce Large-Scale-OR and Air-NRM, the first comprehensive benchmarks for large-scale optimization auto-formulation. The code and data of this work is available at https://github.com/CoraLiang01/lean-llm-opt.
Large-scale optimization is a key backbone of modern business decision-making. However, building these models is often labor-intensive and time-consuming. We address this by proposing LEAN-LLM-OPT, a LightwEight AgeNtic workflow construction framework for LLM-assisted large-scale OPTimization auto-formulation. LEAN-LLM-OPT takes as input a problem description together with associated datasets and orchestrates a team of LLM agents to produce an optimization formulation. Specifically, upon receiving a query, two upstream LLM agents dynamically construct a workflow that specifies, step-by-step, how optimization models for similar problems can be formulated. A downstream LLM agent then follows this workflow to generate the final output. Leveraging LLMs' text-processing capabilities and common modeling practices, the workflow decomposes the modeling task into a sequence of structured sub-tasks and offloads mechanical data-handling operations to auxiliary tools. This design alleviates the downstream agent's burden related to planning and data handling, allowing it to focus on the most challenging components that cannot be readily standardized. Extensive simulations show that LEAN-LLM-OPT, instantiated with GPT-4.1 and the open source gpt-oss-20B, achieves strong performance on large-scale optimization modeling tasks and is competitive with state-of-the-art approaches. In addition, in a Singapore Airlines choice-based revenue management use case, LEAN-LLM-OPT demonstrates practical value by achieving leading performance across a range of scenarios. Along the way, we introduce Large-Scale-OR and Air-NRM, the first comprehensive benchmarks for large-scale optimization auto-formulation. The code and data of this work is available at https://github.com/CoraLiang01/lean-llm-opt.
Why we are recommending this paper?
Due to your Interest in CRM Optimization
This paper’s exploration of LLMs for large-scale optimization is a strong fit, given the user's interest in CRM optimization and automated model formulation. The approach offers a potentially powerful tool for streamlining optimization processes.
Due to your Interest in CRM Optimization
This paper’s exploration of LLMs for large-scale optimization is a strong fit, given the user's interest in CRM optimization and automated model formulation. The approach offers a potentially powerful tool for streamlining optimization processes.
PingCAP
AI Insights
- HDC generation involves extracting representative entities for each database to facilitate efficient data exploration across multiple databases. [3]
- It also includes a self-refinement chain to correct errors in generated SQL statements. [3]
- The system demonstrates its capabilities through two real-world scenarios: the Financial dataset and the Bird dataset, showcasing its ability to provide insights and facilitate user-system interaction. [3]
- HDC: Hierarchical Data Context - a summary of the data that includes a description, keywords, table information, and more. [3]
- TiChart: Chart Selection - a component that selects the most suitable chart type to present analysis results by visualization. [3]
- Exploration Efficiency: The ability of the system to efficiently explore data across multiple databases. [3]
- TiInsight is a SQL-based automated cross-domain exploratory data analysis system that utilizes large language models to facilitate user-system interaction and provide powerful hierarchical data context (HDC) generation, text-to-SQL (TiSQL), chart selection (TiChart), and exploration efficiency. [2]
- TiSQL is a schema filtering framework based on the map-reduce paradigm that filters tables and columns using clarified questions and cosine similarity. [1]
Abstract
The SQL-based exploratory data analysis has garnered significant attention within the data analysis community. The emergence of large language models (LLMs) has facilitated the paradigm shift from manual to automated data exploration. However, existing methods generally lack the ability for cross-domain analysis, and the exploration of LLMs capabilities remains insufficient. This paper presents TiInsight, an SQL-based automated cross-domain exploratory data analysis system. First, TiInsight offers a user-friendly GUI enabling users to explore data using natural language queries. Second, TiInsight offers a robust cross-domain exploratory data analysis pipeline: hierarchical data context (i.e., HDC) generation, question clarification and decomposition, text-to-SQL (i.e., TiSQL), and data visualization (i.e., TiChart). Third, we have implemented and deployed TiInsight in the production environment of PingCAP and demonstrated its capabilities using representative datasets. The demo video is available at https://youtu.be/JzYFyYd-emI.
The SQL-based exploratory data analysis has garnered significant attention within the data analysis community. The emergence of large language models (LLMs) has facilitated the paradigm shift from manual to automated data exploration. However, existing methods generally lack the ability for cross-domain analysis, and the exploration of LLMs capabilities remains insufficient. This paper presents TiInsight, an SQL-based automated cross-domain exploratory data analysis system. First, TiInsight offers a user-friendly GUI enabling users to explore data using natural language queries. Second, TiInsight offers a robust cross-domain exploratory data analysis pipeline: hierarchical data context (i.e., HDC) generation, question clarification and decomposition, text-to-SQL (i.e., TiSQL), and data visualization (i.e., TiChart). Third, we have implemented and deployed TiInsight in the production environment of PingCAP and demonstrated its capabilities using representative datasets. The demo video is available at https://youtu.be/JzYFyYd-emI.
Why we are recommending this paper?
Due to your Interest in Data Driven CRM
The paper’s use of LLMs for automated data exploration directly supports the user’s interest in data-driven CRM and leveraging data for insights. This approach could be valuable for understanding customer behavior and patterns.
Due to your Interest in Data Driven CRM
The paper’s use of LLMs for automated data exploration directly supports the user’s interest in data-driven CRM and leveraging data for insights. This approach could be valuable for understanding customer behavior and patterns.
University of Technology Sydney
AI Insights
- The second stage clusters consumers based on their representative load sets (RLS) obtained in the first stage, allowing for more accurate segmentation. [3]
- Smart Meter: An intelligent device that measures and records energy consumption in real-time. [3]
- The proposed CROCS framework is a two-stage clustering approach designed to avoid the limitations of existing methodologies. [2]
Abstract
With grid operators confronting rising uncertainty from renewable integration and a broader push toward electrification, Demand-Side Management (DSM) -- particularly Demand Response (DR) -- has attracted significant attention as a cost-effective mechanism for balancing modern electricity systems. Unprecedented volumes of consumption data from a continuing global deployment of smart meters enable consumer segmentation based on real usage behaviours, promising to inform the design of more effective DSM and DR programs. However, existing clustering-based segmentation methods insufficiently reflect the behavioural diversity of consumers, often relying on rigid temporal alignment, and faltering in the presence of anomalies, missing data, or large-scale deployments. To address these challenges, we propose a novel two-stage clustering framework -- Clustered Representations Optimising Consumer Segmentation (CROCS). In the first stage, each consumer's daily load profiles are clustered independently to form a Representative Load Set (RLS), providing a compact summary of their typical diurnal consumption behaviours. In the second stage, consumers are clustered using the Weighted Sum of Minimum Distances (WSMD), a novel set-to-set measure that compares RLSs by accounting for both the prevalence and similarity of those behaviours. Finally, community detection on the WSMD-induced graph reveals higher-order prototypes that embody the shared diurnal behaviours defining consumer groups, enhancing the interpretability of the resulting clusters. Extensive experiments on both synthetic and real Australian smart meter datasets demonstrate that CROCS captures intra-consumer variability, uncovers both synchronous and asynchronous behavioural similarities, and remains robust to anomalies and missing data, while scaling efficiently through natural parallelisation. These results...
With grid operators confronting rising uncertainty from renewable integration and a broader push toward electrification, Demand-Side Management (DSM) -- particularly Demand Response (DR) -- has attracted significant attention as a cost-effective mechanism for balancing modern electricity systems. Unprecedented volumes of consumption data from a continuing global deployment of smart meters enable consumer segmentation based on real usage behaviours, promising to inform the design of more effective DSM and DR programs. However, existing clustering-based segmentation methods insufficiently reflect the behavioural diversity of consumers, often relying on rigid temporal alignment, and faltering in the presence of anomalies, missing data, or large-scale deployments. To address these challenges, we propose a novel two-stage clustering framework -- Clustered Representations Optimising Consumer Segmentation (CROCS). In the first stage, each consumer's daily load profiles are clustered independently to form a Representative Load Set (RLS), providing a compact summary of their typical diurnal consumption behaviours. In the second stage, consumers are clustered using the Weighted Sum of Minimum Distances (WSMD), a novel set-to-set measure that compares RLSs by accounting for both the prevalence and similarity of those behaviours. Finally, community detection on the WSMD-induced graph reveals higher-order prototypes that embody the shared diurnal behaviours defining consumer groups, enhancing the interpretability of the resulting clusters. Extensive experiments on both synthetic and real Australian smart meter datasets demonstrate that CROCS captures intra-consumer variability, uncovers both synchronous and asynchronous behavioural similarities, and remains robust to anomalies and missing data, while scaling efficiently through natural parallelisation. These results...
Why we are recommending this paper?
Due to your Interest in Data Driven CRM
Due to your Interest in Data Driven CRM
Interests not found
We did not find any papers that match the below interests.
Try other terms also consider if the content exists in arxiv.org.
- Personalization Platform
- Email Marketing
- MLOps
You can edit or add more interests any time.
💬 Help Shape Our Pricing
We're exploring pricing options to make this project sustainable. Take 3 minutes to share what you'd be willing to pay (if anything). Your input guides our future investment.
Help us improve your experience!
This project is on its early stages your feedback can be pivotal on the future of the project.
Let us know what you think about this week's papers and suggestions!
Give Feedback