Think-like-LSTM: Memory-Augmented Large Language Models via Dynamic Fine-Tuning for Financial Risk Assessment

MINT: A Universal Zero-Shot Predictor for Transaction Data

Banks analyse sequential financial transaction data to perform many tasks, including fraud prevention, credit risk assessment and offer personalization. To improve the predictive accuracy of these tasks, Payments Foundation Models encode transaction sequence data as rich contextual embeddings, which can then be provided to task-specific models as features. However, these Foundation Models are not designed for flexible zero-shot reasoning across novel downstream prediction tasks, limiting their adaptability and utility. Existing LLM-based approaches to zero-shot prediction often fail to fully exploit the predictive signal within transaction data, while relying on costly text serialization or task-specific architectures that scale poorly. To address these limitations, we present the Multimodal Instruction Network for Transactions (MINT), a framework that connects a pretrained transaction sequence encoder to a decoder-only LLM through lightweight embedding injection, transaction-language alignment, and instruction tuning. We find that MINT achieves state-of-the-art predictive question-answering performance in both in-distribution and out-of-distribution questions, while substantially reducing input tokens, latency, and memory consumption compared to text-serialization baselines. Through comprehensive analyses of representations, alignment strategies, training data, and history length, we establish that compact transaction embeddings are a superior approach to transaction representation than text serialization for multimodal reasoning and zero-shot prediction tasks.

Bridging the Semantic Gap: Leveraging LLMs for Hierarchical Interest Evolution in Sequential Recommendation

A Survey of Zero-Shot and Few-Shot Learning with Large Language Models for Financial Sentiment Analysis

Overall, LLMs offer a flexible and annotation-light alternative to traditional fine-tuning, but their successful deployment in finance will depend on robust prompt engineering and solid factual grounding.

TaSC-LLM: A Large Language Model–Enabled Business Intelligence Framework for Topic Analytics in Live-Streaming E-Commerce Systems

In live-streaming e-commerce systems, massive volumes of user-generated danmaku constitute a critical yet underutilized source of business intelligence. However, transforming such unstructured, noisy, and highly context-dependent textual data into structured and actionable knowledge remains a fundamental challenge for enterprise information systems. To address this issue, this study proposes TaSC-LLM, an LLM-enabled topic recognition method for constructing interpretable topic measurements from unstructured user-generated content. The proposed framework integrates topic taxonomy construction and zero-shot classification into a unified semantic reasoning pipeline. Unlike conventional topic modeling or supervised classification approaches, TaSC-LLM leverages chain-of-thought reasoning, multi-stage taxonomy induction, sliding window context modeling, and self-consistency verification to eliminate reliance on predefined label spaces and annotated training data. This design allows the system to dynamically construct and update topic taxonomies while ensuring interpretability, robustness, and cross-scenario adaptability. Empirical evaluation on three large-scale live-streaming e-commerce danmaku datasets shows that TaSC-LLM achieves strong taxonomy coverage, classification accuracy, and agreement with expert annotations. The findings suggest that LLM-based reasoning can help convert unstructured user-generated text into interpretable topic measures for downstream empirical and managerial analysis. While the present evaluation is conducted offline, TaSC-LLM provides a methodological foundation for future business applications that can be further examined under multi-session, multi-platform, and deployment-oriented conditions.

SeqLLM: Augmenting LLMs with Behavioral-Sequence Modeling for High-Stakes Decisions at WeChat Pay

Merchant risk control at large payment platforms screens tens of millions of merchants daily, where false positives harm legitimate merchants and false negatives leave harmful activity undetected. The hardest cases require jointly understanding a merchant's textual profile and long behavioral sequence. Large language models (LLMs) excel at text but cannot natively model such sequences, while adapting them often causes catastrophic forgetting. We present SeqLLM, a framework that adds behavioral-sequence modeling to a pretrained LLM while preserving its language ability. SeqLLM combines three components: a compact discrete vocabulary that represents behavioral events as native tokens; a lightweight projector, trained with a two-stage alignment curriculum, that grounds these tokens in the LLM's semantic space; and prefix-guided capability injection, which acquires sequence-modeling ability through task-prefixed supervised fine-tuning rather than continual pre-training. SeqLLM is deployed at WeChat Pay, screening millions of merchants daily. Against the production DeepSeek-based LLM baseline, it raises screening precision from 92.0% to 97.5%. Its pretrained behavior-token embeddings also improve Precision@Top-0.01% by 26.8 percentage points in a production fraud detector serving billion-scale transaction traffic. Beyond payments, SeqLLM achieves state-of-the-art results on public recommendation benchmarks. On MovieLens and Amazon, it surpasses the strong User-LLM baseline by up to 32% relative Recall@5 while retaining markedly stronger language ability. On RecIF, it improves Pass@32 by 14.2% over the full OneRec-8B pipeline using only one-fifth of its GPU-days.

A Survey of Joint Online-Offline Fine-tuning for Large Language Models

This work advocates for Joint Online-Offline Fine-Tuning as a superior paradigm that breaks the convention of restricting offline data to SFT and online data to RFT, and provides the first comprehensive survey focusing specifically on the synchronization of data provenance.